     ______   ___    ___
    /\  _  \ /\_ \  /\_ \
    \ \ \L\ \\//\ \ \//\ \      __     __   _ __   ___ 
     \ \  __ \ \ \ \  \ \ \   /'__`\ /'_ `\/\`'__\/ __`\
      \ \ \/\ \ \_\ \_ \_\ \_/\  __//\ \L\ \ \ \//\ \L\ \
       \ \_\ \_\/\____\/\____\ \____\ \____ \ \_\\ \____/
        \/_/\/_/\/____/\/____/\/____/\/___L\ \/_/ \/___/
                                       /\____/
                                       \_/__/     Version 4.0.3


                A game programming library.

             By Shawn Hargreaves, Apr 19, 2003.

                See the AUTHORS file for a
               complete list of contributors.



#include <std_disclaimer.h>

   "I do not accept responsibility for any effects, adverse or otherwise, 
    that this code may have on you, your computer, your sanity, your dog, 
    and anything else that you can think of. Use it at your own risk."



=======================================
============ Using Allegro ============
=======================================

See readme.txt for a general introduction, copyright details, and 
information about how to install Allegro and link your program with it.

int install_allegro(int system_id, int *errno_ptr, int (*atexit_ptr)());
   Initialises the Allegro library. You must call either this or 
   allegro_init() before doing anything other than using the Unicode 
   routines. If you want to use a text mode other than UTF-8, you can set
   it with set_uformat() before you call this. The available system ID codes 
   will vary from one platform to another, but you will almost always want 
   to pass SYSTEM_AUTODETECT. Alternatively, SYSTEM_NONE installs a stripped 
   down version of Allegro that won't even try to touch your hardware or do 
   anything platform specific: this can be useful for situations where you 
   only want to manipulate memory bitmaps, such as the text mode datafile 
   tools or the Windows GDI interfacing functions. The errno_ptr and 
   atexit_ptr parameters should point to the errno variable and atexit 
   function from your libc: these are required because when Allegro is 
   linked as a DLL, it doesn't have direct access to your local libc data.
   atexit_ptr may be NULL, in which case it is your responsibility to call
   allegro_exit manually. Currently this function always returns zero. If no
   system driver can be used, the program will abort.

int allegro_init();
   Macro which initialises the Allegro library. This is the same thing as
   calling install_allegro(SYSTEM_AUTODETECT, &errno, atexit).

void allegro_exit();
   Closes down the Allegro system. This includes returning the system to 
   text mode and removing whatever mouse, keyboard, and timer routines have 
   been installed. You don't normally need to bother making an explicit call 
   to this function, because allegro_init() installs it as an atexit() 
   routine so it will be called automatically when your program exits.

Macro END_OF_MAIN()
   In order to maintain cross-platform compatibility, you have to put this
   macro at the very end of your main function. This macro uses some `magic'
   to mangle your main procedure on platforms that need it like Windows
   or Linux. On the other platforms this macro compiles to nothing, so you
   don't have to #ifdef around it. Example:

      int main(void)
      {
         allegro_init();
         /* more stuff goes here */
         ...
         return 0;
      }
      END_OF_MAIN()


extern char allegro_id[];
   Text string containing a date and version number for the library, in case 
   you want to display these somewhere.

extern char allegro_error[ALLEGRO_ERROR_SIZE];
   Text string used by set_gfx_mode() and install_sound() to report error 
   messages. If they fail and you want to tell the user why, this is the 
   place to look for a description of the problem.

extern int os_type;
   Set by allegro_init() to one of the values:

      OSTYPE_UNKNOWN    - unknown, or regular MSDOS
      OSTYPE_WIN3       - Windows 3.1 or earlier
      OSTYPE_WIN95      - Windows 95
      OSTYPE_WIN98      - Windows 98
      OSTYPE_WINME      - Windows ME
      OSTYPE_WINNT      - Windows NT
      OSTYPE_WIN2000    - Windows 2000
      OSTYPE_WINXP      - Windows XP
      OSTYPE_OS2        - OS/2
      OSTYPE_WARP       - OS/2 Warp 3
      OSTYPE_DOSEMU     - Linux DOSEMU
      OSTYPE_OPENDOS    - Caldera OpenDOS
      OSTYPE_LINUX      - Linux
      OSTYPE_SUNOS      - SunOS/Solaris
      OSTYPE_FREEBSD    - FreeBSD
      OSTYPE_NETBSD     - NetBSD
      OSTYPE_IRIX       - IRIX
      OSTYPE_QNX        - QNX
      OSTYPE_UNIX       - Unknown Unix variant
      OSTYPE_BEOS       - BeOS
      OSTYPE_MACOS      - MacOS

extern int os_version;
extern int os_revision;
   The major and minor version of the Operating System currently running.
   Set by allegro_init(). If Allegro for some reason was not able to
   retrieve the version of the Operating System, os_version and
   os_revision will be set to -1. For example: Under Win98 SE (v4.10.2222)
   os_version will be set to 4 and os_revision to 10.

extern int os_multitasking;
   Set by allegro_init() to either TRUE or FALSE depending on whether your
   Operating System is multitasking or not.

void allegro_message(const char *msg, ...);
   Outputs a message, using a printf() format string. This function must 
   only be used when you aren't in graphics mode, eg. before calling 
   set_gfx_mode(), or after a set_gfx_mode(GFX_TEXT). On platforms that have 
   a text console (DOS and Unix) it will print the string to that console,
   attempting to work around codepage differences by reducing any accented
   characters to 7 bit ASCII approximations, and on platforms featuring a
   windowing system it will bring up a GUI message box.

void set_window_title(const char *name);
   On platforms that are capable of it, this routine alters the window title 
   for your Allegro program. Note that Allegro cannot set the window title 
   when running in a DOS box under Windows.

int set_window_close_button(int enable);
   On platforms that are capable of it, this routine disables or enables the 
   window close button for your Allegro program. You can call it before the 
   window is created if you wish. If the close button is successfully 
   disabled, this function returns zero.

   On platforms where the close button either does not exist or cannot be 
   disabled, this function returns -1. If this happens, you may wish 
   to use set_window_close_hook() to handle the close event yourself.

   When enabling the close button, the function will return the same value 
   for your platform as when disabling. That means it will return non-zero 
   if the button cannot be disabled, even though you are not trying to 
   disable it.

   Note that Allegro cannot manipulate the close button of a DOS box in 
   Windows.

void set_window_close_hook(void (*proc)());
   On platforms that have a close button, this routine installs a hook 
   function to handle the close event. In other words, when the user clicks 
   the close button on your program's window, the function you specify here 
   will be called.

   This function should not generally attempt to exit the program or save 
   any data itself. The function could be called at any time, and there is 
   usually a risk of conflict with the main thread of the program. Instead, 
   you should set a flag during this function, and test it on a regular 
   basis in the main loop of the program.

   Pass NULL to this function to restore the close button's default 
   functionality. On Windows and BeOS, the following message will appear:

      Warning: forcing program shutdown may lead to data loss and unexpected 
      results. It is preferable to use the exit command inside the window.

      Proceed anyway?

      [Yes] [No]

   This message will be translated into your selected language if a 
   translation is available in language.dat (see get_config_text()).

   If the user clicks [Yes], the program will exit immediately in the same 
   style as Ctrl+Alt+End (see three_finger_flag).

   In other operating systems, the program will exit immediately without 
   prompting the user.

   Note that Allegro cannot intercept the close button of a DOS box in 
   Windows.

int desktop_color_depth();
   On platforms that can run Allegro programs in a window on an existing 
   desktop, returns the currently selected desktop color depth (your program 
   is likely to run faster if it uses this same depth). On platforms where 
   this information is not available or does not apply, returns zero.

int get_desktop_resolution(int *width, int *height);
   On platforms that can run Allegro programs in a window on an existing
   desktop, this retrieves the current desktop resolution (e.g. you may want
   to call this function before creating a large window because, with some
   windowed drivers, a window cannot be created if it is larger than the
   desktop). Returns zero on success, or a negative number if this
   information is not available or does not apply, in which case the values
   stored in width and height are unspecified.

void yield_timeslice();
   On systems that support this, gives up the rest of the current scheduler 
   timeslice. Also known as the "play nice with multitasking" option.

void check_cpu();
   Detects the CPU type, setting the following global variables. You don't 
   normally need to call this, because allegro_init() will do it for you.

extern char cpu_vendor[];
   Contains the CPU vendor name, if known (empty string on non-Intel 
   platforms).

extern int cpu_family;
   Contains the Intel CPU type, where applicable: 3=386, 4=486, 5=Pentium, 
   6=PPro, etc.

extern int cpu_model;
   Contains the Intel CPU submodel, where applicable. On a 486 
   (cpu_family=4), zero or one indicates a DX chip, 2 an SX, 3 a 487 (SX) or 
   486 DX, 4 an SL, 5 an SX2, 7 a DX2 write-back enhanced, 8 a DX4 or DX4 
   overdrive, 14 a Cyrix, and 15 is unknown. On a Pentium chip 
   (cpu_family=5), 1 indicates a Pentium (510\66, 567\66), 2 is a Pentium 
   P54C, 3 is a Pentium overdrive processor, 5 is a Pentium overdrive for 
   IntelDX4, 14 is a Cyrix, and 15 is unknown.

extern int cpu_capabilities;
   Contains CPU flags indicating what features are available on the current 
   CPU. The flags can be any combination of these:


   CPU_ID       - Indicates that the "cpuid" instruction is available. If this 
                  is set, then all Allegro CPU variables are 100% reliable,
                  otherwise there may be some mistakes.
   CPU_FPU      - An x87 FPU is available.
   CPU_MMX      - Intel MMX  instruction set is available.
   CPU_MMXPLUS  - Intel MMX+ instruction set is available.
   CPU_SSE      - Intel SSE  instruction set is available.
   CPU_SSE2     - Intel SSE2 instruction set is available.
   CPU_3DNOW    - AMD 3DNow! instruction set is available.
   CPU_ENH3DNOW - AMD Enhanced 3DNow! instruction set is available.
   CPU_CMOV     - Pentium Pro "cmov" instruction is available.


   You can check for multiple features by OR-ing the flags together.
   For example, to check if the CPU has an FPU and MMX instructions
   available, you'd do:

   if ((cpu_capabilities & (CPU_FPU | CPU_MMX)) == (CPU_FPU | CPU_MMX))
      printf("CPU has both an FPU and MMX instructions!\n");




==========================================
============ Unicode routines ============
==========================================

Allegro can manipulate and display text using any character values from 0 
right up to 2^32-1 (although the current implementation of the grabber can 
only create fonts using characters up to 2^16-1). You can choose between a 
number of different text encoding formats, which controls how strings are 
stored and how Allegro interprets strings that you pass to it. This setting 
affects all aspects of the system: whenever you see a function that returns 
a char * type, or that takes a char * as an argument, that text will be in 
whatever format you have told Allegro to use.

By default, Allegro uses UTF-8 encoded text (U_UTF8). This is a 
variable-width format, where characters can occupy anywhere from one to six 
bytes. The nice thing about it is that characters ranging from 0-127 are 
encoded directly as themselves, so UTF-8 is upwardly compatible with 7 bit 
ASCII ("Hello, World!" means the same thing regardless of whether you 
interpret it as ASCII or UTF-8 data). Any character values above 128, such 
as accented vowels, the UK currency symbol, and Arabic or Chinese 
characters, will be encoded as a sequence of two or more bytes, each in the 
range 128-255. This means you will never get what looks like a 7 bit ASCII 
character as part of the encoding of a different character value, which 
makes it very easy to manipulate UTF-8 strings.

There are a few editing programs that understand UTF-8 format text files. 
Alternatively, you can write your strings in plain ASCII or 16 bit Unicode 
formats, and then use the Allegro textconv program to convert them into 
UTF-8.

If you prefer to use some other text format, you can set Allegro to work 
with normal 8 bit ASCII (U_ASCII), or 16 bit Unicode (U_UNICODE) instead, or 
you can provide some handler functions to make it support whatever other 
text encoding you like (for example it would be easy to add support for 32 
bit UCS-4 characters, or the Chinese GB-code format).

There is some limited support for alternative 8 bit codepages, via the 
U_ASCII_CP mode. This is very slow, so you shouldn't use it for serious 
work, but it can be handy as an easy way to convert text between different 
codepages. By default the U_ASCII_CP mode is set up to reduce text to a 
clean 7 bit ASCII format, trying to replace any accented vowels with their 
simpler equivalents (this is used by the allegro_message() function when it 
needs to print an error report onto a text mode DOS screen). If you want to 
work with other codepages, you can do this by passing a character mapping 
table to the set_ucodepage() function.

Note that you can use the Unicode routines before you call install_allegro() 
or allegro_init(). If you want to work in a text mode other than UTF-8, it 
is best to set it with set_uformat() just before you call these.

void set_uformat(int type);
   Sets the current text encoding format. This will affect all parts of 
   Allegro, wherever you see a function that returns a char *, or takes a 
   char * as a parameter. The type should be one of the values:

      U_ASCII     - fixed size, 8 bit ASCII characters
      U_ASCII_CP  - alternative 8 bit codepage (see set_ucodepage())
      U_UNICODE   - fixed size, 16 bit Unicode characters
      U_UTF8      - variable size, UTF-8 format Unicode characters

   Although you can change the text format on the fly, this is not a good 
   idea. Many strings, for example the names of your hardware drivers and 
   any language translations, are loaded when you call allegro_init(), so if 
   you change the encoding format after this, they will be in the wrong 
   format, and things will not work properly. Generally you should only call 
   set_uformat() once, before allegro_init(), and then leave it on the same 
   setting for the duration of your program.

int get_uformat(void);
   Returns the currently selected text encoding format.

void register_uformat(int type,
                      int (*u_getc)(const char *s),
                      int (*u_getx)(char **s),
                      int (*u_setc)(char *s, int c),
                      int (*u_width)(const char *s),
                      int (*u_cwidth)(int c),
                      int (*u_isok)(int c));
   Installs a set of custom handler functions for a new text encoding 
   format. The type is the ID code for your new format, which should be a 
   4-character string as produced by the AL_ID() macro, and which can later 
   be passed to functions like set_uformat() and uconvert(). The function 
   parameters are handlers that implement the character access for your new 
   type: see below for details of these.

void set_ucodepage(const unsigned short *table,
                   const unsigned short *extras);
   When you select the U_ASCII_CP encoding mode, a set of tables are used to 
   convert between 8 bit characters and their Unicode equivalents. You can 
   use this function to specify a custom set of mapping tables, which allows 
   you to support different 8 bit codepages. The table parameter points to 
   an array of 256 shorts, which contain the Unicode value for each 
   character in your codepage. The extras parameter, if not NULL, points to 
   a list of mapping pairs, which will be used when reducing Unicode data to 
   your codepage. Each pair consists of a Unicode value, followed by the way 
   it should be represented in your codepage. The table is terminated by a 
   zero Unicode value. This allows you to create a many->one mapping, where 
   many different Unicode characters can be represented by a single codepage 
   value (eg. for reducing accented vowels to 7 bit ASCII).

int need_uconvert(const char *s, int type, int newtype);
   Given a pointer to a string, a description of the type of the string, and 
   the type that you would like this string to be converted into, this 
   function tells you whether any conversion is required. No conversion will 
   be needed if type and newtype are the same, or if one type is ASCII, the 
   other is UTF-8, and the string contains only character values less than 
   128. As a convenience shortcut, you can pass the value U_CURRENT as 
   either of the type parameters, to represent whatever text format is 
   currently selected.

int uconvert_size(const char *s, int type, int newtype);
   Returns the number of bytes that will be required to store the specified 
   string after a conversion from type to newtype, including the zero 
   terminator. The type parameters can use the value U_CURRENT as a shortcut 
   to represent the currently selected encoding format.

void do_uconvert(const char *s, int type,
                 char *buf, int newtype, int size);
   Converts the specified string from type to newtype, storing at most size 
   bytes into the output buf. The type parameters can use the value 
   U_CURRENT as a shortcut to represent the currently selected encoding 
   format.

char *uconvert(const char *s, int type,
               char *buf, int newtype, int size);
   Higher level function running on top of do_uconvert(). This function 
   converts the specified string from type to newtype, storing at most size 
   bytes into the output buf, but it checks before doing the conversion, and 
   doesn't bother if the string formats are already the same (either both 
   types are equal, or one is ASCII, the other is UTF-8, and the string 
   contains only 7 bit ASCII characters). If a conversion was performed it 
   returns a pointer to buf, otherwise it returns a copy of s, so you must 
   use the return value rather than assuming that the string will always be 
   moved to buf. As a convenience, if buf is NULL it will convert the string 
   into an internal static buffer. You should be wary of using this feature, 
   though, because that buffer will be overwritten the next time this 
   routine is called, so don't expect the data to persist across any other 
   library calls.

char *uconvert_ascii(const char *s, char buf[]);
   Helper macro for converting strings from ASCII into the current encoding 
   format. Expands to uconvert(s, U_ASCII, buf, U_CURRENT, sizeof(buf)).

char *uconvert_toascii(const char *s, char buf[]);
   Helper macro for converting strings from the current encoding format into 
   ASCII. Expands to uconvert(s, U_CURRENT, buf, U_ASCII, sizeof(buf)).

extern char empty_string[];
   You can't just rely on "" to be a valid empty string in any encoding 
   format. This global buffer contains a number of consecutive zeros, so it 
   will be a valid empty string no matter whether the program is running in 
   ASCII, Unicode, or UTF-8 mode.

int ugetc(const char *s);
   Low level helper function for reading Unicode text data. Given a pointer 
   to a string in the current encoding format, it returns the next character 
   from the string.

int ugetx(char **s);
int ugetxc(const char **s);
   Low level helper function for reading Unicode text data. Given the 
   address of a pointer to a string in the current encoding format, it 
   returns the next character from the string, and advances the pointer to 
   the character after the one just read.

   ugetxc is provided for working with pointer-to-pointer-to-const char 
   data.

int usetc(char *s, int c);
   Low level helper function for writing Unicode text data. It writes the 
   specified character to the given address in the current encoding format, 
   and returns the number of bytes written.

int uwidth(const char *s);
   Low level helper function for testing Unicode text data. It returns the 
   number of bytes occupied by the first character of the specified string, 
   in the current encoding format.

int ucwidth(int c);
   Low level helper function for testing Unicode text data. It returns the 
   number of bytes that would be occupied by the specified character value, 
   when encoded in the current format.

int uisok(int c);
   Low level helper function for testing Unicode text data. Tests whether 
   the specified value can be correctly encoded in the current format.

int uoffset(const char *s, int index);
   Returns the offset in bytes from the start of the string to the character 
   at the specified index. If the index is negative, it counts backward from 
   the end of the string, so an index of -1 will return an offset to the last
   character.

int ugetat(const char *s, int index);
   Returns the character value at the specified index within the string. A 
   zero index parameter will return the first character of the string. If 
   the index is negative, it counts backward from the end of the string, so 
   an index of -1 will return the last character of the string.

int usetat(char *s, int index, int c);
   Replaces the character at the specified index within the string with 
   value c, handling any adjustments for variable width data (ie. if c 
   encodes to a different width than the previous value at that location). 
   Returns the number of bytes by which the trailing part of the string was 
   moved. If the index is negative, it counts backward from the end of the 
   string.

int uinsert(char *s, int index, int c);
   Inserts the character c at the specified index within the string, sliding 
   the rest of the data along to make room. Returns the number of bytes by 
   which the trailing part of the string was moved. If the index is 
   negative, it counts backward from the end of the string.

int uremove(char *s, int index);
   Removes the character at the specified index within the string, sliding 
   the rest of the data back to fill the gap. Returns the number of bytes by 
   which the trailing part of the string was moved. If the index is 
   negative, it counts backward from the end of the string.

int ustrsize(const char *s);
   Returns the size of the specified string in bytes, not including the 
   trailing zero.

int ustrsizez(const char *s);
   Returns the size of the specified string in bytes, including the trailing 
   zero.

int uwidth_max(int type);
   Low level helper function for working with Unicode text data. Returns the 
   largest number of bytes that one character can occupy in the given 
   encoding format. Pass U_CURRENT to represent the current format.

int utolower(int c);
   This function returns c, converting it to lower case if it is upper case.

int utoupper(int c);
   This function returns c, converting it to upper case if it is lower case.

int uisspace(int c);
   Returns nonzero if c is whitespace, that is, carriage return, newline,
   form feed, tab, vertical tab, or space.

int uisdigit(int c);
   Returns nonzero if c is a digit.

char *ustrdup(const char *src)
   This functions copies the NULL-terminated string src into a newly
   allocated area of memory. The memory returned by this call must be freed
   by the caller. Returns NULL if it cannot allocate space for the duplicated
   string.

char *_ustrdup(const char *src, void* (*malloc_func) (size_t))
   Does the same as ustrdup(), but allows the user to specify his own memory
   allocater function.

char *ustrcpy(char *dest, const char *src);
   This function copies src (including the terminating NULL character) into 
   dest. The return value is the value of dest.

char *ustrzcpy(char *dest, int size, const char *src);
   This function copies src (including the terminating NULL character) into 
   dest, whose length in bytes is specified by size and which is guaranteed 
   to be NULL-terminated. The return value is the value of dest.

char *ustrcat(char *dest, const char *src);
   This function concatenates src to the end of dest. The return value is the
   value of dest.

char *ustrzcat(char *dest, int size, const char *src);
   This function concatenates src to the end of dest, whose length in bytes 
   is specified by size and which is guaranteed to be NULL-terminated. The 
   return value is the value of dest.

int ustrlen(const char *s);
   This function returns the number of characters in s. Note that this
   doesn't have to equal the string's size in bytes.

int ustrcmp(const char *s1, const char *s2);
   This function compares s1 and s2. Returns zero if the strings are equal,
   a positive number if s1 comes after s2 in the ASCII collating sequence,
   else a negative number.

char *ustrncpy(char *dest, const char *src, int n);
   This function is like ustrcpy() except that no more than n characters 
   from src are copied into dest. If src is shorter than n characters, NULL 
   characters are appended to dest as padding until n characters have been 
   written. Note that if src is longer than n characters, dest will not be 
   NULL-terminated. The return value is the value of dest.

char *ustrzncpy(char *dest, int size, const char *src, int n);
   This function is like ustrzcpy() except that no more than n characters 
   from src are copied into dest. If src is shorter than n characters, NULL 
   characters are appended to dest as padding until n characters have been 
   written. Note that dest is guaranteed to be NULL-terminated. The return 
   value is the value of dest.

char *ustrncat(char *dest, const char *src, int n);
   This function is like ustrcat() except that no more than n characters 
   from src are appended to the end of dest. If the terminating NULL 
   character in src is reached before n characters have been written, the 
   NULL character is copied, but no other characters are written. If n 
   characters are written before a terminating NULL is encountered, the 
   function appends its own NULL character to dest, so that n+1 characters 
   are written. The return value is the value of dest.

char *ustrzncat(char *dest, int size, const char *src, int n);
   This function is like ustrzcat() except that no more than n characters 
   from src are appended to the end of dest. If the terminating NULL 
   character in src is reached before n characters have been written, the 
   NULL character is copied, but no other characters are written. Note that 
   dest is guaranteed to be NULL-terminated. The return value is the value 
   of dest.

int ustrncmp(const char *s1, const char *s2, int n);
   This function compares up to n characters of s1 and s2. Returns zero if
   the substrings are equal, a positive number if s1 comes after s2 in the
   ASCII collating sequence, else a negative number.

int ustricmp(const char *s1, const char *s2);
   This function compares s1 and s2, ignoring case.

char *ustrlwr(char *s);
   This function replaces all upper case letters in s with lower case
   letters.

char *ustrupr(char *s);
   This function replaces all lower case letters in s with upper case
   letters.

char *ustrchr(const char *s, int c);
   This function returns a pointer to the first occurrence of c in s, or
   NULL if no match was found. Note that if c is NULL, this will return a
   pointer to the end of the string.

char *ustrrchr(const char *s, int c);
   This function returns a pointer to the last occurrence of c in s, or NULL
   if no match was found.

char *ustrstr(const char *s1, const char *s2);
   This function finds the first occurence of s2 in s1. Returns a pointer
   within s1, or NULL if s2 wasn't found.

char *ustrpbrk(const char *s, const char *set);
   This function finds the first character in s that matches any character in
   set. Returns a pointer to the first match, or NULL if none are found.

char *ustrtok(char *s, const char *set);
   This function retrieves tokens from s which are delimited by characters
   from set. To initiate the search, pass the string to be searched as s.
   For the remaining tokens, pass NULL instead. Returns a pointer to the
   token, or NULL if no more are found. Warning: Since ustrtok alters the
   string it is parsing, you should always copy the string to a temporary
   buffer before parsing it. Also, this function is not reentrant (ie. you
   cannot parse two strings at the same time).

char *ustrtok_r(char *s, const char *set, char **last);
   Reentrant version of ustrtok. The last parameter is used to keep track
   of where the parsing is up to and must be a pointer to a char * variable
   allocated by the user that remains the same while parsing the same
   string.

double uatof(const char *s);
   Convert as much of the string as possible to an equivalent double
   precision real number. This function is almost like `ustrtod(s, NULL)'.
   Returns the equivalent value, or zero if the string does not represent a
   number.

long ustrtol(const char *s, char **endp, int base);
   This function converts the initial part of s to a signed integer, which
   is returned as a value of type `long int', setting *endp to point to the
   first unused character, if endp is not a NULL pointer. The base argument
   indicates what base the digits (or letters) should be treated as. If base
   is zero, the base is determined by looking for `0x', `0X', or `0' as the
   first part of the string, and sets the base used to 16, 16, or 8 if it
   finds one. The default base is 10 if none of those prefixes are found.

double ustrtod(const char *s, char **endp);
   This function converts as many characters of s that look like a floating
   point number into one, and sets *endp to point to the first unused
   character, if endp is not a NULL pointer.

const char *ustrerror(int err);
   This function returns a string that describes the error code `err', which
   normally comes from the variable `errno'. Returns a pointer to a static
   string that should not be modified or free'd. If you make subsequent
   calls to ustrerror, the string might be overwritten.

int usprintf(char *buf, const char *format, ...);
   This function writes formatted data into the output buffer. A NULL
   character is written to mark the end of the string. Returns the number of
   characters written, not including the terminating NULL character.

int uszprintf(char *buf, int size, const char *format, ...);
   This function writes formatted data into the output buffer, whose length
   in bytes is specified by size and which is guaranteed to be NULL
   terminated. Returns the number of characters that would have been written
   without eventual truncation (like with usprintf), not including the
   terminating NULL character.
   
int uvsprintf(char *buf, const char *format, va_list args);
   This is like usprintf(), but you pass the variable argument list directly,
   instead of the arguments themselves.
   
int uvszprintf(char *buf, int size, const char *format, va_list args);
   This is like uszprintf(), but you pass the variable argument list
   directly, instead of the arguments themselves.



================================================
============ Configuration routines ============
================================================

Various parts of Allegro, such as the sound routines and the 
load_joystick_data() function, require some configuration information. This 
data is stored in text files as a collection of "variable=value" lines, 
along with comments that begin with a '#' character and continue to the end 
of the line. The configuration file may optionally be divided into sections, 
which begin with a "[sectionname]" line. Each section has a unique 
namespace, to prevent variable name conflicts, but any variables that aren't 
in a section are considered to belong to all the sections simultaneously.

By default the configuration data is read from a file called allegro.cfg, 
which can be located either in the same directory as the program executable, 
or the directory pointed to by the ALLEGRO environment variable. Under Unix, 
it also checks for ~/allegro.cfg, ~/.allegrorc, /etc/allegro.cfg, and 
/etc/allegrorc, in that order; under BeOS only the last two are also checked.
If you don't like this approach, you can specify any filename you like, or
use a block of binary configuration data provided by your program (which
could for example be loaded from a datafile).

You can store whatever custom information you like in the config file, along 
with the standard variables that are used by Allegro (see below).

void set_config_file(const char *filename);
   Sets the configuration file to be used by all subsequent config 
   functions. If you don't call this function, Allegro will use the default 
   allegro.cfg file, looking first in the same directory as your program and 
   then in the directory pointed to by the ALLEGRO environment variable.

   All pointers returned by previous calls to get_config_string() and
   other related functions are invalidated when you call this function!

void set_config_data(const char *data, int length);
   Specifies a block of data to be used by all subsequent config functions, 
   which you have already loaded from disk (eg. as part of some more 
   complicated format of your own, or in a grabber datafile). This routine 
   makes a copy of the information, so you can safely free the data after 
   calling it.

void override_config_file(const char *filename);
   Specifies a file containing config overrides. These settings will be used 
   in addition to the parameters in the main config file, and where a 
   variable is present in both files this version will take priority. This 
   can be used by application programmers to override some of the config 
   settings from their code, while still leaving the main config file free 
   for the end user to customise. For example, you could specify a 
   particular sample frequency and IBK instrument file, but the user could 
   still use an allegro.cfg file to specify the port settings and irq 
   numbers.

void override_config_data(const char *data, int length);
   Version of override_config_file() which uses a block of data that has 
   already been read into memory.

void push_config_state();
   Pushes the current configuration state (filename, variable values, etc). 
   onto an internal stack, allowing you to select some other config source 
   and later restore the current settings by calling pop_config_state(). 
   This function is mostly intended for internal use by other library 
   functions, for example when you specify a config filename to the 
   save_joystick_data() function, it pushes the config state before 
   switching to the file you specified.

void pop_config_state();
   Pops a configuration state previously stored by push_config_state(), 
   replacing the current config source with it.

void flush_config_file();
   Writes the current config file to disk if the contents have changed 
   since it was loaded or since the latest call to the function.

void reload_config_texts(const char *new_language);
   Reloads the translated strings returned by get_config_text. This is
   useful to switch to another language in your program at runtime. If you
   want to modify the [system] language configuration variable yourself, or
   you have switched configuration files, you will want to pass NULL to
   just reload whatever language is currently selected. Or you can pass a
   string containing the two letter code of the language you desire to
   switch to, and the function will modify the language variable. After you
   call this function, the previously returned pointers of get_config_text
   will be invalid.

void hook_config_section(const char *section,
       int (*intgetter)(const char *name, int def),
       const char *(*stringgetter)(const char *name, const char *def),
       void (*stringsetter)(const char *name, const char *value));
   Takes control of the specified config file section, so that your hook 
   functions will be used to manipulate it instead of the normal disk file 
   access. If both the getter and setter functions are NULL, a currently 
   present hook will be unhooked. Hooked functions have the highest 
   priority. If a section is hooked, the hook will always be called, so you 
   can also hook a '#' section: even override_config_file() cannot override 
   a hooked section.

int config_is_hooked(const char *section);
   Returns TRUE if the specified config section has been hooked.

const char *get_config_string(const char *section,
                               const char *name, const char *def);
   Retrieves a string variable from the current config file. If the named
   variable cannot be found, or its entry in the config file is empty, the
   value of def is returned. The section name may be set to NULL to read
   variables from the root of the file, or used to control which set of
   parameters (eg. sound or joystick) you are interested in reading.

int get_config_int(const char *section, const char *name, int def);
   Reads an integer variable from the current config file. See the comments 
   about get_config_string().

int get_config_hex(const char *section, const char *name, int def);
   Reads an integer variable from the current config file, in hexadecimal 
   format. See the comments about get_config_string().

float get_config_float(const char *section, const char *name, float def);
   Reads a floating point variable from the current config file. See the 
   comments about get_config_string().

int get_config_id(const char *section, const char *name, int def);
   Reads a 4-letter driver ID variable from the current config file. See the 
   comments about get_config_string().

char **get_config_argv(const char *section, const char *name, int *argc);
   Reads a token list (words separated by spaces) from the current config 
   file, returning a an argv style argument list, and setting argc to the 
   number of tokens (unlike argc/argv, this list is zero based). Returns 
   NULL and sets argc to zero if the variable is not present. The token list 
   is stored in a temporary buffer that will be clobbered by the next call 
   to get_config_argv(), so the data should not be expected to persist.

const char *get_config_text(const char *msg);
   This function is primarily intended for use by internal library code, but 
   it may perhaps be helpful to application programmers as well. It uses the 
   language.dat or XXtext.cfg files (where XX is a language code) to look up
   a translated version of the parameter in the currently selected language,
   returning a suitable translation if one can be found or a copy of the
   parameter if nothing else is available. This is basically the same thing
   as calling get_config_string() with [language] as the section, msg as the
   variable name, and msg as the default value, but it contains some special
   code to handle Unicode format conversions. The msg parameter is always
   given in ASCII format, but the returned string will be converted into the
   current text encoding, with memory being allocated as required, so you
   can assume that this pointer will persist without having to manually
   allocate storage space for each string.

void set_config_string(const char *section, const char *name,
                        const char *val);
   Writes a string variable to the current config file, replacing any 
   existing value it may have, or removes the variable if val is NULL. The 
   section name may be set to NULL to write the variable to the root of the 
   file, or used to control which section the variable is inserted into. The 
   altered file will be cached in memory, and not actually written to disk 
   until you call allegro_exit(). Note that you can only write to files in 
   this way, so the function will have no effect if the current config 
   source was specified with set_config_data() rather than set_config_file().

   As a special case, variable or section names that begin with a '#' 
   character are treated specially and will not be read from or written to 
   the disk. Addon packages can use this to store version info or other 
   status information into the config module, from where it can be read with 
   the get_config_string() function.

void set_config_int(const char *section, const char *name, int val);
   Writes an integer variable to the current config file. See the comments 
   about set_config_string().

void set_config_hex(const char *section, const char *name, int val);
   Writes an integer variable to the current config file, in hexadecimal 
   format. See the comments about set_config_string().

void set_config_float(const char *section, const char *name, float val);
   Writes a floating point variable to the current config file. See the 
   comments about set_config_string().

void set_config_id(const char *section, const char *name, int val);
   Writes a 4-letter driver ID variable to the current config file. See the 
   comments about set_config_string().

Allegro uses these standard variables from the configuration file:

[system]
   Section containing general purpose variables:

system = x
   Specifies which system driver to use. This is currently only useful on 
   Linux, for choosing between the XWindows ("XWIN") or console ("LNXC") 
   modes.

keyboard = x
   Specifies which keyboard layout to use. The parameter is the name of a 
   keyboard mapping file produced by the keyconf utility, and can either be 
   a fully qualified file path or a basename like "us" or "uk". If the 
   latter, Allegro will look first for a separate config file with that name 
   (eg. "uk.cfg") and then for an object with that name in the keyboard.dat 
   file (eg. "UK_CFG"). The config file or keyboard.dat file can be stored 
   in the same directory as the program, or in the location pointed to by 
   the ALLEGRO environment variable. Look in the keyboard.dat file to see 
   what mappings are currently available.

language = x
   Specifies which language file to use for error messages and other bits of 
   system text. The parameter is the name of a translation file, and can 
   either be a fully qualified file path or a basename like "en" or "sp". If 
   the latter, Allegro will look first for a separate config file with a 
   name in the form "entext.cfg", and then for an object with that name in 
   the language.dat file (eg. "ENTEXT_CFG"). The config file or language.dat 
   file can be stored in the same directory as the program, or in the 
   location pointed to by the ALLEGRO environment variable. Look in the 
   language.dat file to see which mappings are currently available.

menu_opening_delay = x
   Sets how long the menus take to auto-open. The time is given in
   milliseconds (default is 300). Specifying -1 will disable the auto-opening
   feature.

dga_mouse = x
   X only: disable to work around a bug in some X servers' DGA modes, 
   concerning the mouse.  Default is on; enable the workaround by setting the
   variable to "0".

dga_centre = x
   X only: instructs the DGA driver to centre the Allegro screen if the
   actual screen resolution is higher than Allegro's.  Default is on; disable
   this feature by setting the variable to "0".

dga_clear = x
   X only: instructs the DGA driver to clear visible video memory on startup.
   Default is on; disable this feature by setting the variable to "0".

[graphics]
   Section containing graphics configuration information, using the
   variables:

gfx_card = x
   Specifies which graphics driver to use when the program requests 
   GFX_AUTODETECT. Multiple possible drivers can be suggested with extra 
   lines in the form "gfx_card1 = x", "gfx_card2 = x", etc, or you can 
   specify different drivers for each mode and color depth with variables in 
   the form "gfx_card_24bpp = x", "gfx_card_640x480x16 = x", etc.

gfx_cardw = x
   Specifies which graphics driver to use when the program requests 
   GFX_AUTODETECT_WINDOWED. This variable functions exactly like
   gfx_card in all other respects. If it is not set, Allegro will look
   for the gfx_card variable.

vbeaf_driver = x
   DOS and Linux only: specifies where to look for the VBE/AF driver 
   (vbeaf.drv). If this variable is not set, Allegro will look in the same 
   directory as the program, and then fall back on the standard locations 
   (c:\ for DOS, /usr/local/lib, /usr/lib, /lib, and / for Linux, or the 
   directory specified with the VBEAF_PATH environment variable).

framebuffer = x
   Linux only: specifies what device file to use for the fbcon driver. If 
   this variable is not set, Allegro checks the FRAMEBUFFER environment 
   variable, and then defaults to /dev/fb0.

force_centering = x
   Unix/X11 only: specifies whether to force window centering in fullscreen
   mode when the XWFS driver is used (yes or no). Enabling this setting may
   cause some artifacts to appear on KDE desktops.

disable_direct_updating = x
   Windows only: specifies whether to disable direct updating when the 
   GFX_DIRECTX_WIN driver is used in color conversion mode (yes or no).
   Direct updating can cause artifacts to be left on the desktop when the 
   window is moved or minimized; disabling it results in a significant
   performance loss.

[mouse]
   Section containing mouse configuration information, using the variables:

mouse = x
   Mouse driver type. Available DOS drivers are:
      MICK - mickey mode driver (normally the best)
      I33  - int 0x33 callback driver
      POLL - timer polling (for use under NT)
   Linux console mouse drivers are:
      MS   - Microsoft serial mouse
      IMS  - Microsoft serial mouse with Intellimouse extension
      LPS2 - PS2 mouse
      LIPS - PS2 mouse with Intellimouse extension
      GPMD - GPM repeater data (Mouse Systems protocol)

num_buttons = x
   Sets the number of mouse buttons viewed by Allegro. You don't normally
   need to set this variable because Allegro will autodetect it. You can only
   use it to restrict the set of actual mouse buttons.

emulate_three = x
   Sets whether to emulate a third mouse button by detecting chords of the 
   left and right buttons (yes or no). Defaults to yes if you have a two 
   button mouse, no otherwise.

mouse_device = x
   Linux only: specifies the name of the mouse device file (eg. /dev/mouse).

mouse_accel_factor = x
   Windows only: specifies the mouse acceleration factor. Defaults to 1.
   Set it to 0 in order to disable mouse acceleration. 2 accelerates twice
   as much as 1.

[sound]
   Section containing sound configuration information, using the variables:

digi_card = x
   Sets the driver to use for playing digital samples.

midi_card = x
   Sets the driver to use for MIDI music.

digi_input_card = x
   Sets the driver to use for digital sample input.

midi_input_card = x
   Sets the driver to use for MIDI data input.

digi_voices = x
   Specifies the minimum number of voices to reserve for use by the digital 
   sound driver. How many are possible depends on the driver.

midi_voices = x
   Specifies the minimum number of voices to reserve for use by the MIDI 
   sound driver. How many are possible depends on the driver.

digi_volume = x
   Sets the volume for digital sample playback, from 0 to 255.

midi_volume = x
   Sets the volume for midi music playback, from 0 to 255.

quality = x
   Controls the sound quality vs. performance tradeoff for the sample mixing 
   code. This can be set to any of the values:
      0 - fast mixing of 8 bit data into 16 bit buffers
      1 - true 16 bit mixing (requires a 16 bit stereo soundcard)
      2 - interpolated 16 bit mixing

flip_pan = x
   Toggling this between 0 and 1 reverses the left/right panning of samples, 
   which might be needed because some SB cards (including mine :-) get the 
   stereo image the wrong way round.

sound_freq = x
   DOS, Unix and BeOS: sets the sample frequency. With the SB driver, 
   possible rates are 11906 (any), 16129 (any), 22727 (SB 2.0 and above), 
   and 45454 (only on SB 2.0 or SB16, not the stereo SB Pro driver). On the 
   ESS Audiodrive, possible rates are 11363, 17046, 22729, or 44194. On the 
   Ensoniq Soundscape, possible rates are 11025, 16000, 22050, or 48000. On 
   the Windows Sound System, possible rates are 11025, 22050, 44100, or 
   48000. Don't worry if you set some other number by mistake: Allegro will 
   automatically round it to the closest supported frequency.

sound_bits = x
   Unix and BeOS: sets the preferred number of bits (8 or 16).

sound_stereo = x
   Unix and BeOS: selects mono or stereo output (0 or 1).

sound_port = x
   DOS only: sets the soundcard port address (this is usually 220).

sound_dma = x
   DOS only: sets the soundcard DMA channel (this is usually 1).

sound_irq = x
   DOS only: sets the soundcard IRQ number (this is usually 7).

fm_port = x
   DOS only: sets the port address of the OPL synth (this is usually 388).

mpu_port = x
   DOS only: sets the port address of the MPU-401 MIDI interface (this is 
   usually 330).

mpu_irq = x
   DOS only: sets the IRQ for the MPU-401 (this is usually the same as 
   sound_irq).

ibk_file = x
   DOS only: specifies the name of a .IBK file which will be used to replace 
   the standard Adlib patch set.

ibk_drum_file = x
   DOS only: specifies the name of a .IBK file which will be used to replace 
   the standard set of Adlib percussion patches.

oss_driver = x
   Unix only: sets the OSS device driver name. Usually /dev/dsp or 
   /dev/audio, but could be a particular device (e.g. /dev/dsp2).

oss_numfrags = x
oss_fragsize = x
   Unix only: sets number of OSS driver fragments (buffers) and size of each 
   buffer in samples. Buffers are filled with data in the interrupts where 
   interval between subsequent interrupts is not less than 10 ms. If 
   hardware can play all information from buffers faster than 10 ms, then 
   there will be clicks, when hardware have played all data and library has 
   not prepared new data yet. On the other hand, if it takes too long for 
   device driver to play data from all buffers, then there will be delays 
   between action which triggers sound and sound itself.

oss_midi_driver = x
   Unix only: sets the OSS MIDI device name. Usually /dev/sequencer.

oss_mixer_driver = x
   Unix only: sets the OSS mixer device name. Usually /dev/mixer.

esd_server = x
   Unix only: where to find the ESD (Enlightened Sound Daemon) server.

alsa_card = x
alsa_pcmdevice = x
   Unix only: paramaters for the ALSA sound driver system.

alsa_numfrags = x
   Unix only: number of ALSA driver fragments (buffers).

alsa_fragsize = x
   Unix only: size of each ALSA fragment, in samples.

alsa_rawmidi_card = x
   Unix only: ALSA card to use for midi output.

alsa_rawmidi_device = x
   Unix only: ALSA rawmidi device to use for output.

be_midi_quality = x
   BeOS only: system MIDI synthesizer instruments quality. 0 uses low
   quality 8-bit 11 kHz samples, 1 uses 16-bit 22 kHz samples.

be_midi_freq = x
   BeOS only: MIDI sample mixing frequency in Hz. Can be 11025, 22050 or
   44100.

be_midi_interpolation = x
   BeOS only: specifies the MIDI samples interpolation method. 0 doesn't
   interpolate, it's fast but has the worst quality; 1 does a fast
   interpolation with better performances, but it's a bit slower than the
   previous method; 2 does a linear interpolation between samples, it is the
   slowest method but gives the best performances.

be_midi_reverb = x
   BeOS only: reverberation intensity, from 0 to 5. 0 disables it, 5 is the
   strongest one.

patches = x
   Specifies where to find the sample set for the DIGMID driver. This can 
   either be a Gravis style directory containing a collection of .pat files 
   and a default.cfg index, or an Allegro datafile produced by the pat2dat 
   utility. If this variable is not set, Allegro will look either for a 
   default.cfg or patches.dat file in the same directory as the program, the 
   directory pointed to by the ALLEGRO environment variable, and the 
   standard GUS directory pointed to by the ULTRASND environment variable.

[midimap]
   If you are using the SB MIDI output or MPU-401 drivers with an external 
   synthesiser that is not General MIDI compatible, you can use the midimap 
   section of the config file to specify a patch mapping table for 
   converting GM patch numbers into whatever bank and program change 
   messages will select the appropriate sound on your synth. This is a real 
   piece of self-indulgence. I have a Yamaha TG500, which has some great 
   sounds but no GM patch set, and I just had to make it work somehow...

   This section consists of a set of lines in the form:

p<n> = bank0 bank1 prog pitch
   With this statement, n is the GM program change number (1-128), bank0 and 
   bank1 are the two bank change messages to send to your synth (on 
   controllers #0 and #32), prog is the program change message to send to 
   your synth, and pitch is the number of semitones to shift everything that 
   is played with that sound. Setting the bank change numbers to -1 will 
   prevent them from being sent.

   For example, the line:

      p36 = 0 34 9 12

   specifies that whenever GM program 36 (which happens to be a fretless 
   bass) is selected, Allegro should send a bank change message #0 with a 
   parameter of 0, a bank change message #32 with a parameter of 34, a 
   program change with a parameter of 9, and then should shift everything up 
   by an octave.

[joystick]
   Section containing joystick configuration information, using the
   variables:

joytype = x
   Specifies which joystick driver to use when the program requests 
   JOY_TYPE_AUTODETECT.

joystick_device = x
   BeOS only: specifies the name of the joystick device to be used. First
   device found is used by default.

throttle_axis = x
   Linux only: sets which axis number the throttle is located at. This
   variable will be used for every detected joystick. If you want to specify
   the axis number for each joystick individually, use variables of the form
   throttle_axis_n, where n is the joystick number.




========================================
============ Mouse routines ============
========================================

int install_mouse();
   Installs the Allegro mouse handler. You must do this before using any 
   other mouse functions. Returns -1 on failure, otherwise the number of 
   buttons on the mouse.

void remove_mouse();
   Removes the mouse handler. You don't normally need to bother calling 
   this, because allegro_exit() will do it for you.

int poll_mouse();
   Wherever possible, Allegro will read the mouse input asynchronously (ie. 
   from inside an interrupt handler), but on some platforms that may not be 
   possible, in which case you must call this routine at regular intervals 
   to update the mouse state variables. To help you test your mouse polling 
   code even if you are programming on a platform that doesn't require it, 
   after the first time that you call this function Allegro will switch into 
   polling mode, so from that point onwards you will have to call this 
   routine in order to get any mouse input at all, regardless of whether the 
   current driver actually needs to be polled or not. Returns zero on
   success, or a negative number on failure (ie. no mouse driver installed).

int mouse_needs_poll();
   Returns TRUE if the current mouse driver is operating in polling mode.

extern volatile int mouse_x;
extern volatile int mouse_y;
extern volatile int mouse_z;
extern volatile int mouse_b;
extern volatile int mouse_pos;
   Global variables containing the current mouse position and button state. 
   Wherever possible these values will be updated asynchronously, but if 
   mouse_needs_poll() returns TRUE, you must manually call poll_mouse() to 
   update them with the current input state. The mouse_x and mouse_y 
   positions are integers ranging from zero to the bottom right corner of 
   the screen. The mouse_z variable holds the current wheel position, when 
   using an input driver that supports wheel mice. The mouse_b variable is a 
   bitfield indicating the state of each button: bit 0 is the left button, 
   bit 1 the right, and bit 2 the middle button. For example:

      if (mouse_b & 1)
         printf("Left button is pressed\n");

      if (!(mouse_b & 2))
         printf("Right button is not pressed\n");

   The mouse_pos variable has the current X coordinate in the high word and 
   the Y in the low word. This may be useful in tight polling loops where a 
   mouse interrupt could occur between your reading of the two separate 
   variables, since you can copy this value into a local variable with a 
   single instruction and then split it up at your leisure.

extern BITMAP *mouse_sprite;
extern int mouse_x_focus;
extern int mouse_y_focus;
   Global variables containing the current mouse sprite and the focus
   point.  These are read-only, and only to be modified using the
   set_mouse_sprite() and set_mouse_sprite_focus() functions.

void show_mouse(BITMAP *bmp);
   Tells Allegro to display a mouse pointer on the screen. This will only 
   work if the timer module has been installed. The mouse pointer will be 
   drawn onto the specified bitmap, which should normally be 'screen' (see 
   later for information about bitmaps). To hide the mouse pointer, call 
   show_mouse(NULL). Warning: if you draw anything onto the screen while the 
   pointer is visible, a mouse movement interrupt could occur in the middle 
   of your drawing operation. If this happens the mouse buffering and SVGA 
   bank switching code will get confused and will leave 'mouse droppings' 
   all over the screen. To prevent this, you must make sure you turn off the 
   mouse pointer whenever you draw onto the screen.

void scare_mouse();
   Helper for hiding the mouse pointer prior to a drawing operation. This 
   will temporarily get rid of the pointer, but only if that is really 
   required (ie. the mouse is visible, and is displayed on the physical 
   screen rather than some other memory surface, and it is not a hardware 
   cursor). The previous mouse state is stored for subsequent calls to 
   unscare_mouse().

void scare_mouse_area(int x, int y, int w, int h);
   Like scare_mouse(), but will only hide the cursor if it is inside the 
   specified rectangle. Otherwise the cursor will simply be frozen in place 
   until you call unscare_mouse(), so it cannot interfere with your drawing.

void unscare_mouse();
   Undoes the effect of a previous call to scare_mouse() or 
   scare_mouse_area(), restoring the original pointer state.

extern volatile int freeze_mouse_flag;
   If this flag is set, the mouse pointer won't be redrawn when the mouse 
   moves. This can avoid the need to hide the pointer every time you draw to 
   the screen, as long as you make sure your drawing doesn't overlap with 
   the current pointer position.

void position_mouse(int x, int y);
   Moves the mouse to the specified screen position. It is safe to call even 
   when a mouse pointer is being displayed.

void position_mouse_z(int z);
   Sets the mouse wheel position variable to the specified value.

void set_mouse_range(int x1, int y1, int x2, int y2);
   Sets the area of the screen within which the mouse can move. Pass the top 
   left corner and the bottom right corner (inclusive). If you don't call 
   this function the range defaults to (0, 0, SCREEN_W-1, SCREEN_H-1).

void set_mouse_speed(int xspeed, int yspeed);
   Sets the mouse speed. Larger values of xspeed and yspeed represent slower 
   mouse movement: the default for both is 2.

void set_mouse_sprite(BITMAP *sprite);
   You don't like my mouse pointer? No problem. Use this function to supply 
   an alternative of your own. If you change the pointer and then want to 
   get my lovely arrow back again, call set_mouse_sprite(NULL).

   As a bonus, set_mouse_sprite(NULL) uses the current palette in choosing
   colors for the arrow. So if your arrow mouse sprite looks ugly after
   changing the palette, call set_mouse_sprite(NULL).

void set_mouse_sprite_focus(int x, int y);
   The mouse focus is the bit of the pointer that represents the actual 
   mouse position, ie. the (mouse_x, mouse_y) position. By default this is 
   the top left corner of the arrow, but if you are using a different mouse 
   pointer you might need to alter it.

void get_mouse_mickeys(int *mickeyx, int *mickeyy);
   Measures how far the mouse has moved since the last call to this 
   function. The mouse will continue to generate movement mickeys even when 
   it reaches the edge of the screen, so this form of input can be useful 
   for games that require an infinite range of mouse movement.

extern void (*mouse_callback)(int flags);
   Called by the interrupt handler whenever the mouse moves or one of the 
   buttons changes state. This function must be in locked memory, and must 
   execute _very_ quickly! It is passed the event flags that triggered the 
   call, which is a bitmask containing any of the values MOUSE_FLAG_MOVE, 
   MOUSE_FLAG_LEFT_DOWN, MOUSE_FLAG_LEFT_UP, MOUSE_FLAG_RIGHT_DOWN, 
   MOUSE_FLAG_RIGHT_UP, MOUSE_FLAG_MIDDLE_DOWN, MOUSE_FLAG_MIDDLE_UP, and 
   MOUSE_FLAG_MOVE_Z.



========================================
============ Timer routines ============
========================================

Allegro can set up several virtual timer functions, all going at different
speeds.

Under DOS it will constantly reprogram the clock to make sure they are all
called at the correct times. Because they alter the low level timer chip
settings, these routines should not be used together with other DOS timer
functions like the djgpp uclock() routine. Moreover, the FPU state is not
preserved across Allegro interrupts so you ought not to use floating point
or MMX code inside timer interrupt handlers.

Under other platforms, they are usually implemented using threads, which run
parallel to the main thread. Therefore timer callbacks on such platforms
will not block the main thread when called, so you may need to use
appropriate synchronisation devices (eg. mutexes, semaphores, etc.) when
accessing data that is shared by a callback and the main thread. (Currently
Allegro does not provide such devices.)

int install_timer();
   Installs the Allegro timer interrupt handler. You must do this before 
   installing any user timer routines, and also before displaying a mouse 
   pointer, playing FLI animations or MIDI music, and using any of the GUI 
   routines. Returns zero on success, or a negative number on failure (but
   you may decide not to check the return value as this function is very
   unlikely to fail).

void remove_timer();
   Removes the Allegro timer handler (and, under DOS, passes control of the
   clock back to the operating system). You don't normally need to bother
   calling this, because allegro_exit() will do it for you.

int install_int(void (*proc)(), int speed);
   Installs a user timer handler, with the speed given as the number of 
   milliseconds between ticks. This is the same thing as 
   install_int_ex(proc, MSEC_TO_TIMER(speed)). If you call this routine 
   without having first installed the timer module, install_timer() will be 
   called automatically. If there is no room to add a new user timer,
   install_int() will return a negative number, otherwise it returns zero.

int install_int_ex(void (*proc)(), int speed);
   Adds a function to the list of user timer handlers or, if it is already 
   installed, retroactively adjusts its speed (i.e makes as though the speed
   change occured precisely at the last tick). The speed is given in hardware
   clock ticks, of which there are 1193181 a second. You can convert from 
   other time formats to hardware clock ticks with the macros:

      SECS_TO_TIMER(secs)  - give the number of seconds between
                             each tick
      MSEC_TO_TIMER(msec)  - give the number of milliseconds
                             between ticks
      BPS_TO_TIMER(bps)    - give the number of ticks each second
      BPM_TO_TIMER(bpm)    - give the number of ticks per minute

   If there is no room to add a new user timer, install_int_ex() will return 
   a negative number, otherwise it returns zero. There can only be sixteen 
   timers in use at a time, and some other parts of Allegro (the GUI code, 
   the mouse pointer display routines, rest(), the FLI player, and the MIDI 
   player) need to install handlers of their own, so you should avoid using 
   too many at the same time. If you call this routine without having first 
   installed the timer module, install_timer() will be called automatically.

   Your function will be called by the Allegro interrupt handler and not 
   directly by the processor, so it can be a normal C function and does not 
   need a special wrapper. You should be aware, however, that it will be 
   called in an interrupt context, which imposes a lot of restrictions on 
   what you can do in it. It should not use large amounts of stack, it must 
   not make any calls to the operating system, use C library functions, or 
   contain any floating point code, and it must execute very quickly. Don't 
   try to do lots of complicated code in a timer handler: as a general rule 
   you should just set some flags and respond to these later in your main 
   control loop.

   In a DOS protected mode environment like djgpp, memory is virtualised and 
   can be swapped to disk. Due to the non-reentrancy of DOS, if a disk swap 
   occurs inside an interrupt handler the system will die a painful death, 
   so you need to make sure you lock all the memory (both code and data) 
   that is touched inside timer routines. Allegro will lock everything it 
   uses, but you are responsible for locking your handler functions. The 
   macros LOCK_VARIABLE (variable), END_OF_FUNCTION (function_name),
   END_OF_STATIC_FUNCTION (function_name), and LOCK_FUNCTION (function_name)
   can be used to simplify this task. For example, if you want an interrupt 
   handler that increments a counter variable, you should write:

      volatile int counter;

      void my_timer_handler()
      {
         counter++;
      }

      END_OF_FUNCTION(my_timer_handler)

   and in your initialisation code you should lock the memory:

      LOCK_VARIABLE(counter);
      LOCK_FUNCTION(my_timer_handler);

   Obviously this can get awkward if you use complicated data structures and 
   call other functions from within your handler, so you should try to keep 
   your interrupt routines as simple as possible.

void remove_int(void (*proc)());
   Removes a function from the list of user interrupt routines. At program 
   termination, allegro_exit() does this automatically.

int install_param_int(void (*proc)(void *), void *param, int speed);
   Like install_int(), but the callback routine will be passed a copy of the 
   specified void pointer parameter. To disable the handler, use 
   remove_param_int() instead of remove_int().

int install_param_int_ex(void (*proc)(void *), void *param, int speed);
   Like install_int_ex(), but the callback routine will be passed a copy of 
   the specified void pointer parameter. To disable the handler, use 
   remove_param_int() instead of remove_int().

void remove_param_int(void (*proc)(void *), void *param);
   Like remove_int(), but for use with timer callbacks that have parameter 
   values. If there is more than one copy of the same callback active at a 
   time, it identifies which one to remove by checking the parameter value 
   (so you can't have more than one copy of a handler using an identical 
   parameter).

int timer_can_simulate_retrace()
   Checks whether it is possible to sync the timer module with the monitor 
   retrace, given the current platform and environment (at the moment this 
   is only possible when running in clean DOS mode in a VGA or mode-X 
   resolution). Returns non-zero if simulation is possible.

void timer_simulate_retrace(int enable);
   The DOS timer handler can be used to simulate vertical retrace 
   interrupts. A retrace interrupt can be extremely useful for implementing 
   smooth animation, but unfortunately the VGA hardware doesn't support it. 
   The EGA did, and some SVGA chipsets do, but not enough, and not in a 
   sufficiently standardised way, for it to be useful. Allegro works around 
   this by programming the timer to generate an interrupt when it thinks a 
   retrace is next likely to occur, and polling the VGA inside the interrupt 
   handler to make sure it stays in sync with the monitor refresh. This 
   works quite well in some situations, but there are a lot of caveats:

   - You can't use the retrace simulator in SVGA modes. It will work with 
     some chipsets, but not others, and it conflicts with most VESA 
     implementations. Retrace simulation is only reliable in VGA mode 13h 
     and mode-X.

   - Retrace simulation doesn't work under win95, because win95 returns 
     garbage when I try to read the elapsed time from the PIT. If anyone 
     knows how I can make this work, please tell me!

   - Retrace simulation involves a lot of waiting around in the timer 
     handler with interrupts disabled. This will significantly slow down 
     your entire system, and may also cause static when playing samples on 
     SB 1.0 cards (because they don't support auto-initialised DMA: SB 2.0 
     and above will be fine).

   Bearing all those problems in mind, I'd strongly advise against relying 
   on the retrace simulator. If you are coding in mode-X, and don't care 
   about your program working under win95, it is great, but it would be a 
   good idea to give the user an option to disable it.

   Retrace simulation must be enabled before you use the triple buffering 
   functions in a mode-X resolution. It can also be useful for simple 
   retrace detection, because the polling vsync() function can occasionally 
   miss retraces if a soundcard or timer interrupt occurs at exactly the 
   same time as the retrace. When retrace interrupt simulation is enabled, 
   vsync() will check the retrace_count variable rather than polling the 
   VGA, so it won't miss retraces even if they are masked by other 
   interrupts.

int timer_is_using_retrace()
   Checks whether the timer module is currently synced with the monitor 
   retrace or not. Returns non-zero if it is.

extern volatile int retrace_count;
   If the retrace simulator is installed, this is incremented on each 
   vertical retrace, otherwise it is incremented 70 times a second (ignoring 
   retraces). This provides a useful way of controlling the speed of your 
   program without the hassle of installing user timer functions.

   The speed of retraces varies depending on the graphics mode. In mode 13h 
   and 200/400 line mode-X resolutions there are 70 retraces a second, and 
   in 240/480 line modes there are 60. It can be as low as 50 (in 376x282 
   mode) or as high as 92 (in 400x300 mode).

extern void (*retrace_proc)();
   If the retrace simulator is installed, this function is called during 
   every vertical retrace, otherwise it is called 70 times a second 
   (ignoring retraces). Set it to NULL to disable the callback. The function 
   must obey the same rules as regular timer handlers (ie. it must be 
   locked, and mustn't call OS or libc functions) but even more so: it must 
   execute _very_ quickly, or it will mess up the timer synchronisation. The 
   only use I can see for this function is for doing palette manipulations, 
   because triple buffering can be done with the request_scroll() function, 
   and the retrace_count variable can be used for timing your code. If you 
   want to alter the palette in the retrace_proc you should use the inline 
   _set_color() function rather than the regular set_color() or 
   set_palette(), and you shouldn't try to alter more than two or three 
   palette entries in a single retrace.

void rest(long time);
   Once Allegro has taken over the timer the standard delay() function will 
   no longer work, so you should use this routine instead. The time is given 
   in milliseconds.

void rest_callback(long time, void (*callback)())
   Like rest(), but continually calls the specified function while it is 
   waiting for the required time to elapse.



===========================================
============ Keyboard routines ============
===========================================

The Allegro keyboard handler provides both buffered input and a set of flags 
storing the current state of each key. Note that it is not possible to 
correctly detect every combination of keys, due to the design of the PC 
keyboard. Up to two or three keys at a time will work fine, but if you press 
more than that the extras are likely to be ignored (exactly which 
combinations are possible seems to vary from one keyboard to another).

int install_keyboard();
   Installs the Allegro keyboard interrupt handler. You must call this 
   before using any of the keyboard input routines. Once you have set up the 
   Allegro handler, you can no longer use operating system calls or C 
   library functions to access the keyboard. Returns zero on success, or a
   negative number on failure (but you may decide not to check the return
   value as this function is very unlikely to fail). Note that on some
   platforms the keyboard won't work unless you have set a graphic mode,
   even if this function returns zero before calling set_gfx_mode.

void remove_keyboard();
   Removes the keyboard handler, returning control to the operating system. 
   You don't normally need to bother calling this, because allegro_exit() 
   will do it for you.

void install_keyboard_hooks(int (*keypressed)(), int (*readkey)());
   You should only use this function if you *aren't* using the rest of the 
   keyboard handler. It should be called in the place of install_keyboard(), 
   and lets you provide callback routines to detect and read keypresses, 
   which will be used by the main keypressed() and readkey() functions. This 
   can be useful if you want to use Allegro's GUI code with a custom 
   keyboard handler, as it provides a way for the GUI to get keyboard input 
   from your own code, bypassing the normal Allegro input system.

int poll_keyboard();
   Wherever possible, Allegro will read the keyboard input asynchronously 
   (ie. from inside an interrupt handler), but on some platforms that may 
   not be possible, in which case you must call this routine at regular 
   intervals to update the keyboard state variables. To help you test your 
   keyboard polling code even if you are programming on a platform that 
   doesn't require it, after the first time that you call this function 
   Allegro will switch into polling mode, so from that point onwards you 
   will have to call this routine in order to get any keyboard input at all, 
   regardless of whether the current driver actually needs to be polled or 
   not. The keypressed(), readkey(), and ureadkey() functions call 
   poll_keyboard() automatically, so you only need to use this function when 
   accessing the key[] array and key_shifts variable. Returns zero on
   success, or a negative number on failure (ie. no keyboard driver
   installed).

int keyboard_needs_poll();
   Returns TRUE if the current keyboard driver is operating in polling mode.

extern volatile char key[KEY_MAX];
   Array of flags indicating the state of each key, ordered by scancode. 
   Wherever possible these values will be updated asynchronously, but if 
   keyboard_needs_poll() returns TRUE, you must manually call 
   poll_keyboard() to update them with the current input state. The 
   scancodes are defined in allegro/keyboard.h as a series of KEY_*
   constants (and are also listed below). For example, you could write:

      if (key[KEY_SPACE])
         printf("Space is pressed\n");


   Note that the array is supposed to represent which keys are physically
   held down and which keys are not, so it is semantically read-only.

   These are the keyboard scancodes:

      KEY_A ... KEY_Z,
      KEY_0 ... KEY_9,
      KEY_0_PAD ... KEY_9_PAD,
      KEY_F1 ... KEY_F12,

      KEY_ESC, KEY_TILDE, KEY_MINUS, KEY_EQUALS,
      KEY_BACKSPACE, KEY_TAB, KEY_OPENBRACE, KEY_CLOSEBRACE,
      KEY_ENTER, KEY_COLON, KEY_QUOTE, KEY_BACKSLASH,
      KEY_BACKSLASH2, KEY_COMMA, KEY_STOP, KEY_SLASH,
      KEY_SPACE,

      KEY_INSERT, KEY_DEL, KEY_HOME, KEY_END, KEY_PGUP,
      KEY_PGDN, KEY_LEFT, KEY_RIGHT, KEY_UP, KEY_DOWN,

      KEY_SLASH_PAD, KEY_ASTERISK, KEY_MINUS_PAD,
      KEY_PLUS_PAD, KEY_DEL_PAD, KEY_ENTER_PAD,

      KEY_PRTSCR, KEY_PAUSE,

      KEY_ABNT_C1, KEY_YEN, KEY_KANA, KEY_CONVERT, KEY_NOCONVERT,
      KEY_AT, KEY_CIRCUMFLEX, KEY_COLON2, KEY_KANJI,

      KEY_LSHIFT, KEY_RSHIFT,
      KEY_LCONTROL, KEY_RCONTROL,
      KEY_ALT, KEY_ALTGR,
      KEY_LWIN, KEY_RWIN, KEY_MENU,
      KEY_SCRLOCK, KEY_NUMLOCK, KEY_CAPSLOCK
   
extern volatile int key_shifts;
   Bitmask containing the current state of shift/ctrl/alt, the special 
   Windows keys, and the accent escape characters. Wherever possible this 
   value will be updated asynchronously, but if keyboard_needs_poll() 
   returns TRUE, you must manually call poll_keyboard() to update it with 
   the current input state. This can contain any of the flags:

      KB_SHIFT_FLAG
      KB_CTRL_FLAG
      KB_ALT_FLAG
      KB_LWIN_FLAG
      KB_RWIN_FLAG
      KB_MENU_FLAG
      KB_SCROLOCK_FLAG
      KB_NUMLOCK_FLAG
      KB_CAPSLOCK_FLAG
      KB_INALTSEQ_FLAG
      KB_ACCENT1_FLAG
      KB_ACCENT2_FLAG
      KB_ACCENT3_FLAG
      KB_ACCENT4_FLAG

int keypressed();
   Returns TRUE if there are keypresses waiting in the input buffer. This is 
   equivalent to the libc kbhit() function.

int readkey();
   Returns the next character from the keyboard buffer, in ASCII format. If 
   the buffer is empty, it waits until a key is pressed. The low byte of the 
   return value contains the ASCII code of the key, and the high byte the 
   scancode. The scancode remains the same whatever the state of the shift, 
   ctrl and alt keys, while the ASCII code is affected by shift and ctrl in 
   the normal way (shift changes case, ctrl+letter gives the position of 
   that letter in the alphabet, eg. ctrl+A = 1, ctrl+B = 2, etc). Pressing 
   alt+key returns only the scancode, with a zero ASCII code in the low 
   byte. For example:

      if ((readkey() & 0xff) == 'd')         // by ASCII code
         printf("You pressed 'd'\n");

      if ((readkey() >> 8) == KEY_SPACE)     // by scancode
         printf("You pressed Space\n");

      if ((readkey() & 0xff) == 3)           // ctrl+letter
         printf("You pressed Control+C\n");

      if (readkey() == (KEY_X << 8))         // alt+letter
         printf("You pressed Alt+X\n");

   This function cannot return character values greater than 255. If you 
   need to read Unicode input, use ureadkey() instead.

int ureadkey(int *scancode);
   Returns the next character from the keyboard buffer, in Unicode format. 
   If the buffer is empty, it waits until a key is pressed. The return value 
   contains the Unicode value of the key, and if not NULL, the pointer 
   argument will be set to the scancode. Unlike readkey(), this function is 
   able to return character values greater than 255.

int scancode_to_ascii(int scancode);
   Converts the given scancode to an ASCII character for that key, returning 
   the unshifted uncapslocked result of pressing the key, or zero if the key 
   isn't a character-generating key or the lookup can't be done.

void simulate_keypress(int key);
   Stuffs a key into the keyboard buffer, just as if the user had pressed 
   it. The parameter is in the same format returned by readkey().

void simulate_ukeypress(int key, int scancode);
   Stuffs a key into the keyboard buffer, just as if the user had pressed 
   it. The parameter is in the same format returned by ureadkey().

extern int (*keyboard_callback)(int key);
   If set, this function is called by the keyboard handler in response to 
   every keypress. It is passed a copy of the value that is about to be 
   added into the input buffer, and can either return this value unchanged, 
   return zero to cause the key to be ignored, or return a modified value to 
   change what readkey() will later return. This routine executes in an 
   interrupt context, so it must be in locked memory.

extern int (*keyboard_ucallback)(int key, int *scancode);
   Unicode-aware version of keyboard_callback(). If set, this function is 
   called by the keyboard handler in response to every keypress. It is 
   passed the character value and scancode that are about to be added into 
   the input buffer, can modify the scancode value, and returns a new or 
   modified key code. If it both sets the scancode to zero and returns zero, 
   the keypress will be ignored. This routine executes in an interrupt 
   context, so it must be in locked memory.

extern void (*keyboard_lowlevel_callback)(int scancode);
   If set, this function is called by the keyboard handler in response to 
   every keyboard event, both presses and releases. It will be passed a raw 
   keyboard scancode byte, with the top bit clear if the key has been 
   pressed or set if it was released. This routine executes in an interrupt 
   context, so it must be in locked memory.

void set_leds(int leds);
   Overrides the state of the keyboard LED indicators. The parameter is a 
   bitmask containing any of the values KB_SCROLOCK_FLAG, KB_NUMLOCK_FLAG, 
   and KB_CAPSLOCK_FLAG, or -1 to restore the default behavior.

void set_keyboard_rate(int delay, int repeat);
   Sets the keyboard repeat rate. Times are given in milliseconds. Passing 
   zero times will disable the key repeat.

void clear_keybuf();
   Empties the keyboard buffer.

extern int three_finger_flag;
   The djgpp keyboard handler provides an 'emergency exit' sequence which 
   you can use to kill off your program. If you are running under DOS this 
   is the three finger salute, ctrl+alt+del. Most multitasking OS's will 
   trap this combination before it reaches the Allegro handler, in which 
   case you can use the alternative ctrl+alt+end. If you want to disable 
   this behaviour in release versions of your program, set this flag to 
   FALSE.

extern int key_led_flag;
   By default, the capslock, numlock, and scroll-lock keys toggle the 
   keyboard LED indicators when they are pressed. If you are using these 
   keys for input in your game (eg. capslock to fire) this may not be 
   desirable, so you can clear this flag to prevent the LED's being updated.



===========================================
============ Joystick routines ============
===========================================

int install_joystick(int type);
   Initialises the joystick, and calibrates the centre position value. The 
   type parameter should usually be JOY_TYPE_AUTODETECT, or see the platform 
   specific documentation for a list of the available drivers. You must call 
   this routine before using any other joystick functions, and you should 
   make sure that the joystick is in the middle position at the time. 
   Returns zero on success. As soon as you have installed the joystick 
   module, you will be able to read the button state and digital (on/off 
   toggle) direction information, which may be enough for some games. If you 
   want to get full analogue input, though, you need to use the 
   calibrate_joystick() functions to measure the exact range of the inputs: 
   see below.

void remove_joystick();
   Removes the joystick handler. You don't normally need to bother calling 
   this, because allegro_exit() will do it for you.

int poll_joystick();
   The joystick is not interrupt driven, so you need to call this function 
   every now and again to update the global position values. Returns zero
   on success or a negative number on failure (usually because no joystick
   driver was installed).

extern int num_joysticks;
   Global variable containing the number of active joystick devices. The 
   current drivers support a maximum of four controllers.

extern JOYSTICK_INFO joy[n];
   Global array of joystick state information, which is updated by the 
   poll_joystick() function. Only the first num_joysticks elements will 
   contain meaningful information. The JOYSTICK_INFO structure is defined as:

   typedef struct JOYSTICK_INFO
   {
      int flags;                       - status flags for this
                                         joystick
      int num_sticks;                  - how many stick inputs?
      int num_buttons;                 - how many buttons?
      JOYSTICK_STICK_INFO stick[n];    - stick state information
      JOYSTICK_BUTTON_INFO button[n];  - button state information
   } JOYSTICK_INFO;

   The button status is stored in the structure:

   typedef struct JOYSTICK_BUTTON_INFO
   {
      int b;                           - boolean on/off flag
      char *name;                      - description of this
                                         button
   } JOYSTICK_BUTTON_INFO;

   You may wish to display the button names as part of an input 
   configuration screen to let the user choose what game function will be 
   performed by each button, but in simpler situations you can safely assume 
   that the first two elements in the button array will always be the main 
   trigger controls.

   Each joystick will provide one or more stick inputs, of varying types. 
   These can be digital controls which snap to specific positions (eg. a 
   gamepad controller, the coolie hat on a Flightstick Pro or Wingman 
   Extreme, or a normal joystick which hasn't yet been calibrated), or they 
   can be full analogue inputs with a smooth range of motion. Sticks may 
   also have different numbers of axis, for example a normal directional 
   control has two, but the Flightstick Pro throttle is only a single axis, 
   and it is possible that the system could be extended in the future to 
   support full 3d controllers. A stick input is described by the structure:

   typedef struct JOYSTICK_STICK_INFO
   {
      int flags;                       - status flags for this
                                         input
      int num_axis;                    - how many axis do we
                                         have?
      JOYSTICK_AXIS_INFO axis[n];      - axis state information
      char *name;                      - description of this
                                         input
   } JOYSTICK_STICK_INFO;

   A single joystick may provide several different stick inputs, but you can 
   safely assume that the first element in the stick array will always be 
   the main directional controller.

   Information about each of the stick axis is stored in the substructure:

   typedef struct JOYSTICK_AXIS_INFO
   {
      int pos;                         - analogue axis position
      int d1, d2;                      - digital axis position
      char *name;                      - description of this axis
   } JOYSTICK_AXIS_INFO;

   This provides both analogue input in the pos field (ranging from -128 to 
   128 or from 0 to 255, depending on the type of the control), and digital 
   values in the d1 and d2 fields. For example, when describing the X-axis 
   position, the pos field will hold the horizontal position of the 
   joystick, d1 will be set if it is moved left, and d2 will be set if it is 
   moved right. Allegro will fill in all these values regardless of whether 
   it is using a digital or analogue joystick, emulating the pos field for 
   digital inputs by snapping it to the min, middle, and maximum positions, 
   and emulating the d1 and d2 values for an analogue stick by comparing the 
   current position with the centre point.

   The joystick flags field may contain any combination of the bit flags:

   JOYFLAG_DIGITAL
      This control is currently providing digital input.

   JOYFLAG_ANALOGUE
      This control is currently providing analogue input.

   JOYFLAG_CALIB_DIGITAL
      This control will be capable of providing digital input once it has 
      been calibrated, but is not doing this at the moment.

   JOYFLAG_CALIB_ANALOGUE
      This control will be capable of providing analogue input once it has 
      been calibrated, but is not doing this at the moment.

   JOYFLAG_CALIBRATE
      Indicates that this control needs to be calibrated. Many devices 
      require multiple calibration steps, so you should call the 
      calibrate_joystick() function from a loop until this flag is cleared.

   JOYFLAG_SIGNED
      Indicates that the analogue axis position is in signed format, ranging 
      from -128 to 128. This is the case for all 2d directional controls.

   JOYFLAG_UNSIGNED
      Indicates that the analogue axis position is in unsigned format, 
      ranging from 0 to 255. This is the case for all 1d throttle controls.

   Note for people who spell funny: in case you don't like having to type 
   "analogue", there are some #define aliases in allegro/joystick.h that
   will allow you to write "analog" instead.

const char *calibrate_joystick_name(int n);
   Returns a text description for the next type of calibration that will be 
   done on the specified joystick, or NULL if no more calibration is 
   required.

int calibrate_joystick(int n);
   Most joysticks need to be calibrated before they can provide full 
   analogue input. This function performs the next operation in the 
   calibration series for the specified stick, assuming that the joystick 
   has been positioned in the manner described by a previous call to 
   calibrate_joystick_name(), returning zero on success. For example, a 
   simple routine to fully calibrate all the joysticks might look like:

      int i;

      for (i=0; i<;num_joysticks; i++) {
         while (joy[i].flags & JOYFLAG_CALIBRATE) {
            char *msg = calibrate_joystick_name(i);
            printf("%s, and press a key\n", msg);
            readkey();
            if (calibrate_joystick(i) != 0) {
               printf("oops!\n");
               exit(1);
            }
         }
      }

int save_joystick_data(const char *filename);
   After all the headache of calibrating the joystick, you may not want to 
   make your poor users repeat the process every time they run your program. 
   Call this function to save the joystick calibration data into the 
   specified configuration file, from which it can later be read by 
   load_joystick_data(). Pass a NULL filename to write the data to the 
   currently selected configuration file. Returns zero on success.

int load_joystick_data(const char *filename);
   Restores calibration data previously stored by save_joystick_data() or 
   the setup utility. This sets up all aspects of the joystick code: you 
   don't even need to call install_joystick() if you are using this 
   function. Pass a NULL filename to read the data from the currently 
   selected configuration file. Returns zero on success: if it fails the 
   joystick state is undefined and you must reinitialise it from scratch.

int initialise_joystick();
   Deprecated. Use install_joystick() instead.



========================================
============ Graphics modes ============
========================================

void set_color_depth(int depth);
   Sets the pixel format to be used by subsequent calls to set_gfx_mode() 
   and create_bitmap(). Valid depths are 8 (the default), 15, 16, 24, and 32 
   bits. Note that you can retrieve the pixel format currently in use by
   calling bitmap_color_depth() on the 'screen' bitmap, once a graphics mode
   has been set.

void request_refresh_rate(int rate);
   Requests that the next call to set_gfx_mode() try to use the specified 
   refresh rate, if possible. Not all drivers are able to control this at 
   all, and even when they can, not all rates will be possible on all 
   hardware, so the actual settings may differ from what you requested. 
   After you call set_gfx_mode(), you can use get_refresh_rate() to find out 
   what was actually selected. At the moment only the DOS VESA 3.0, X DGA 2.0
   and some Windows DirectX drivers support this function. The speed is
   specified in Hz, eg. 60, 70. To return to the normal default selection,
   pass a rate value of zero.

int get_refresh_rate(void);
   Returns the current refresh rate, if known (not all drivers are able to 
   report this information). Returns zero if the actual rate is unknown.

GFX_MODE_LIST *get_gfx_mode_list(int card);
   Attempts to create a list of all the supported video modes for a certain
   graphics driver. This function returns a pointer to a list structure of 
   the type GFX_MODE_LIST which has the following definition:

   typedef struct GFX_MODE_LIST {
      int num_modes;
      GFX_MODE *mode;
   } GFX_MODE_LIST;

   If this function returns NULL, it means the call failed. The mode entry
   points to the actual list of video modes.

   typedef struct GFX_MODE {
      int width, height, bpp;
   } GFX_MODE;

   This list is terminated with an { 0, 0, 0 } entry.

   Note that the card parameter must refer to a _real_ driver. This function
   fails if you pass GFX_SAFE, GFX_AUTODETECT, or any other "magic" driver.

void destroy_gfx_mode_list(GFX_MODE_LIST *mode_list);
   Removes the mode list created by get_gfx_mode_list() from memory.

int set_gfx_mode(int card, int w, int h, int v_w, int v_h);
   Switches into graphics mode. The card parameter should usually be 
   GFX_AUTODETECT, GFX_AUTODETECT_FULLSCREEN or GFX_AUTODETECT_WINDOWED, or
   see the platform specific documentation for a list of the available
   drivers. The w and h parameters specify what screen resolution you want.

   The v_w and v_h parameters specify the minimum virtual screen size, in 
   case you need a large virtual screen for hardware scrolling or page 
   flipping. You should set them to zero if you don't care about the virtual 
   screen size. Virtual screens can cause a lot of confusion, but they are 
   really quite simple. Warning: patronising explanation coming up, so you 
   may wish to skip the rest of this paragraph :-) Think of video memory as 
   a rectangular piece of paper which is being viewed through a small hole 
   (your monitor) in a bit of cardboard. Since the paper is bigger than the 
   hole you can only see part of it at any one time, but by sliding the 
   cardboard around you can alter which portion of the image is visible. You 
   could just leave the hole in one position and ignore the parts of video 
   memory that aren't visible, but you can get all sorts of useful effects 
   by sliding the screen window around, or by drawing images in a hidden 
   part of video memory and then flipping across to display them.

   For example, you could select a 640x480 mode in which the monitor acts as 
   a window onto a 1024x1024 virtual screen, and then move the visible 
   screen around in this larger area. Initially, with the visible screen 
   positioned at the top left corner of video memory, this setup would look 
   like:

      (0,0)------------(640,0)----(1024,0)
        |                  |           |
        |  visible screen  |           |
        |                  |           |
      (0,480)----------(640,480)       |
        |                              |
        |   the rest of video memory   |
        |                              |
      (0,1024)--------------------(1024,1024)

   What's that? You are viewing this with a proportional font? Hehehe.

   When you call set_gfx_mode(), the v_w and v_h parameters represent the 
   minimum size of virtual screen that is acceptable for your program. The 
   range of possible sizes is usually very restricted, and Allegro is likely 
   to end up creating a virtual screen much larger than the one you request. 
   On an SVGA card with one megabyte of vram you can count on getting a 
   1024x1024 virtual screen (256 colors) or 1024x512 (15 or 16 bpp), and 
   with 512k vram you can get 1024x512 (256 color). Other sizes may or may 
   not be possible: don't assume that they will work. In mode-X the virtual 
   width can be any multiple of eight greater than or equal to the physical 
   screen width, and the virtual height will be set accordingly (the VGA has 
   256k of vram, so the virtual height will be 256*1024/virtual_width).

   After you select a graphics mode, the physical and virtual screen sizes 
   can be checked with the macros SCREEN_W, SCREEN_H, VIRTUAL_W, and 
   VIRTUAL_H.

   If Allegro is unable to select an appropriate mode, set_gfx_mode() 
   returns a negative number and stores a description of the problem in 
   allegro_error. Otherwise it returns zero.

   As a special case, if you use the magic driver code GFX_SAFE, Allegro 
   will guarantee that the mode will always be set correctly. It will try to 
   select the resolution that you request, and if that fails, it will fall 
   back upon whatever mode is known to be reliable on the current platform 
   (this is 320x200 VGA mode under DOS, a 640x480 resolution under Windows,
   the actual framebuffer's resolution under Linux if it's supported, etc).
   If it absolutely cannot set any graphics mode at all, it will return
   negative as usual, meaning that there's no possible video output on the
   machine, and that you should abort your program immediately, possibly
   after notifying this to the user with allegro_message. This fake driver
   is useful for situations where you just want to get into some kind of
   workable display mode, and can't be bothered with trying multiple
   different resolutions and doing all the error checking yourself. Note
   however, that after a successful call to set_gfx_mode with this driver,
   you cannot make any assumptions about the width, height or color depth
   of the screen: your code will have to deal with this little detail.

   Finally, GFX_TEXT is another magic driver which usually closes any
   previously opened graphic mode, making you unable to use the global
   variable screen, and in those environments that have text modes, sets one
   previously used or the closest match to that (usually 80x25). With this
   driver the size parameters of set_gfx_mode don't mean anything, so you can
   leave them all to zero or any other number you prefer.

int set_display_switch_mode(int mode);
   Sets how the program should handle being switched into the background, 
   if the user tabs away from it. Not all of the possible modes will be 
   supported by every graphics driver on every platform: you must call this 
   routine after initializing the graphics driver and if you request a mode 
   that isn't currently possible, it will return -1. The available modes are:

   SWITCH_NONE
      Disables switching. This is the default in single-tasking systems like 
      DOS. It may be supported on other platforms, but you should use it 
      with caution, because your users won't be impressed if they want to 
      tab away from your program, but you don't let them!

   SWITCH_PAUSE
      Pauses the program whenever it is in the background. Execution will be 
      resumed as soon as the user switches back to it. This is the default 
      in most fullscreen multitasking environments, for example the Linux 
      console, but not under Windows.

   SWITCH_AMNESIA
      Like SWITCH_PAUSE, but this mode doesn't bother to remember the 
      contents of video memory, so the screen, and any video bitmaps that 
      you have created, will be erased after the user switches away and then 
      back to your program. This is not a terribly useful mode to have, but 
      it is the default for the fullscreen drivers under Windows because 
      DirectDraw is too dumb to implement anything better.

   SWITCH_BACKGROUND
      The program will carry on running in the background, with the screen 
      bitmap temporarily being pointed at a memory buffer for the fullscreen 
      drivers. You must take special care when using this mode, because bad 
      things will happen if the screen bitmap gets changed around when your 
      program isn't expecting it (see below).

   SWITCH_BACKAMNESIA
      Like SWITCH_BACKGROUND, but this mode doesn't bother to remember the 
      contents of video memory (see SWITCH_AMNESIA). It is again the only 
      mode supported by the fullscreen drivers under Windows that lets the 
      program keep running in the background.

   Note that you should be very careful when you are using graphics routines 
   in the switching context: you must always call acquire_screen() before the
   start of any drawing code onto the screen and not release it until you are
   completely finished, because the automatic locking mechanism may not be
   good enough to work when the program runs in the background or has just
   been raised in the foreground.

int set_display_switch_callback(int dir, void (*cb)());
   Installs a notification callback for the switching mode that was 
   previously selected by calling set_display_switch_mode(). The direction 
   parameter can either be SWITCH_IN or SWITCH_OUT, depending whether you 
   want to be notified about switches away from your program or back to your 
   program. You can sometimes install callbacks for both directions at the 
   same time, but not every platform supports this, so this function may
   return -1 if your request is impossible. You can install several switch
   callbacks at the same time.

void remove_display_switch_callback(void (*cb)());
   Removes a notification callback that was previously installed by calling 
   set_display_switch_callback(). All the callbacks will automatically be 
   removed when you call set_display_switch_mode().

int get_display_switch_mode();
   Returns the current display switching mode, in the same format passed to 
   set_display_switch_mode().

extern int gfx_capabilities;
   Bitfield describing the capabilities of the current graphics driver and 
   video hardware. This may contain combination any of the flags:

   GFX_CAN_SCROLL:
      Indicates that the scroll_screen() function may be used with this 
      driver.

   GFX_CAN_TRIPLE_BUFFER:
      Indicates that the request_scroll() and poll_scroll() functions may be 
      used with this driver. If this flag is not set, it is possible that 
      the enable_triple_buffer() function may be able to activate it.

   GFX_HW_CURSOR:
      Indicates that a hardware mouse cursor is in use. When this flag is 
      set, it is safe to draw onto the screen without hiding the mouse 
      pointer first. Note that not every cursor graphic can be implemented 
      in hardware: in particular VBE/AF only supports 2-color images up to 
      32x32 in size, where the second color is an exact inverse of the 
      first. This means that Allegro may need to switch between hardware and 
      software cursors at any point during the execution of your program, so 
      you should not assume that this flag will remain constant for long 
      periods of time. It only tells you whether a hardware cursor is in use 
      at the current time, and may change whenever you hide/redisplay the 
      pointer.

   GFX_HW_HLINE:
      Indicates that the normal opaque version of the hline() function is 
      implemented using a hardware accelerator. This will improve the 
      performance not only of hline() itself, but also of many other 
      functions that use it as a workhorse, for example circlefill(), 
      triangle(), and floodfill().

   GFX_HW_HLINE_XOR:
      Indicates that the XOR version of the hline() function, and any other 
      functions that use it as a workhorse, are implemented using a hardware 
      accelerator.

   GFX_HW_HLINE_SOLID_PATTERN:
      Indicates that the solid and masked pattern modes of the hline() 
      function, and any other functions that use it as a workhorse, are 
      implemented using a hardware accelerator (see note below).

   GFX_HW_HLINE_COPY_PATTERN:
      Indicates that the copy pattern mode of the hline() function, and any 
      other functions that use it as a workhorse, are implemented using a 
      hardware accelerator (see note below).

   GFX_HW_FILL:
      Indicates that the opaque version of the rectfill() function, the 
      clear_bitmap() routine, and clear_to_color(), are implemented using a
      hardware accelerator.

   GFX_HW_FILL_XOR:
      Indicates that the XOR version of the rectfill() function is 
      implemented using a hardware accelerator.

   GFX_HW_FILL_SOLID_PATTERN:
      Indicates that the solid and masked pattern modes of the rectfill() 
      function are implemented using a hardware accelerator (see note below).

   GFX_HW_FILL_COPY_PATTERN:
      Indicates that the copy pattern mode of the rectfill() function is 
      implemented using a hardware accelerator (see note below).

   GFX_HW_LINE:
      Indicates that the opaque mode line() and vline() functions are 
      implemented using a hardware accelerator.

   GFX_HW_LINE_XOR:
      Indicates that the XOR version of the line() and vline() functions are 
      implemented using a hardware accelerator.

   GFX_HW_TRIANGLE:
      Indicates that the opaque mode triangle() function is implemented 
      using a hardware accelerator.

   GFX_HW_TRIANGLE_XOR:
      Indicates that the XOR version of the triangle() function is 
      implemented using a hardware accelerator.

   GFX_HW_GLYPH:
      Indicates that monochrome character expansion (for text drawing) is 
      implemented using a hardware accelerator.

   GFX_HW_VRAM_BLIT:
      Indicates that blitting from one part of the screen to another is 
      implemented using a hardware accelerator. If this flag is set, 
      blitting within the video memory will almost certainly be the fastest 
      possible way to display an image, so it may be worth storing some of 
      your more frequently used graphics in an offscreen portion of the 
      video memory.

   GFX_HW_VRAM_BLIT_MASKED:
      Indicates that the masked_blit() routine is capable of a hardware 
      accelerated copy from one part of video memory to another, and that 
      draw_sprite() will use a hardware copy when given a sub-bitmap of the 
      screen or a video memory bitmap as the source image. If this flag is 
      set, copying within the video memory will almost certainly be the 
      fastest possible way to display an image, so it may be worth storing 
      some of your more frequently used sprites in an offscreen portion of 
      the video memory.

      Warning: if this flag is not set, masked_blit() and draw_sprite() will 
      not work correctly when used with a video memory source image! You 
      must only try to use these functions to copy within the video memory 
      if they are supported in hardware.

   GFX_HW_MEM_BLIT:
      Indicates that blitting from a memory bitmap onto the screen is being 
      accelerated in hardware.

   GFX_HW_MEM_BLIT_MASKED:
      Indicates that the masked_blit() and draw_sprite() functions are being 
      accelerated in hardware when the source image is a memory bitmap and 
      the destination is the physical screen.

   GFX_HW_SYS_TO_VRAM_BLIT:
      Indicates that blitting from a system bitmap onto the screen is being 
      accelerated in hardware. Note that some acceleration may be present 
      even if this flag is not set, because system bitmaps can benefit from 
      normal memory to screen blitting as well. This flag will only be set 
      if system bitmaps have further acceleration above and beyond what is 
      provided by GFX_HW_MEM_BLIT.

   GFX_HW_SYS_TO_VRAM_BLIT_MASKED:
      Indicates that the masked_blit() and draw_sprite() functions are being 
      accelerated in hardware when the source image is a system bitmap and 
      the destination is the physical screen. Note that some acceleration 
      may be present even if this flag is not set, because system bitmaps 
      can benefit from normal memory to screen blitting as well. This flag 
      will only be set if system bitmaps have further acceleration above and 
      beyond what is provided by GFX_HW_MEM_BLIT_MASKED.

   Note: even if the capabilities information says that patterned drawing is 
   supported by the hardware, it will not be possible for every size of 
   pattern. VBE/AF only supports patterns up to 8x8 in size, so Allegro will 
   fall back on the original non-accelerated drawing routines whenever you 
   use a pattern larger than this.

   Note2: these hardware acceleration features will only take effect when 
   you are drawing directly onto the screen bitmap, a video memory bitmap, 
   or a sub-bitmap thereof. Accelerated hardware is most useful in a page 
   flipping or triple buffering setup, and is unlikely to make any 
   difference to the classic "draw onto a memory bitmap, then blit to the 
   screen" system.

int enable_triple_buffer();
   If the GFX_CAN_TRIPLE_BUFFER bit of the gfx_capabilities field is not 
   set, you can attempt to enable it by calling this function. In particular 
   if you are running in mode-X in a clean DOS environment, this routine 
   will enable the timer retrace simulator, which will activate the triple 
   buffering functions. Returns zero if triple buffering is enabled.

int scroll_screen(int x, int y);
   Attempts to scroll the hardware screen to display a different part of the 
   virtual screen (initially it will be positioned at 0, 0, which is the top 
   left corner). Returns zero on success: it may fail if the graphics driver 
   can't handle hardware scrolling or the virtual screen isn't large enough. 
   You can use this to move the screen display around in a large virtual 
   screen space, or to page flip back and forth between two non-overlapping 
   areas of the virtual screen. Note that to draw outside the original 
   position in the screen bitmap you will have to alter the clipping 
   rectangle: see below.

   Mode-X scrolling is reliable and will work on any card. Unfortunately 
   most VESA implementations can only handle horizontal scrolling in four 
   pixel increments, so smooth horizontal panning is impossible in SVGA 
   modes. This is a shame, but I can't see any way round it. A significant 
   number of VESA implementations seem to be very buggy when it comes to 
   scrolling in truecolor video modes, so I suggest you don't depend on this 
   routine working correctly in the truecolor resolutions unless you can be 
   sure that SciTech Display Doctor is installed.

   Allegro will handle any necessary vertical retrace synchronisation when 
   scrolling the screen, so you don't need to call vsync() before it. This 
   means that scroll_screen() has the same time delay effects as vsync().

int request_scroll(int x, int y);
   This function is used for triple buffering. It requests a hardware scroll 
   to the specified position, but returns immediately rather than waiting 
   for a retrace. The scroll will then take place during the next vertical 
   retrace, but you can carry on running other code in the meantime and use 
   the poll_scroll() routine to detect when the flip has actually taken 
   place (see examples/ex3buf.c). Triple buffering is only possible on 
   certain hardware: it will work in any mode-X resolution if the timer 
   retrace simulator is active (but this doesn't work correctly under 
   win95), plus it is supported by the VBE 3.0 and VBE/AF drivers for a 
   limited number of high-end graphics cards. You can look at the 
   GFX_CAN_TRIPLE_BUFFER bit in the gfx_capabilities flag to see if it will 
   work with the current driver. This function returns zero on success.

int poll_scroll();
   This function is used for triple buffering. It checks the status of a 
   hardware scroll previously initiated by the request_scroll() routine, 
   returning non-zero if it is still waiting to take place, and zero if it 
   has already happened.

int show_video_bitmap(BITMAP *bitmap);
   Attempts to page flip the hardware screen to display the specified video 
   bitmap object, which must be the same size as the physical screen, and 
   should have been obtained by calling the create_video_bitmap() function. 
   Returns zero on success and non-zero on failure.
   
   Allegro will handle any necessary vertical retrace synchronisation when 
   page flipping, so you don't need to call vsync() before it. This means
   that show_video_bitmap() has the same time delay effects as vsync().

int request_video_bitmap(BITMAP *bitmap);
   This function is used for triple buffering. It requests a page flip to 
   display the specified video bitmap object, but returns immediately rather 
   than waiting for a retrace. The flip will then take place during the next 
   vertical retrace, but you can carry on running other code in the meantime 
   and use the poll_scroll() routine to detect when the flip has actually 
   taken place. Triple buffering is only possible on certain hardware: see 
   the comments about request_scroll(). Returns zero on success.



========================================
============ Bitmap objects ============
========================================

Once you have selected a graphics mode, you can draw things onto the display 
via the 'screen' bitmap. All the Allegro graphics routines draw onto BITMAP 
structures, which are areas of memory containing rectangular images, stored 
as packed byte arrays (in 8 bit modes one byte per pixel, in 15 and 16 bit 
modes sizeof(short) bytes per pixel, in 24 bit modes 3 bytes per pixel and 
in 32 bit modes sizeof(long) bytes per pixel). You can create and manipulate 
bitmaps in system RAM, or you can write to the special 'screen' bitmap which 
represents the video memory in your graphics card.

For example, to draw a pixel onto the screen you would write:

   putpixel(screen, x, y, color);

Or to implement a double-buffered system:

   BITMAP *bmp = create_bitmap(320, 200);    // make a bitmap in system RAM
   clear_bitmap(bmp);                        // zero the memory bitmap
   putpixel(bmp, x, y, color);               // draw onto the memory bitmap
   blit(bmp, screen, 0, 0, 0, 0, 320, 200);  // copy it to the screen

See below for information on how to get direct access to the image memory in 
a bitmap.

Allegro supports several different types of bitmaps:

   - The screen bitmap, which represents the hardware video memory. 
     Ultimately you have to draw onto this in order for your image to be 
     visible. It is destroyed by any subsequent calls to set_gfx_mode().

   - Memory bitmaps, which are located in system RAM and can be used to 
     store graphics or as temporary drawing spaces for double buffered 
     systems. These can be obtained by calling create_bitmap(), load_pcx(), 
     or by loading a grabber datafile.

   - Sub-bitmaps. These share image memory with a parent bitmap (which 
     can be the screen, a memory bitmap, or another sub-bitmap), so drawing 
     onto them will also change their parent. They can be of any size and 
     located anywhere within the parent bitmap, and can have their own 
     clipping rectangles, so they are a useful way of dividing a bitmap into 
     several smaller units, eg. splitting a large virtual screen into 
     multiple sections (see examples/exscroll.c).

   - Video memory bitmaps. These are created by the create_video_bitmap() 
     function, and are usually implemented as sub-bitmaps of the screen 
     object. They must be destroyed by destroy_bitmap() before any subsequent
     calls to set_gfx_mode().

   - System bitmaps. These are created by the create_system_bitmap() 
     function, and are a sort of halfway house between memory and video 
     bitmaps. They live in system memory, so you aren't limited by the 
     amount of video ram in your card, but they are stored in a 
     platform-specific format that may enable better hardware acceleration 
     than is possible with a normal memory bitmap (see the 
     GFX_HW_SYS_TO_VRAM_BLIT and GFX_HW_SYS_TO_VRAM_BLIT_MASKED flags in 
     gfx_capabilities). System bitmaps must be accessed in the same way as 
     video bitmaps, using the bank switch functions and bmp_write*() macros. 
     Not every platform implements this type of bitmap: if they aren't 
     available, create_system_bitmap() will function identically to 
     create_bitmap(). They must be destroyed by destroy_bitmap() before any
     subsequent calls to set_gfx_mode().

extern BITMAP *screen;
   Global pointer to a bitmap, sized VIRTUAL_W x VIRTUAL_H. This is created 
   by set_gfx_mode(), and represents the hardware video memory. Only a part 
   of this bitmap will actually be visible, sized SCREEN_W x SCREEN_H. 
   Normally this is the top left corner of the larger virtual screen, so you 
   can ignore the extra invisible virtual size of the bitmap if you aren't 
   interested in hardware scrolling or page flipping. To move the visible 
   window to other parts of the screen bitmap, call scroll_screen(). 
   Initially the clipping rectangle will be limited to the physical screen 
   size, so if you want to draw onto a larger virtual screen space outside 
   this rectangle, you will need to adjust the clipping.

BITMAP *create_bitmap(int width, int height);
   Creates a memory bitmap sized width by height, and returns a pointer to 
   it. The bitmap will have clipping turned on, and the clipping rectangle 
   set to the full size of the bitmap. The image memory will not be cleared, 
   so it will probably contain garbage: you should clear the bitmap before 
   using it. This routine always uses the global pixel format, as specified 
   by calling set_color_depth().

BITMAP *create_bitmap_ex(int color_depth, int width, int height);
   Creates a bitmap in a specific color depth (8, 15, 16, 24 or 32 bits per 
   pixel).

BITMAP *create_sub_bitmap(BITMAP *parent, int x, y, width, height);
   Creates a sub-bitmap, ie. a bitmap sharing drawing memory with a 
   pre-existing bitmap, but possibly with a different size and clipping 
   settings. When creating a sub-bitmap of the mode-X screen, the x position 
   must be a multiple of four. The sub-bitmap width and height can extend 
   beyond the right and bottom edges of the parent (they will be clipped), 
   but the origin point must lie within the parent region.

BITMAP *create_video_bitmap(int width, int height);
   Allocates a video memory bitmap of the specified size, returning a 
   pointer to it on success or NULL on failure (ie. if you have run out of 
   vram). This can be used to allocate offscreen video memory for storing 
   source graphics ready for a hardware accelerated blitting operation, or 
   to create multiple video memory pages which can then be displayed by 
   calling show_video_bitmap(). Video memory bitmaps are usually allocated 
   from the same space as the screen bitmap, so they may overlap with it: it 
   is not therefore a good idea to use the global screen at the same time as 
   any surfaces returned by this function.

BITMAP *create_system_bitmap(int width, int height);
   Allocates a system memory bitmap of the specified size, returning a 
   pointer to it on success or NULL on failure.

void destroy_bitmap(BITMAP *bitmap);
   Destroys a memory bitmap, sub-bitmap, video memory bitmap, or system 
   bitmap when you are finished with it.

void lock_bitmap(BITMAP *bitmap);
   Under DOS, locks all the memory used by a bitmap. You don't normally need
   to call this function unless you are doing very weird things in your
   program.

int bitmap_color_depth(BITMAP *bmp);
   Returns the color depth of the specified bitmap (8, 15, 16, 24, or 32).
   Note that calling it on the 'screen' bitmap will return the pixel format
   currently in use, as specified by the latest call to set_color_depth(),
   once a graphics mode has been set.

int bitmap_mask_color(BITMAP *bmp);
   Returns the mask color for the specified bitmap (the value which is 
   skipped when drawing sprites). For 256 color bitmaps this is zero, and 
   for truecolor bitmaps it is bright pink (maximum red and blue, zero 
   green).

int is_same_bitmap(BITMAP *bmp1, BITMAP *bmp2);
   Returns TRUE if the two bitmaps describe the same drawing surface, ie. 
   the pointers are equal, one is a sub-bitmap of the other, or they are 
   both sub-bitmaps of a common parent.

int is_linear_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is a linear bitmap, ie. a memory bitmap, mode 13h 
   screen, or SVGA screen. Linear bitmaps can be used with the _putpixel(), 
   _getpixel(), bmp_write_line(), and bmp_read_line() functions.

int is_planar_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is a planar (mode-X or Xtended mode) screen bitmap.

int is_memory_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is a memory bitmap, ie. it was created by calling 
   create_bitmap() or loaded from a grabber datafile or image file. Memory 
   bitmaps can be accessed directly via the line pointers in the bitmap 
   structure, eg. bmp->line[y][x] = color.

int is_screen_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is the screen bitmap, or a sub-bitmap of the screen.

int is_video_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is the screen bitmap, a video memory bitmap, or a 
   sub-bitmap of either.

int is_system_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is a system bitmap object, or a sub-bitmap of one.

int is_sub_bitmap(BITMAP *bmp);
   Returns TRUE if bmp is a sub-bitmap.

void acquire_bitmap(BITMAP *bmp);
   Locks the specified video memory bitmap prior to drawing onto it. This 
   does not apply to memory bitmaps, and only affects some platforms 
   (Windows needs it, DOS does not). These calls are not strictly required, 
   because the drawing routines will automatically acquire the bitmap before 
   accessing it, but locking a DirectDraw surface is very slow, so you will 
   get much better performance if you acquire the screen just once at the 
   start of your main redraw function, and only release it when the drawing 
   is completely finished. Multiple acquire calls may be nested, and the 
   bitmap will only be truly released when the lock count returns to zero. 
   Be warned that DirectX programs activate a mutex lock whenever a surface 
   is locked, which prevents them from getting any input messages, so you 
   must be sure to release all your bitmaps before using any timer, 
   keyboard, or other non-graphics routines!

void release_bitmap(BITMAP *bmp);
   Releases a bitmap that was previously locked by calling acquire_bitmap(). 
   If the bitmap was locked multiple times, you must release it the same 
   number of times before it will truly be unlocked.

void acquire_screen();
   Shortcut version of acquire_bitmap(screen);

void release_screen();
   Shortcut version of release_bitmap(screen);

void set_clip(BITMAP *bitmap, int x1, int y1, int x2, int y2);
   Each bitmap has an associated clipping rectangle, which is the area of 
   the image that it is ok to draw on. Nothing will be drawn to positions 
   outside this space. Pass the two opposite corners of the clipping 
   rectangle: these are inclusive, eg. set_clip(bitmap, 16, 16, 32, 32) will 
   allow drawing to (16, 16) and (32, 32), but not to (15, 15) and (33, 33). 
   If x1, y1, x2, and y2 are all zero, clipping will be turned off, which 
   may slightly speed up some drawing operations (usually a negligible 
   difference, although every little helps) but will result in your program 
   dying a horrible death if you try to draw beyond the edges of the bitmap.



=============================================
============ Loading image files ============
=============================================

Warning: when using truecolor images, you should always set the graphics 
mode before loading any bitmap data! Otherwise the pixel format (RGB or BGR)
will not be known, so the file may be converted wrongly.

BITMAP *load_bitmap(const char *filename, RGB *pal);
   Loads a bitmap from a file, returning a pointer to a bitmap and storing 
   the palette data in the specified location, which should be an array of 
   256 RGB structures. You are responsible for destroying the bitmap when 
   you are finished with it. Returns NULL on error. At present this function 
   supports BMP, LBM, PCX, and TGA files, determining the type from the file 
   extension.

   If the file contains a truecolor image, you must set the video mode or
   call set_color_conversion() before loading it. In this case, if the
   destination color depth is 8-bit, the palette will be generated by calling
   generate_optimized_palette() on the bitmap; otherwise, the returned
   palette will be generated by calling generate_332_palette().

   The pal argument may be NULL. In this case, the palette data are simply
   not returned. Additionally, if the file is a truecolor image and the
   destination color depth is 8-bit, the color conversion process will use
   the current palette instead of generating an optimized one.

BITMAP *load_bmp(const char *filename, RGB *pal);
   Loads a 256 color or 24 bit truecolor Windows or OS/2 BMP file.

BITMAP *load_lbm(const char *filename, RGB *pal);
   Loads a 256 color IFF ILBM/PBM file.

BITMAP *load_pcx(const char *filename, RGB *pal);
   Loads a 256 color or 24 bit truecolor PCX file.

BITMAP *load_tga(const char *filename, RGB *pal);
   Loads a 256 color, 15 bit hicolor, 24 bit truecolor, or 32 bit 
   truecolor+alpha TGA file.

int save_bitmap(const char *filename, BITMAP *bmp, const RGB *pal);
   Writes a bitmap into a file, using the specified palette, which should be 
   an array of 256 RGB structures. Returns non-zero on error. The output 
   format is determined from the filename extension: at present this 
   function supports BMP, PCX and TGA formats.

   Two things to watch out for: on some video cards it may be faster to copy
   the screen to a memory bitmap and save the latter, and if you use this to
   dump the screen into a file you may end up with an image much larger than
   you were expecting, because Allegro often creates virtual screens larger
   than the visible screen. You can get around this by using a sub-bitmap to
   specify which part of the screen to save, eg:

      BITMAP *bmp;
      PALETTE pal;

      get_palette(pal);
      bmp = create_sub_bitmap(screen, 0, 0, SCREEN_W, SCREEN_H);
      save_bitmap("dump.pcx", bmp, pal);
      destroy_bitmap(bmp);

int save_bmp(const char *filename, BITMAP *bmp, const RGB *pal);
   Writes a bitmap into a 256 color or 24 bit truecolor BMP file.

int save_pcx(const char *filename, BITMAP *bmp, const RGB *pal);
   Writes a bitmap into a 256 color or 24 bit truecolor PCX file.

int save_tga (const char *filename, BITMAP *bmp, const RGB *pal);
   Writes a bitmap into a 256 color, 15 bit hicolor, 24 bit truecolor, or 32 
   bit truecolor+alpha TGA file.

void register_bitmap_file_type(const char *ext,
          BITMAP *(*load)(const char *filename, RGB *pal),
          int (*save)(const char *filename, BITMAP *bmp, const RGB *pal));
   Informs the load_bitmap() and save_bitmap() functions of a new file type, 
   providing routines to read and write images in this format (either 
   function may be NULL).

void set_color_conversion(int mode);
   Specifies how to convert images between the various color depths when 
   reading graphics from external bitmap files or datafiles. The mode is a 
   bitmask specifying which types of conversion are allowed. If the 
   appropriate bit is set, data will be converted into the current pixel 
   format (selected by calling the set_color_depth() function), otherwise it 
   will be left in the same format as the disk file, leaving you to convert 
   it manually before the graphic can be displayed. The default mode is 
   total conversion, so that all images will be loaded in the appropriate 
   format for the current video mode. Valid bit flags are:

      COLORCONV_NONE                // disable all format
                                    // conversions
      COLORCONV_8_TO_15             // expand 8 bits to 15 bits
      COLORCONV_8_TO_16             // expand 8 bits to 16 bits
      COLORCONV_8_TO_24             // expand 8 bits to 24 bits
      COLORCONV_8_TO_32             // expand 8 bits to 32 bits
      COLORCONV_15_TO_8             // reduce 15 bits to 8 bits
      COLORCONV_15_TO_16            // expand 15 bits to 16 bits
      COLORCONV_15_TO_24            // expand 15 bits to 24 bits
      COLORCONV_15_TO_32            // expand 15 bits to 32 bits
      COLORCONV_16_TO_8             // reduce 16 bits to 8 bits
      COLORCONV_16_TO_15            // reduce 16 bits to 15 bits
      COLORCONV_16_TO_24            // expand 16 bits to 24 bits
      COLORCONV_16_TO_32            // expand 16 bits to 32 bits
      COLORCONV_24_TO_8             // reduce 24 bits to 8 bits
      COLORCONV_24_TO_15            // reduce 24 bits to 15 bits
      COLORCONV_24_TO_16            // reduce 24 bits to 16 bits
      COLORCONV_24_TO_32            // expand 24 bits to 32 bits
      COLORCONV_32_TO_8             // reduce 32 bit RGB to 8 bits
      COLORCONV_32_TO_15            // reduce 32 bit RGB to 15 bits
      COLORCONV_32_TO_16            // reduce 32 bit RGB to 16 bits
      COLORCONV_32_TO_24            // reduce 32 bit RGB to 24 bits
      COLORCONV_32A_TO_8            // reduce 32 bit RGBA to 8 bits
      COLORCONV_32A_TO_15           // reduce 32 bit RGBA to 15 bits
      COLORCONV_32A_TO_16           // reduce 32 bit RGBA to 16 bits
      COLORCONV_32A_TO_24           // reduce 32 bit RGBA to 24 bits
      COLORCONV_DITHER_PAL          // dither when reducing to 8 bit
      COLORCONV_DITHER_HI           // dither when reducing to
                                    // hicolor
      COLORCONV_KEEP_TRANS          // keep original transparency

   For convenience, the following macros can be used to select common 
   combinations of these flags:

      COLORCONV_EXPAND_256          // expand 256 colors to
                                    // hi/truecolor
      COLORCONV_REDUCE_TO_256       // reduce hi/truecolor to 256
                                    // colors
      COLORCONV_EXPAND_15_TO_16     // expand 15 bit hicolor to
                                    // 16 bits
      COLORCONV_REDUCE_16_TO_15     // reduce 16 bit hicolor to
                                    // 15 bits
      COLORCONV_EXPAND_HI_TO_TRUE   // expand 15/16 bits to
                                    // 24/32 bits
      COLORCONV_REDUCE_TRUE_TO_HI   // reduce 24/32 bits to
                                    // 15/16 bits
      COLORCONV_24_EQUALS_32        // convert between 24 and
                                    // 32 bits
      COLORCONV_TOTAL               // everything to current format
      COLORCONV_PARTIAL             // convert 15 <-> 16 and
                                    // 24 <-> 32
      COLORCONV_MOST                // all but hi/truecolor <-> 256
      COLORCONV_DITHER              // dither during all color
                                    // reductions

   If you enable the COLORCONV_DITHER flag, dithering will be performed 
   whenever truecolor graphics are converted into a hicolor or paletted 
   format, including by the blit() function, and any automatic conversions 
   that take place while reading graphics from disk. This can produce much 
   better looking results, but is obviously slower than a direct conversion.

   If you intend using converted bitmaps with functions like masked_blit() 
   or draw_sprite(), you should specify the COLORCONV_KEEP_TRANS flag. It 
   will ensure that the masked areas in the bitmap before and after the 
   conversion stay exactly the same, by mapping transparent colors to each 
   other and adjusting colors which would be converted to the transparent 
   color otherwise. It affects every blit() operation between distinct pixel 
   formats and every automatic conversion.



==========================================
============ Palette routines ============
==========================================

All the Allegro drawing functions use integer parameters to represent 
colors. In truecolor resolutions these numbers encode the color directly as 
a collection of red, green, and blue bits, but in a regular 256 color mode 
the values are treated as indexes into the current palette, which is a table 
listing the red, green and blue intensities for each of the 256 possible 
colors.

Palette entries are stored in an RGB structure, which contains red, green 
and blue intensities in the VGA hardware format, ranging from 0-63, and is 
defined as:

typedef struct RGB
{
   unsigned char r, g, b;
} RGB;

For example:

   RGB black = { 0,  0,  0  };
   RGB white = { 63, 63, 63 };
   RGB green = { 0,  63, 0  };
   RGB grey  = { 32, 32, 32 };

The type PALETTE is defined to be an array of 256 RGB structures.

You may notice that a lot of the code in Allegro spells 'palette' as 
'pallete'. This is because the headers from my old Mark Williams compiler on 
the Atari spelt it with two l's, so that is what I'm used to. Allegro will 
happily accept either spelling, due to some #defines in allegro/alcompat.h.

void vsync();
   Waits for a vertical retrace to begin. The retrace happens when the 
   electron beam in your monitor has reached the bottom of the screen and is 
   moving back to the top ready for another scan. During this short period 
   the graphics card isn't sending any data to the monitor, so you can do 
   things to it that aren't possible at other times, such as altering the 
   palette without causing flickering (snow). Allegro will automatically 
   wait for a retrace before altering the palette or doing any hardware 
   scrolling, though, so you don't normally need to bother with this 
   function.

void set_color(int index, const RGB *p);
   Sets the specified palette entry to the specified RGB triplet. Unlike the 
   other palette functions this doesn't do any retrace synchronisation, so 
   you should call vsync() before it to prevent snow problems.

void _set_color(int index, const RGB *p);
   This is an inline version of set_color(), intended for use in the 
   vertical retrace simulator callback function. It should only be used in 
   VGA mode 13h and mode-X, because some of the more recent SVGA chipsets 
   aren't VGA compatible (set_color() and set_palette() will use VESA calls 
   on these cards, but _set_color() doesn't know about that).

void set_palette(const PALETTE p);
   Sets the entire palette of 256 colors. You should provide an array of 256 
   RGB structures. Unlike set_color(), there is no need to call vsync() 
   before this function.

void set_palette_range(const PALETTE p, int from, int to, int vsync);
   Sets the palette entries between from and to (inclusive: pass 0 and 255 
   to set the entire palette). If vsync is set it waits for the vertical 
   retrace, otherwise it sets the colors immediately.

void get_color(int index, RGB *p);
   Retrieves the specified palette entry.

void get_palette(PALETTE p);
   Retrieves the entire palette of 256 colors. You should provide an array 
   of 256 RGB structures to store it in.

void get_palette_range(PALETTE p, int from, int to);
   Retrieves the palette entries between from and to (inclusive: pass 0 and 
   255 to get the entire palette).

void fade_interpolate(const PALETTE source, const PALETTE dest,
                      PALETTE output, int pos, int from, to);
   Calculates a temporary palette part way between source and dest, 
   returning it in the output parameter. The position between the two 
   extremes is specified by the pos value: 0 returns an exact copy of 
   source, 64 returns dest, 32 returns a palette half way between the two, 
   etc. This routine only affects colors between from and to (inclusive: 
   pass 0 and 255 to interpolate the entire palette).

void fade_from_range(const PALETTE source, const PALETTE dest,
                     int speed, int from, to);
   Gradually fades a part of the palette from the source palette to the dest 
   palette. The speed is from 1 (the slowest) up to 64 (instantaneous). This 
   routine only affects colors between from and to (inclusive: pass 0 and 
   255 to fade the entire palette).

void fade_in_range(const PALETTE p, int speed, int from, to);
   Gradually fades a part of the palette from a black screen to the 
   specified palette. The speed is from 1 (the slowest) up to 64 
   (instantaneous). This routine only affects colors between from and to 
   (inclusive: pass 0 and 255 to fade the entire palette).

void fade_out_range(int speed, int from, to);
   Gradually fades a part of the palette from the current palette to a black 
   screen. The speed is from 1 (the slowest) up to 64 (instantaneous). This 
   routine only affects colors between from and to (inclusive: pass 0 and 
   255 to fade the entire palette).

void fade_from(const PALETTE source, const PALETTE dest, int speed);
   Fades gradually from the source palette to the dest palette. The speed is 
   from 1 (the slowest) up to 64 (instantaneous).

void fade_in(const PALETTE p, int speed);
   Fades gradually from a black screen to the specified palette. The speed 
   is from 1 (the slowest) up to 64 (instantaneous).

void fade_out(int speed);
   Fades gradually from the current palette to a black screen. The speed is 
   from 1 (the slowest) up to 64 (instantaneous).

void select_palette(const PALETTE p);
   Ugly hack for use in various dodgy situations where you need to convert 
   between paletted and truecolor image formats. Sets the internal palette 
   table in the same way as the set_palette() function, so the conversion 
   will use the specified palette, but without affecting the display 
   hardware in any way. The previous palette settings are stored in an 
   internal buffer, and can be restored by calling unselect_palette().

void unselect_palette();
   Restores the palette tables that were in use before the last call to 
   select_palette().

void generate_332_palette(PALETTE pal);
   Constructs a fake truecolor palette, using three bits for red and green 
   and two for the blue. The load_bitmap() function returns this if the file 
   does not contain a palette itself (ie. you are reading a truecolor 
   bitmap).

int generate_optimized_palette(BITMAP *bmp, PALETTE pal,
                               const char rsvd[256]);
   Generates a 256 color palette suitable for making a reduced color version 
   of the specified truecolor image. The rsvd parameter points to a table 
   indicating which colors it is allowed to modify: zero for free colors
   which may be set to whatever the optimiser likes, negative values for
   reserved colors which cannot be used, and positive values for fixed
   palette entries that must not be changed, but can be used in the
   optimisation.

extern PALETTE default_palette;
   The default IBM BIOS palette. This will be automatically selected 
   whenever you set a new graphics mode.

extern PALETTE black_palette;
   A palette containing solid black colors, used by the fade routines.

extern PALETTE desktop_palette;
   The palette used by the Atari ST low resolution desktop. I'm not quite 
   sure why this is still here, except that the grabber and test programs 
   use it. It is probably the only Atari legacy code left in Allegro, and it 
   would be a shame to remove it :-)



=================================================
============ Truecolor pixel formats ============
=================================================

In a truecolor video mode the red, green, and blue components for each pixel 
are packed directly into the color value, rather than using a palette lookup 
table. In a 15 bit mode there are 5 bits for each color, in 16 bit modes 
there are 5 bits each of red and blue and six bits of green, and both 24 and 
32 bit modes use 8 bits for each color (the 32 bit pixels simply have an 
extra padding byte to align the data nicely). The layout of these components 
can vary depending on your hardware, but will generally either be RGB or 
BGR. Since the layout is not known until you select the video mode you will 
be using, you must call set_gfx_mode() before using any of the following 
routines!

int makecol8(int r, int g, int b);
int makecol15(int r, int g, int b);
int makecol16(int r, int g, int b);
int makecol24(int r, int g, int b);
int makecol32(int r, int g, int b);
   These functions convert colors from a hardware independent form (red, 
   green, and blue values ranging 0-255) into various display dependent 
   pixel formats. Converting to 15, 16, 24, or 32 bit formats only takes a 
   few shifts, so it is fairly efficient. Converting to an 8 bit color 
   involves searching the palette to find the closest match, which is quite 
   slow unless you have set up an RGB mapping table (see below).

int makeacol32(int r, int g, int b, int a);
   Converts an RGBA color into a 32 bit display pixel format, which includes 
   an alpha (transparency) value. There are no versions of this routine for 
   other color depths, because only the 32 bit format has enough room to 
   store a proper alpha channel. You should only use RGBA format colors as 
   the input to draw_trans_sprite() or draw_trans_rle_sprite() after calling 
   set_alpha_blender(), rather than drawing them directly to the screen.

int makecol(int r, int g, int b);
   Converts colors from a hardware independent format (red, green, and blue 
   values ranging 0-255) to the pixel format required by the current video 
   mode, calling the preceding 8, 15, 16, 24, or 32 bit makecol functions as 
   appropriate.

int makecol_depth(int color_depth, int r, int g, int b);
   Converts colors from a hardware independent format (red, green, and blue 
   values ranging 0-255) to the pixel format required by the specified color 
   depth.

int makeacol(int r, int g, int b, int a);
int makeacol_depth(int color_depth, int r, int g, int b, int a);
   Convert RGBA colors into display dependent pixel formats. In anything 
   less than a 32 bit mode, these are the same as calling makecol() or 
   makecol_depth(), but by using these routines it is possible to create 32 
   bit color values that contain a true 8 bit alpha channel along with the 
   red, green, and blue components. You should only use RGBA format colors 
   as the input to draw_trans_sprite() or draw_trans_rle_sprite() after 
   calling set_alpha_blender(), rather than drawing them directly to the 
   screen.

int makecol15_dither(int r, int g, int b, int x, int y);
int makecol16_dither(int r, int g, int b, int x, int y);
   Given both a color value and a pixel coordinate, calculate a dithered 15 
   or 16 bit RGB value. This can produce better results when reducing images 
   from truecolor to hicolor. In addition to calling these functions 
   directly, hicolor dithering can be automatically enabled when loading 
   graphics by calling the set_color_conversion() function, for example 
   set_color_conversion (COLORCONV_REDUCE_TRUE_TO_HI | COLORCONV_DITHER).

int getr8(int c);
int getg8(int c);
int getb8(int c);
int getr15(int c);
int getg15(int c);
int getb15(int c);
int getr16(int c);
int getg16(int c);
int getb16(int c);
int getr24(int c);
int getg24(int c);
int getb24(int c);
int getr32(int c);
int getg32(int c);
int getb32(int c);
   Given a color in a display dependent format, these functions extract one 
   of the red, green, or blue components (ranging 0-255).

int geta32(int c);
   Given a color in a 32 bit pixel format, this function extracts the alpha 
   component (ranging 0-255).

int getr(int c);
int getg(int c);
int getb(int c);
int geta(int c);
   Given a color in the format being used by the current video mode, these 
   functions extract one of the red, green, blue, or alpha components 
   (ranging 0-255), calling the preceding 8, 15, 16, 24, or 32 bit get 
   functions as appropriate. The alpha part is only meaningful for 32 bit 
   pixels.

int getr_depth(int color_depth, int c);
int getg_depth(int color_depth, int c);
int getb_depth(int color_depth, int c);
int geta_depth(int color_depth, int c);
   Given a color in the format being used by the specified color depth, 
   these functions extract one of the red, green, blue, or alpha components 
   (ranging 0-255). The alpha part is only meaningful for 32 bit pixels.

extern int palette_color[256];
   Table mapping palette index colors (0-255) into whatever pixel format is 
   being used by the current display mode. In a 256 color mode this just 
   maps onto the array index. In truecolor modes it looks up the specified 
   entry in the current palette, and converts that RGB value into the 
   appropriate packed pixel format.

#define MASK_COLOR_8 0
#define MASK_COLOR_15 (5.5.5 pink)
#define MASK_COLOR_16 (5.6.5 pink)
#define MASK_COLOR_24 (8.8.8 pink)
#define MASK_COLOR_32 (8.8.8 pink)
   Constants representing the colors used to mask transparent sprite pixels 
   for each color depth. In 256 color resolutions this is zero, and in 
   truecolor modes it is bright pink (maximum red and blue, zero green).



============================================
============ Drawing primitives ============
============================================

Except for _putpixel(), all these routines are affected by the current 
drawing mode and the clipping rectangle of the destination bitmap.

void putpixel(BITMAP *bmp, int x, int y, int color);
   Writes a pixel to the specified position in the bitmap, using the current 
   drawing mode and the bitmap's clipping rectangle.

void _putpixel(BITMAP *bmp, int x, int y, int color);
void _putpixel15(BITMAP *bmp, int x, int y, int color);
void _putpixel16(BITMAP *bmp, int x, int y, int color);
void _putpixel24(BITMAP *bmp, int x, int y, int color);
void _putpixel32(BITMAP *bmp, int x, int y, int color);
   Like the regular putpixel(), but much faster because they are implemented 
   as an inline assembler functions for specific color depths. These won't 
   work in mode-X graphics modes, don't perform any clipping (they will 
   crash if you try to draw outside the bitmap!), and ignore the drawing 
   mode.

int getpixel(BITMAP *bmp, int x, int y);
   Reads a pixel from point x, y in the bitmap. Returns -1 if the point lies 
   outside the bitmap.

int _getpixel(BITMAP *bmp, int x, int y);
int _getpixel15(BITMAP *bmp, int x, int y);
int _getpixel16(BITMAP *bmp, int x, int y);
int _getpixel24(BITMAP *bmp, int x, int y);
int _getpixel32(BITMAP *bmp, int x, int y);
   Faster inline versions of getpixel() for specific color depths. These 
   won't work in mode-X, and don't do any clipping, so you must make sure 
   the point lies inside the bitmap.

void vline(BITMAP *bmp, int x, int y1, int y2, int color);
   Draws a vertical line onto the bitmap, from point (x, y1) to (x, y2).

void hline(BITMAP *bmp, int x1, int y, int x2, int color);
   Draws a horizontal line onto the bitmap, from point (x1, y) to (x2, y).

void do_line(BITMAP *bmp, int x1, y1, x2, y2, int d, 
             void (*proc)(BITMAP *bmp, int x, int y, int d));
   Calculates all the points along a line from point (x1, y1) to (x2, y2), 
   calling the supplied function for each one. This will be passed a copy of 
   the bmp parameter, the x and y position, and a copy of the d parameter, 
   so it is suitable for use with putpixel().

void line(BITMAP *bmp, int x1, int y1, int x2, int y2, int color);
   Draws a line onto the bitmap, from point (x1, y1) to (x2, y2).

void triangle(BITMAP *bmp, int x1, y1, x2, y2, x3, y3, int color);
   Draws a filled triangle between the three points.

void polygon(BITMAP *bmp, int vertices, const int *points, int color);
   Draws a filled polygon with an arbitrary number of corners. Pass the 
   number of vertices and an array containing a series of x, y points (a 
   total of vertices*2 values).

void rect(BITMAP *bmp, int x1, int y1, int x2, int y2, int color);
   Draws an outline rectangle with the two points as its opposite corners.

void rectfill(BITMAP *bmp, int x1, int y1, int x2, int y2, int color);
   Draws a solid, filled rectangle with the two points as its opposite 
   corners.

void do_circle(BITMAP *bmp, int x, int y, int radius, int d,
               void (*proc)(BITMAP *bmp, int x, int y, int d));
   Calculates all the points in a circle around point (x, y) with radius r, 
   calling the supplied function for each one. This will be passed a copy of 
   the bmp parameter, the x and y position, and a copy of the d parameter, 
   so it is suitable for use with putpixel().

void circle(BITMAP *bmp, int x, int y, int radius, int color);
   Draws a circle with the specified centre and radius.

void circlefill(BITMAP *bmp, int x, int y, int radius, int color);
   Draws a filled circle with the specified centre and radius.

void do_ellipse(BITMAP *bmp, int x, int y, int rx, ry, int d,
                void (*proc)(BITMAP *bmp, int x, int y, int d));
   Calculates all the points in an ellipse around point (x, y) with radius 
   rx and ry, calling the supplied function for each one. This will be 
   passed a copy of the bmp parameter, the x and y position, and a copy of 
   the d parameter, so it is suitable for use with putpixel().

void ellipse(BITMAP *bmp, int x, int y, int rx, int ry, int color);
   Draws an ellipse with the specified centre and radius.

void ellipsefill(BITMAP *bmp, int x, int y, int rx, int ry, int color);
   Draws a filled ellipse with the specified centre and radius.

void do_arc(BITMAP *bmp, int x, int y, fixed a1, fixed a2, int r, int d,
            void (*proc)(BITMAP *bmp, int x, int y, int d));
   Calculates all the points in a circular arc around point (x, y) with 
   radius r, calling the supplied function for each one. This will be passed 
   a copy of the bmp parameter, the x and y position, and a copy of the d 
   parameter, so it is suitable for use with putpixel(). The arc will be 
   plotted in an anticlockwise direction starting from the angle a1 and 
   ending when it reaches a2. These values are specified in 16.16 fixed 
   point format, with 256 equal to a full circle, 64 a right angle, etc. 
   Zero is to the right of the centre point, and larger values rotate 
   anticlockwise from there.

void arc(BITMAP *bmp, int x, y, fixed ang1, ang2, int r, int color);
   Draws a circular arc with centre x, y and radius r, in an anticlockwise 
   direction starting from the angle a1 and ending when it reaches a2. These 
   values are specified in 16.16 fixed point format, with 256 equal to a 
   full circle, 64 a right angle, etc. Zero is to the right of the centre 
   point, and larger values rotate anticlockwise from there.

void calc_spline(const int points[8], int npts, int *x, int *y);
   Calculates a series of npts values along a bezier spline, storing them in 
   the output x and y arrays. The bezier curve is specified by the four x/y 
   control points in the points array: points[0] and points[1] contain the 
   coordinates of the first control point, points[2] and points[3] are the 
   second point, etc. Control points 0 and 3 are the ends of the spline, and 
   points 1 and 2 are guides. The curve probably won't pass through points 1 
   and 2, but they affect the shape of the curve between points 0 and 3 (the 
   lines p0-p1 and p2-p3 are tangents to the spline). The easiest way to 
   think of it is that the curve starts at p0, heading in the direction of 
   p1, but curves round so that it arrives at p3 from the direction of p2. 
   In addition to their role as graphics primitives, spline curves can be 
   useful for constructing smooth paths around a series of control points, 
   as in exspline.c.

void spline(BITMAP *bmp, const int points[8], int color);
   Draws a bezier spline using the four control points specified in the 
   points array.

void floodfill(BITMAP *bmp, int x, int y, int color);
   Floodfills an enclosed area, starting at point (x, y), with the specified 
   color.



==============================================
============ Blitting and sprites ============
==============================================

All these routines are affected by the clipping rectangle of the destination 
bitmap.

void clear_bitmap(BITMAP *bitmap);
   Clears the bitmap to color 0.

void clear(BITMAP *bitmap);
   An alias for clear_bitmap(), provided for backwards compatibility. It is
   implemented as a static inline function. The aliasing may be switched off
   by defining the preprocessor symbol ALLEGRO_NO_CLEAR_BITMAP_ALIAS before
   including Allegro headers, eg:

      #define ALLEGRO_NO_CLEAR_BITMAP_ALIAS
      #include <allegro.h>

void clear_to_color(BITMAP *bitmap, int color);
   Clears the bitmap to the specified color.

void blit(BITMAP *source, BITMAP *dest, int source_x, int source_y,
          int dest_x, int dest_y, int width, int height);
   Copies a rectangular area of the source bitmap to the destination bitmap. 
   The source_x and source_y parameters are the top left corner of the area 
   to copy from the source bitmap, and dest_x and dest_y are the 
   corresponding position in the destination bitmap. This routine respects 
   the destination clipping rectangle, and it will also clip if you try to 
   blit from areas outside the source bitmap.

   You can blit between any parts of any two bitmaps, even if the two memory 
   areas overlap (ie. source and dest are the same, or one is sub-bitmap of 
   the other). You should be aware, however, that a lot of SVGA cards don't 
   provide separate read and write banks, which means that blitting from one 
   part of the screen to another requires the use of a temporary bitmap in 
   memory, and is therefore extremely slow. As a general rule you should 
   avoid blitting from the screen onto itself in SVGA modes.

   In mode-X, on the other hand, blitting from one part of the screen to 
   another can be significantly faster than blitting from memory onto the 
   screen, as long as the source and destination are correctly aligned with 
   each other. Copying between overlapping screen rectangles is slow, but if 
   the areas don't overlap, and if they have the same plane alignment (ie. 
   (source_x%4) == (dest_x%4)), the VGA latch registers can be used for a 
   very fast data transfer. To take advantage of this, in mode-X it is often 
   worth storing tile graphics in a hidden area of video memory (using a 
   large virtual screen), and blitting them from there onto the visible part 
   of the screen.

   If the GFX_HW_VRAM_BLIT bit in the gfx_capabilities flag is set, the 
   current driver supports hardware accelerated blits from one part of the 
   screen onto another. This is extremely fast, so when this flag is set it 
   may be worth storing some of your more frequently used graphics in an 
   offscreen portion of the video memory.

   Unlike most of the graphics routines, blit() allows the source and 
   destination bitmaps to be of different color depths, so it can be used to 
   convert images from one pixel format to another.

void stretch_blit(BITMAP *source, BITMAP *dest,
                  int source_x, source_y, source_width, source_height,
                  int dest_x, dest_y, dest_width, dest_height);
   Like blit(), except it can scale images (so the source and destination 
   rectangles don't need to be the same size) and requires the source and 
   destination bitmaps to be of the same color depth. This routine doesn't 
   do as much safety checking as the regular blit(): in particular you must 
   take care not to copy from areas outside the source bitmap, and you 
   cannot blit between overlapping regions, ie. you must use different 
   bitmaps for the source and the destination. Moreover, the source must 
   be a memory bitmap.

void masked_blit(BITMAP *source, BITMAP *dest, int source_x, int source_y,
                  int dest_x, int dest_y, int width, int height);
   Like blit(), but skips transparent pixels, which are marked by a zero in 
   256 color modes or bright pink for truecolor data (maximum red and blue, 
   zero green), and requires the source and destination bitmaps to be of 
   the same color depth. The source and destination regions must not overlap.

   If the GFX_HW_VRAM_BLIT_MASKED bit in the gfx_capabilities flag is set, 
   the current driver supports hardware accelerated masked blits from one 
   part of the screen onto another. This is extremely fast, so when this 
   flag is set it may be worth storing some of your more frequently used 
   sprites in an offscreen portion of the video memory.

   Warning: if the hardware acceleration flag is not set, masked_blit() will 
   not work correctly when used with a source image in system or video 
   memory so the latter must be a memory bitmap.

void masked_stretch_blit(BITMAP *source, BITMAP *dest,
                         int source_x, source_y, source_w, source_h,
                         int dest_x, dest_y, dest_w, dest_h);
   Like masked_blit(), except it can scale images (so the source and 
   destination rectangles don't need to be the same size). This routine 
   doesn't do as much safety checking as the regular masked_blit(): in 
   particular you must take care not to copy from areas outside the source 
   bitmap. Moreover, the source must be a memory bitmap.

void draw_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y);
   Draws a copy of the sprite bitmap onto the destination bitmap at the 
   specified position. This is almost the same as blit(sprite, bmp, 0, 0, x, 
   y, sprite->w, sprite->h), but it uses a masked drawing mode where 
   transparent pixels are skipped, so the background image will show through 
   the masked parts of the sprite. Transparent pixels are marked by a zero 
   in 256 color modes or bright pink for truecolor data (maximum red and 
   blue, zero green).

   If the GFX_HW_VRAM_BLIT_MASKED bit in the gfx_capabilities flag is set, 
   the current driver supports hardware accelerated sprite drawing when the 
   source image is a video memory bitmap or a sub-bitmap of the screen. This 
   is extremely fast, so when this flag is set it may be worth storing some 
   of your more frequently used sprites in an offscreen portion of the video 
   memory.

   Warning: if the hardware acceleration flag is not set, draw_sprite() will 
   not work correctly when used with a sprite image in system or video 
   memory so the latter must be a memory bitmap.

   Although generally not supporting graphics of mixed color depths, as a 
   special case this function can be used to draw 256 color source images 
   onto truecolor destination bitmaps, so you can use palette effects on 
   specific sprites within a truecolor program.

void stretch_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y, int w, int h);
   Like draw_sprite(), except it can stretch the sprite image to the 
   specified width and height and requires the sprite image and destination 
   bitmap to be of the same color depth. Moreover, the sprite image must 
   be a memory bitmap.

void draw_sprite_v_flip(BITMAP *bmp, BITMAP *sprite, int x, int y);
void draw_sprite_h_flip(BITMAP *bmp, BITMAP *sprite, int x, int y);
void draw_sprite_vh_flip(BITMAP *bmp, BITMAP *sprite, int x, int y);
   These are like draw_sprite(), but they flip the image about the vertical, 
   horizontal, or both, axes. This produces exact mirror images, which
   is not the same as rotating the sprite (and it is a lot faster than the 
   rotation routine). The sprite must be a memory bitmap.

void draw_trans_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y);
   Uses the global color_map table or truecolor blender functions to overlay 
   the sprite on top of the existing image. This must only be used after you 
   have set up the color mapping table (for 256 color modes) or blender 
   functions (for truecolor modes). Because it involves reading as well as 
   writing the bitmap memory, translucent drawing is very slow if you draw 
   directly to video RAM, so wherever possible you should use a memory 
   bitmap instead. The bitmap and sprite must normally be in the same color 
   depth, but as a special case you can draw 32 bit RGBA format sprites onto 
   any hicolor or truecolor bitmap, as long as you call set_alpha_blender() 
   first, and you can draw 8 bit alpha images onto a 32 bit RGBA 
   destination, as long as you call set_write_alpha_blender() first.

void draw_lit_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y, int color);
   In 256 color modes, uses the global color_map table to tint the sprite
   image to the specified color or to light it to the level specified by
   'color', depending on the function which was used to build the table
   (create_trans_table or create_light_table), and draws the resulting image
   to the destination bitmap. In truecolor modes, uses the blender functions
   to light the sprite image using the alpha level specified by 'color' (the
   alpha level which was passed to the blender functions is ignored) and
   draws the resulting image to the destination bitmap. The 'color' parameter
   must be in the range [0-255] whatever its actual meaning is. This must
   only be used after you have set up the color mapping table (for 256 color
   modes) or blender functions (for truecolor modes).

void draw_gouraud_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y,
                         int c1, int c2, int c3, int c4);
   More sophisticated version of draw_lit_sprite(): the 'color' parameter is
   not constant across the sprite image anymore but interpolated between the
   four specified corner colors, which have the same actual meaning as it.

void draw_character(BITMAP *bmp, BITMAP *sprite, int x, int y, int color);
   Draws a copy of the sprite bitmap onto the destination bitmap at the 
   specified position, drawing transparent pixels in the current text mode 
   (skipping them if the text mode is -1, otherwise drawing them in the text 
   background color), and setting all other pixels to the specified color. 
   Transparent pixels are marked by a zero in 256 color modes or bright pink 
   for truecolor data (maximum red and blue, zero green). The sprite must be 
   an 8 bit image, even if the destination is a truecolor bitmap.

void rotate_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y, fixed angle);
   Draws the sprite image onto the bitmap. It is placed with its top left
   corner at the specified position, then rotated by the specified angle
   around its centre. The angle is a fixed point 16.16 number in the same
   format used by the fixed point trig routines, with 256 equal to a full
   circle, 64 a right angle, etc. All rotation functions can draw between any
   two bitmaps, even screen bitmaps or bitmaps of different color depth.

void rotate_sprite_v_flip(BITMAP *bmp, BITMAP *sprite, int x, int y, fixed angle);
   Like rotate_sprite, but also flips the image vertically.  To flip 
   horizontally, use this routine but add itofix(128) to the angle. To flip
   in both directions, use rotate_sprite() and add itofix(128) to its angle.

void rotate_scaled_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y,
                          fixed angle, fixed scale);
   Like rotate_sprite(), but stretches or shrinks the image at the same time 
   as rotating it.

void rotate_scaled_sprite_v_flip(BITMAP *bmp, BITMAP *sprite, int x, int y,
                                 fixed angle, fixed scale)
   Draws the sprite, similar to rotate_scaled_sprite() except that it flips
   the sprite vertically first.

void pivot_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y,
                  int cx, int cy, fixed angle);
   Like rotate_sprite(), but aligns the point in the sprite given by (cx, cy)
   to (x, y) in the bitmap, then rotates around this point.

void pivot_sprite_v_flip(BITMAP *bmp, BITMAP *sprite, int x, int y,
                         int cx, int cy, fixed angle);
   Like rotate_sprite_v_flip(), but aligns the point in the sprite given by
   (cx, cy) to (x, y) in the bitmap, then rotates around this point.

void pivot_scaled_sprite(BITMAP *bmp, BITMAP *sprite, int x, int y,
                         int cx, int cy, fixed angle, fixed scale));
   Like rotate_scaled_sprite(), but aligns the point in the sprite given by
   (cx, cy) to (x, y) in the bitmap, then rotates and scales around this
   point.

void pivot_scaled_sprite_v_flip(BITMAP *bmp, BITMAP *sprite, int x, int y,
                                 fixed angle, fixed scale)
   Like rotate_scaled_sprite_v_flip(), but aligns the point in the sprite
   given by (cx, cy) to (x, y) in the bitmap, then rotates and scales around
   this point.



=====================================
============ RLE sprites ============
=====================================

Because bitmaps can be used in so many different ways, the bitmap structure 
is quite complicated, and it contains a lot of data. In many situations, 
though, you will find yourself storing images that are only ever copied to 
the screen, rather than being drawn onto or used as filling patterns, etc. 
If this is the case you may be better off storing your images in RLE_SPRITE 
or COMPILED_SPRITE structures rather than regular bitmaps.

RLE sprites store the image in a simple run-length encoded format, where 
repeated zero pixels are replaced by a single length count, and strings of 
non-zero pixels are preceded by a counter giving the length of the solid 
run. RLE sprites are usually much smaller then normal bitmaps, both because 
of the run length compression, and because they avoid most of the overhead 
of the bitmap structure. They are often also faster than normal bitmaps, 
because rather than having to compare every single pixel with zero to 
determine whether it should be drawn, it is possible to skip over a whole 
run of zeros with a single add, or to copy a long run of non-zero pixels 
with fast string instructions.

Every silver lining has a cloud, though, and in the case of RLE sprites it 
is a lack of flexibility. You can't draw onto them, and you can't flip them, 
rotate them, or stretch them. In fact the only thing you can do with them is 
to blast them onto a bitmap with the draw_rle_sprite() function, which is 
equivalent to using draw_sprite() with a regular bitmap. You can convert 
bitmaps into RLE sprites at runtime, or you can create RLE sprite structures 
in grabber datafiles by making a new object of type 'RLE sprite'.

RLE_SPRITE *get_rle_sprite(BITMAP *bitmap);
   Creates an RLE sprite based on the specified bitmap (which must be a 
   memory bitmap).

void destroy_rle_sprite(RLE_SPRITE *sprite);
   Destroys an RLE sprite structure previously returned by get_rle_sprite().

void draw_rle_sprite(BITMAP *bmp, const RLE_SPRITE *sprite,
                     int x, int y);
   Draws an RLE sprite onto a bitmap at the specified position.

void draw_trans_rle_sprite(BITMAP *bmp, const RLE_SPRITE *sprite,
                           int x, int y);
   Translucent version of draw_rle_sprite(). See the description of 
   draw_trans_sprite(). This must only be used after you have set up the 
   color mapping table (for 256 color modes) or blender functions (for 
   truecolor modes). The bitmap and sprite must normally be in the same 
   color depth, but as a special case you can draw 32 bit RGBA format 
   sprites onto any hicolor or truecolor bitmap, as long as you call 
   set_alpha_blender() first.

void draw_lit_rle_sprite(BITMAP *bmp, const RLE_SPRITE *sprite,
                         int x, y, color);
   Tinted version of draw_rle_sprite(). See the description of
   draw_lit_sprite(). This must only be used after you have set up the color 
   mapping table (for 256 color modes) or blender functions (for truecolor 
   modes).



==========================================
============ Compiled sprites ============
==========================================

Compiled sprites are stored as actual machine code instructions that draw a 
specific image onto a bitmap, using mov instructions with immediate data 
values. This is the fastest way to draw a masked image: on my machine 
drawing compiled sprites is about five times as fast as using draw_sprite() 
with a regular bitmap. Compiled sprites are big, so if memory is tight you 
should use RLE sprites instead, and what you can do with them is even more 
restricted than with RLE sprites, because they don't support clipping. If 
you try to draw one off the edge of a bitmap, you will corrupt memory and 
probably crash the system. You can convert bitmaps into compiled sprites at 
runtime, or you can create compiled sprite structures in grabber datafiles 
by making a new object of type 'Compiled sprite' or 'Compiled x-sprite'.

COMPILED_SPRITE *get_compiled_sprite(BITMAP *bitmap, int planar);
   Creates a compiled sprite based on the specified bitmap (which must be a 
   memory bitmap). Compiled sprites are device-dependent, so you have to 
   specify whether to compile it into a linear or planar format. Pass FALSE 
   as the second parameter if you are going to be drawing it onto memory 
   bitmaps or mode 13h and SVGA screen bitmaps, and pass TRUE if you are 
   going to draw it onto mode-X or Xtended mode screen bitmaps.

void destroy_compiled_sprite(COMPILED_SPRITE *sprite);
   Destroys a compiled sprite structure previously returned by 
   get_compiled_sprite().

void draw_compiled_sprite(BITMAP *bmp, const COMPILED_SPRITE *sprite,
                          int x, int y);
   Draws a compiled sprite onto a bitmap at the specified position. The 
   sprite must have been compiled for the correct type of bitmap (linear or 
   planar). This function does not support clipping.

   Hint: if not being able to clip compiled sprites is a problem, a neat 
   trick is to set up a work surface (memory bitmap, mode-X virtual screen, 
   or whatever) a bit bigger than you really need, and use the middle of it 
   as your screen. That way you can draw slightly off the edge without any 
   trouble... 



=====================================
============ Text output ============
=====================================

Allegro provides text output routines that work with both monochrome and 
color fonts, which can contain any number of Unicode character ranges. The 
grabber program can create fonts from sets of characters drawn in a bitmap 
file (see grabber.txt for more information), and can also import GRX or BIOS 
format font files. The font structure contains a number of hooks that can be 
used to extend it with your own custom drawing code: see the definition in 
allegro/text.h for details.

extern FONT *font;
   A simple 8x8 fixed size font (the mode 13h BIOS default). If you want to 
   alter the font used by the GUI routines, change this to point to one of 
   your own fonts. This font contains the standard ASCII (U+20 to U+7F), 
   Latin-1 (U+A1 to U+FF), and Latin Extended-A (U+0100 to U+017F) character 
   ranges.

extern int allegro_404_char;
    When Allegro cannot find a glyph it needs in a font, it will instead
    output the character given in allegro_404_char. By default, this is
    set to the caret symbol, '^'.

int text_mode(int mode);
   Sets the mode in which text will be drawn. Returns previous mode.
   If mode is zero or positive, text output will be opaque and the 
   background of the characters will be set to color #mode. If mode is 
   negative, text will be drawn transparently (ie. the background of the 
   characters will not be altered). The default is a mode of zero.

void textout(BITMAP *bmp, const FONT *f, const char *s,
             int x, y, int color);
   Writes the string s onto the bitmap at position x, y, using the current 
   text mode and the specified font and foreground color. If the color is -1 
   and a color font is in use, it will be drawn using the colors from the 
   original font bitmap (the one you imported into the grabber program), 
   which allows multicolored text output.

void textout_centre(BITMAP *bmp, const FONT *f, const char *s,
                    int x, y, color);
   Like textout(), but interprets the x coordinate as the centre rather than 
   the left edge of the string.

void textout_right(BITMAP *bmp, const FONT *f, const char *s,
                   int x, y, color);
   Like textout(), but interprets the x coordinate as the right rather than 
   the left edge of the string.

void textout_justify(BITMAP *bmp, const FONT *f, const char *s,
                     int x1, int x2, int y, int diff, int color);
   Draws justified text within the region x1-x2. If the amount of spare 
   space is greater than the diff value, it will give up and draw regular 
   left justified text instead.

void textprintf(BITMAP *bmp, const FONT *f, int x, y, color,
                const char *fmt, ...);
   Formatted text output, using a printf() style format string.

void textprintf_centre(BITMAP *bmp, const FONT *f, int x, y, color,
                       const char *fmt, ...);
   Like textprintf(), but interprets the x coordinate as the centre rather 
   than the left edge of the string.

void textprintf_right(BITMAP *bmp, const FONT *f, int x, y, color,
                      const char *fmt, ...);
   Like textprintf(), but interprets the x coordinate as the right rather 
   than the left edge of the string.

void textprintf_justify(BITMAP *bmp, const FONT *f, int x1, int x2, int y,
                        int diff, int color, const char *fmt, ...);
   Like textout_justify, but using a printf() style format string.

int text_length(const FONT *f, const char *str);
   Returns the length (in pixels) of a string in the specified font.

int text_height(const FONT *f)
   Returns the height (in pixels) of the specified font.

void destroy_font(FONT *f);
   Frees the memory being used by a font structure.



===========================================
============ Polygon rendering ============
===========================================

void polygon3d(BITMAP *bmp, int type, BITMAP *texture, int vc, V3D *vtx[]);
void polygon3d_f(BITMAP *bmp, int type, BITMAP *texture, int vc, V3D_f *vtx[]);
   Draw 3d polygons onto the specified bitmap, using the specified rendering 
   mode. Unlike the regular polygon() function, these routines don't support 
   concave or self-intersecting shapes, and they can't draw onto mode-X 
   screen bitmaps (if you want to write 3d code in mode-X, draw onto a 
   memory bitmap and then blit to the screen). The width and height of the 
   texture bitmap must be powers of two, but can be different, eg. a 64x16 
   texture is fine, but a 17x3 one is not. The vertex count parameter (vc) 
   should be followed by an array containing the appropriate number of 
   pointers to vertex structures: polygon3d() uses the fixed point V3D 
   structure, while polygon3d_f() uses the floating point V3D_f structure. 
   These are defined as:

   typedef struct V3D
   {
      fixed x, y, z;       - position
      fixed u, v;          - texture map coordinates
      int c;               - color
   } V3D;

   typedef struct V3D_f
   {
      float x, y, z;       - position
      float u, v;          - texture map coordinates
      int c;               - color
   } V3D_f;

   How the vertex data is used depends on the rendering mode:

   The x and y values specify the position of the vertex in 2d screen 
   coordinates.

   The z value is only required when doing perspective correct texture 
   mapping, and specifies the depth of the point in 3d world coordinates.

   The u and v coordinates are only required when doing texture mapping, and 
   specify a point on the texture plane to be mapped on to this vertex. The 
   texture plane is an infinite plane with the texture bitmap tiled across 
   it. Each vertex in the polygon has a corresponding vertex on the texture 
   plane, and the image of the resulting polygon in the texture plane will 
   be mapped on to the polygon on the screen.

   We refer to pixels in the texture plane as texels. Each texel is a block, 
   not just a point, and whole numbers for u and v refer to the top-left 
   corner of a texel. This has a few implications. If you want to draw a 
   rectangular polygon and map a texture sized 32x32 on to it, you would use 
   the texture coordinates (0,0), (0,32), (32,32) and (32,0), assuming the 
   vertices are specified in anticlockwise order. The texture will then be 
   mapped perfectly on to the polygon. However, note that when we set u=32, 
   the last column of texels seen on the screen is the one at u=31, and the 
   same goes for v. This is because the coordinates refer to the top-left 
   corner of the texels. In effect, texture coordinates at the right and 
   bottom on the texture plane are exclusive.

   There is another interesting point here. If you have two polygons side 
   by side sharing two vertices (like the two parts of folded piece of 
   cardboard), and you want to map a texture across them seamlessly, the 
   values of u and v on the vertices at the join will be the same for both 
   polygons. For example, if they are both rectangular, one polygon may use 
   (0,0), (0,32), (32,32) and (32,0), and the other may use (32,0), (32,32), 
   (64,32), (64,0). This would create a seamless join.

   Of course you can specify fractional numbers for u and v to indicate a 
   point part-way across a texel. In addition, since the texture plane is 
   infinite, you can specify larger values than the size of the texture. 
   This can be used to tile the texture several times across the polygon.

   The c value specifies the vertex color, and is interpreted differently by 
   various rendering modes.

   The type parameter specifies the polygon rendering mode, and can be any 
   of the values:

   POLYTYPE_FLAT:
      A simple flat shaded polygon, taking the color from the c value of the 
      first vertex. This polygon type is affected by the drawing_mode() 
      function, so it can be used to render XOR or translucent polygons.

   POLYTYPE_GCOL:
      A single-color gouraud shaded polygon. The colors for each vertex are 
      taken from the c value, and interpolated across the polygon. This is 
      very fast, but will only work in 256 color modes if your palette 
      contains a smooth gradient between the colors. In truecolor modes it 
      interprets the color as a packed, display-format value as produced by 
      the makecol() function.

   POLYTYPE_GRGB:
      A gouraud shaded polygon which interpolates RGB triplets rather than a 
      single color. In 256 color modes this uses the global rgb_map table to 
      convert the result to an 8 bit paletted color, so it must only be used 
      after you have set up the RGB mapping table! The colors for each 
      vertex are taken from the c value, which is interpreted as a 24 bit 
      RGB triplet (0xFF0000 is red, 0x00FF00 is green, and 0x0000FF is 
      blue).

   POLYTYPE_ATEX:
      An affine texture mapped polygon. This stretches the texture across 
      the polygon with a simple 2d linear interpolation, which is fast but 
      not mathematically correct. It can look ok if the polygon is fairly 
      small or flat-on to the camera, but because it doesn't deal with 
      perspective foreshortening, it can produce strange warping artifacts. 
      To see what I mean, run test.exe and see what happens to the 
      polygon3d() test when you zoom in very close to the cube.

   POLYTYPE_PTEX:
      A perspective-correct texture mapped polygon. This uses the z value 
      from the vertex structure as well as the u/v coordinates, so textures 
      are displayed correctly regardless of the angle they are viewed from. 
      Because it involves division calculations in the inner texture mapping 
      loop, this mode is a lot slower than POLYTYPE_ATEX, and it uses 
      floating point so it will be very slow on anything less than a Pentium 
      (even with an FPU, a 486 can't overlap floating point division with 
      other integer operations like the Pentium can).

   POLYTYPE_ATEX_MASK:
   POLYTYPE_PTEX_MASK:
      Like POLYTYPE_ATEX and POLYTYPE_PTEX, but zero texture map pixels are 
      skipped, allowing parts of the texture map to be transparent.

   POLYTYPE_ATEX_LIT:
   POLYTYPE_PTEX_LIT:
      Like POLYTYPE_ATEX and POLYTYPE_PTEX, but the global color_map table 
      (for 256 color modes) or blender function (for non-MMX truecolor 
      modes) is used to blend the texture with a light level taken from the 
      c value in the vertex structure. This must only be used after you have 
      set up the color mapping table or blender functions!

   POLYTYPE_ATEX_MASK_LIT:
   POLYTYPE_PTEX_MASK_LIT:
      Like POLYTYPE_ATEX_LIT and POLYTYPE_PTEX_LIT, but zero texture map 
      pixels are skipped, allowing parts of the texture map to be 
      transparent.

   POLYTYPE_ATEX_TRANS:
   POLYTYPE_PTEX_TRANS:
      Render translucent textures. All the general rules for drawing
      translucent things apply. However, these modes have a major
      limitation: they only work with memory bitmaps or linear frame
      buffers (not with banked frame buffers). Don't even try, they do not
      check and your program will die horribly (or at least draw wrong
      things).

   POLYTYPE_ATEX_MASK_TRANS:
   POLYTYPE_PTEX_MASK_TRANS:
      Like POLYTYPE_ATEX_TRANS and POLYTYPE_PTEX_TRANS, but zero texture map 
      pixels are skipped.

   If the CPU_MMX flag of the cpu_capabilities global variable is set, the 
   GRGB and truecolor *LIT routines will be optimised using MMX 
   instructions. If the CPU_3DNOW flag is set, the truecolor PTEX*LIT 
   routines will take advantage of the 3DNow! CPU extensions.

   Using MMX for *LIT routines has a side effect: normally (without MMX), 
   these routines use the blender functions used also for other lighting 
   functions, set with set_trans_blender() or set_blender_mode(). The MMX 
   versions only use the RGB value passed to set_trans_blender() and do the 
   linear interpolation themselves. Therefore a new set of blender functions 
   passed to set_blender_mode() is ignored.

void triangle3d(BITMAP *bmp, int type, BITMAP *tex, V3D *v1, *v2, *v3);
void triangle3d_f(BITMAP *bmp, int type, BITMAP *tex, V3D_f *v1, *v2, *v3);
   Draw 3d triangles, using either fixed or floating point vertex structures.
   Unlike quad3d[_f], triangle3d[_f] functions are not wrappers of
   polygon3d[_f]. The triangle3d[_f] functions use their own routines taking
   into account the constantness of the gradients.
   Therefore triangle3d[_f](bmp, type, tex, v1, v2, v3) is faster than
   polygon3d[_f](bmp, type, tex, 3, v[]).

void quad3d(BITMAP *bmp, int type, BITMAP *tex, V3D *v1, *v2, *v3, *v4);
void quad3d_f(BITMAP *bmp, int type, BITMAP *tex, V3D_f *v1, *v2, *v3, *v4);
   Draw 3d quads, using either fixed or floating point vertex structures. 
   These are equivalent to calling
      polygon3d(bmp, type, tex, 4, v[]);
   or
      polygon3d_f(bmp, type, tex, 4, v[]);

int clip3d_f(int type, float min_z, float max_z, int vc,
             const V3D_f *vtx[], V3D_f *vout[], V3D_f *vtmp[], int out[]);
   Clips the polygon given in vtx. The number of vertices is vc, the result 
   goes in vout, and vtmp and out are needed for internal purposes. The 
   pointers in vtx, vout and vtmp must point to valid V3D_f structures. 
   As additional vertices may appear in the process of clipping, so the 
   size of vout, vtmp and out should be at least vc * (1.5 ^ n), where n is
   the number of clipping planes (5 or 6), and '^' denotes "to the power of".
   The frustum (viewing volume) is defined by -z<x<z, -z<y<z,
   0<min_z<z<max_z. If  max_z<=min_z, the z<max_z clipping is
   not done. As you can see, clipping is done in the camera space, with
   perspective in mind, so this routine should be called after you apply
   the camera matrix, but before the perspective projection. The routine
   will correctly interpolate u, v, and c in the vertex structure. However,
   no provision is made for high/truecolor GCOL.

int clip3d(int type, fixed min_z, fixed max_z, int vc,
           const V3D *vtx[], V3D *vout[], V3D *vtmp[], int out[]);
   Fixed point version of clip3d_f(). This function should be used with 
   caution, due to the limited precision of fixed point arithmetic and high 
   chance of rounding errors: the floating point code is better for most 
   situations.

A Z-buffer stores the depth of each pixel that is drawn on a viewport.
When a 3D object is rendered, the depth of each of its pixels is compared
against the value stored into the Z-buffer: if the pixel is closer it is
drawn, otherwise it is skipped.

No polygon sorting is needed. However, backface culling should be done
because it prevents many invisible polygons being compared against the
Z-buffer. Z-buffered rendering is the only algorithm supported by Allegro
that directly solves penetrating shapes (see example exzbuf.c, for instance).
The price to pay is more complex (and slower) routines.

Z-buffered polygons are designed as an extension of the normal POLYTYPE_*
rendering styles. Just OR the POLYTYPE with the value POLYTYPE_ZBUF, and
the normal polygon3d(), polygon3d_f(), quad3d(), etc. functions will
render z-buffered polygons.

Example:

   polygon3d(bmp, POLYTYPE_ATEX | POLYTYPE_ZBUF, tex, vc, vtx);


Of course, the z coordinates have to be valid regardless of rendering style.

A Z-buffered rendering procedure looks like a double-buffered rendering
procedure. You should follow four steps: create a Z-buffer at the beginning
of the program and make the library use it by calling set_zbuffer(). Then,
for each frame, clear the Z-buffer and draw polygons with
POLYTYPE_* | POLYTYPE_ZBUF and finally destroy the Z-buffer when leaving the
program.

Notes on Z-buffered renderers:

Unlike the normal POLYTYPE_FLAT renderers, the Z-buffered ones don't use
the hline() routine. Therefore DRAW_MODE has no effect.

The *LIT* routines work the traditional way - through the set of
blender routines.

All the Z-buffered routines are much slower than their normal counterparts
(they all use the FPU to interpolate and test 1/z values).

ZBUFFER *create_zbuffer(BITMAP *bmp);
   Creates a Z-buffer using the size of the BITMAP you are planning to draw
   on. Several Z-buffers can be defined but only one can be used at the same
   time, so you must call set_zbuffer() to make this Z-buffer active.

ZBUFFER *create_sub_zbuffer(ZBUFFER *parent, int x, int y, int width, int height);
   Creates a sub-z-buffer, ie. a z-buffer sharing drawing memory with a
   pre-existing z-buffer, but possibly with a different size. The same rules
   as for sub-bitmaps apply: the sub-z-buffer width and height can extend
   beyond the right and bottom edges of the parent (they will be clipped), 
   but the origin point must lie within the parent region.

   When drawing z-buffered to a bitmap, the top left corner of the bitmap is
   always mapped to the top left corner of the current z-buffer. So this
   function is primarily useful if you want to draw to a sub-bitmap and use
   the corresponding sub-area of the z-buffer. In other cases, eg. if you
   just want to draw to a sub-bitmap of screen (and not to other parts of
   screen), then you would usually want to create a normal z-buffer (not
   sub-z-buffer) the size of the visible screen. You don't need to first
   create a z-buffer the size of the virtual screen and then a sub-z-buffer
   of that.

void set_zbuffer(ZBUFFER *zbuf);
   Makes the given Z-buffer be the active one. This should have been
   previously created with create_zbuffer().

void clear_zbuffer(ZBUFFER *zbuf, float z);
   Writes z into the given Z-buffer (0 means far away). This function should
   be used to initialize the Z-buffer before each frame. Actually, low-level
   routines compare depth of the current pixel with 1/z: for example, if you
   want to clip polygons farther than 10, you must call
   clear_zbuffer(zbuf, 0.1);

void destroy_zbuffer(ZBUFFER *zbuf);
   Destroys the Z-buffer when you are finished with it.

Allegro provides two simple approaches to remove hidden surfaces:

   Z-buffering - (see above)

   Scan-line algorithms - along each scanline on your screen, you keep
   track of what polygons you are "in" and which is the nearest. This
   status changes only where the scanline crosses some polygon edge. So you
   have to juggle an edge list and a polygon list. And you have to sort the
   edges for each scanline (this can be countered by keeping the order of
   the previous scanline - it won't change much). The BIG advantage is that
   you write each pixel only once. If you have a lot of overlapping
   polygons you can get incredible speeds compared to any of the previous
   algorithms. This algorithm is covered by the *_scene routines.


The scene rendering has approximately the following steps:

   Initialize the scene (set the clip area, clear the bitmap, blit a
   background, etc.)

   Call clear_scene().

   Transform all your points to camera space.

   Clip polygons.

   Project with persp_project() or persp_project_f().

   "Draw" polygons with scene_polygon3d() and/or scene_polygon3d_f().
   This doesn't do any actual drawing, only initializes tables.

   Render all the polygons defined previously to the bitmap with
   render_scene().

   Overlay some non-3D graphics.

   Show the bitmap (blit it to screen, flip the page, etc).


For each horizontal line in the viewport an x-sorted edge list is used to
keep track of what polygons are "in" and which is the nearest. Vertical
coherency is used - the edge list for a scanline is sorted starting from
the previous one - it won't change much. The scene rendering routines use
the same low-level asm routines as normal polygon3d().

Notes on scene rendering:

   Unlike polygon3d(), scene_polygon3d() requires valid z coordinates
   for all vertices, regardless of rendering style (unlike
   polygon3d(), which only uses z coordinate for *PTEX*).

   All polygons passed to scene_polygon3d() have to be
   persp_project()'ed.

   After render_scene() the mode is reset to SOLID.

Using a lot of *MASK* polygons drastically reduces performance, because
when a MASKed polygon is the first in line of sight, the polygons
underneath have to be drawn too. The same applies to FLAT polygons drawn
with DRAW_MODE_TRANS.

Z-buffered rendering works also within the scene renderer. It may be
helpful when you have a few intersecting polygons, but most of the
polygons may be safely rendered by the normal scanline sorting algo.
Same as before: just OR the POLYTYPE with POLYTYPE_ZBUF. Also, you
have to clear the z-buffer at the start of the frame. Example:

   clear_scene(buffer);
   if (some_polys_are_zbuf) clear_zbuffer(0.);
   while (polygons) {
      ...
      if (this_poly_is_zbuf) type |= POLYTYPE_ZBUF;
      scene_polygon3d(type, tex, vc, vtx);
   }
   render_scene();


int create_scene(int nedge, int npoly);
   Allocates memory for a scene, nedge and npoly are your estimates of how
   many edges and how many polygons you will render (you cannot get over the
   limit specified here). If you use same values in succesive calls, the
   space will be reused (no new malloc()).

   The memory allocated is a little less than 150 * (nedge + npoly) bytes.
   Returns zero on success, or a negative number if allocations fail.

void clear_scene(BITMAP *bmp);
   Initializes a scene. The bitmap is the bitmap you will eventually render
   on.

void destroy_scene();
   Deallocate memory previously allocated by create_scene.

int scene_polygon3d(int type, BITMAP *texture, int vc, V3D *vtx[]);
int scene_polygon3d_f(int type, BITMAP *texture, int vc, V3D_f *vtx[]);
   Puts a polygon in the rendering list. Nothing is really rendered at this
   moment. Should be called between clear_scene() and render_scene().

   Arguments are the same as for polygon3d(), except the bitmap is missing.
   The one passed to clear_scene() will be used.

   Unlike polygon3d(), the polygon may be concave or self-intersecting.
   Shapes that penetrate one another may look OK, but they are not really
   handled by this code.

   Note that the texture is stored as a pointer only, and you should keep
   the actual bitmap around until render_scene(), where it is used.

   Since the FLAT style is implemented with the low-level hline() funtion,
   the FLAT style is subject to DRAW_MODEs. All these modes are valid. Along
   with the polygon, this mode will be stored for the rendering moment, and
   also all the other related variables (color_map pointer, pattern pointer,
   anchor, blender values).

   The settings of the CPU_MMX and CPU_3DNOW flags of the cpu_capabilities 
   global variable on entry in this routine affect the choice of low-level 
   asm routine that will be used by render_scene() for this polygon.

   Returns zero on success, or a negative number if it won't be rendered for
   lack of a rendering routine.

void render_scene();
   Renders all the specified scene_polygon3d()'s on the bitmap passed to
   clear_scene(). Rendering is done one scanline at a time, with no pixel
   being processed more than once.

   Note that between clear_scene() and render_scene() you shouldn't change
   the clip rectangle of the destination bitmap. For speed reasons, you
   should set the clip rectangle to the minimum.

   Note also that all the textures passed to scene_polygon3d() are stored as
   pointers only and actually used in render_scene().

extern float scene_gap;
   This number (default value = 100.0) controls the behaviour of the
   z-sorting algorithm. When an edge is very close to another's polygon
   plane, there is an interval of uncertainty in which you cannot tell which
   object is visible (which z is smaller). This is due to cumulative
   numerical errors for edges that have undergone a lot of transformations
   and interpolations.

   The default value means that if the 1/z values (in projected space)
   differ by only 1/100 (one percent), they are considered to be equal and
   the x-slopes of the planes are used to find out which plane is getting
   closer when we move to the right.

   Larger values means narrower margins, and increasing the chance of
   missing true adjacent edges/planes. Smaller values means larger margins,
   and increasing the chance of mistaking close polygons for adjacent ones.
   The value of 100 is close to the optimum. However, the optimum shifts
   slightly with resolution, and may be application-dependent. It is here 
   for you to fine-tune.



============================================================
============ Transparency and patterned drawing ============
============================================================

void drawing_mode(int mode, BITMAP *pattern, int x_anchor, int y_anchor);
   Sets the graphics drawing mode. This only affects the geometric routines 
   like putpixel, lines, rectangles, circles, polygons, floodfill, etc, not 
   the text output, blitting, or sprite drawing functions. The mode should 
   be one of the following constants:

      DRAW_MODE_SOLID               - the default, solid color
                                      drawing
      DRAW_MODE_XOR                 - exclusive-or drawing
      DRAW_MODE_COPY_PATTERN        - multicolored pattern fill
      DRAW_MODE_SOLID_PATTERN       - single color pattern fill
      DRAW_MODE_MASKED_PATTERN      - masked pattern fill
      DRAW_MODE_TRANS               - translucent color blending

   In DRAW_MODE_SOLID, pixels of the bitmap being drawn onto are simply 
   replaced by those produced by the drawing function.

   In DRAW_MODE_XOR, pixels are written to the bitmap with an exclusive-or 
   operation rather than a simple copy, so drawing the same shape twice will 
   erase it. Because it involves reading as well as writing the bitmap 
   memory, xor drawing is a lot slower than the normal replace mode.

   With the patterned modes, you provide a pattern bitmap which is tiled 
   across the surface of the shape. Allegro stores a pointer to this bitmap 
   rather than copying it, so you must not destroy the bitmap while it is 
   still selected as the pattern. The width and height of the pattern must 
   be powers of two, but they can be different, eg. a 64x16 pattern is fine, 
   but a 17x3 one is not. The pattern is tiled in a grid starting at point 
   (x_anchor, y_anchor). Normally you should just pass zero for these 
   values, which lets you draw several adjacent shapes and have the patterns 
   meet up exactly along the shared edges. Zero alignment may look peculiar 
   if you are moving a patterned shape around the screen, however, because 
   the shape will move but the pattern alignment will not, so in some 
   situations you may wish to alter the anchor position.

   When you select DRAW_MODE_COPY_PATTERN, pixels are simply copied from the 
   pattern bitmap onto the destination bitmap. This allows the use of 
   multicolored patterns, and means that the color you pass to the drawing 
   routine is ignored. This is the fastest of the patterned modes.

   In DRAW_MODE_SOLID_PATTERN, each pixel in the pattern bitmap is compared 
   with the mask color, which is zero in 256 color modes or bright pink for 
   truecolor data (maximum red and blue, zero green). If the pattern pixel 
   is solid, a pixel of the color you passed to the drawing routine is 
   written to the destination bitmap, otherwise a zero is written. The 
   pattern is thus treated as a monochrome bitmask, which lets you use the 
   same pattern to draw different shapes in different colors, but prevents 
   the use of multicolored patterns.

   DRAW_MODE_MASKED_PATTERN is almost the same as DRAW_MODE_SOLID_PATTERN, 
   but the masked pixels are skipped rather than being written as zeros, so 
   the background shows through the gaps.

   In DRAW_MODE_TRANS, the global color_map table or truecolor blender 
   functions are used to overlay pixels on top of the existing image. This 
   must only be used after you have set up the color mapping table (for 256 
   color modes) or blender functions (for truecolor modes). Because it 
   involves reading as well as writing the bitmap memory, translucent 
   drawing is very slow if you draw directly to video RAM, so wherever 
   possible you should use a memory bitmap instead.

void xor_mode(int on);
   This is a shortcut for toggling xor drawing mode on and off. Calling 
   xor_mode(TRUE) is equivalent to drawing_mode (DRAW_MODE_XOR, NULL, 0, 0);
   Calling xor_mode(FALSE) is equivalent to
   drawing_mode(DRAW_MODE_SOLID, NULL, 0, 0);

void solid_mode();
   This is a shortcut for selecting solid drawing mode. It is equivalent to 
   calling drawing_mode(DRAW_MODE_SOLID, NULL, 0, 0);

In paletted video modes, translucency and lighting are implemented with a 
64k lookup table, which contains the result of combining any two colors c1 
and c2. You must set up this table before you use any of the translucency 
or lighting routines. Depending on how you construct the table, a range of 
different effects are possible. For example, translucency can be implemented 
by using a color halfway between c1 and c2 as the result of the combination. 
Lighting is achieved by treating one of the colors as a light level (0-255) 
rather than a color, and setting up the table appropriately. A range of 
specialised effects are possible, for instance replacing any color with any 
other color and making individual source or destination colors completely 
solid or invisible.

Color mapping tables can be precalculated with the colormap utility, or 
generated at runtime. The COLOR_MAP structure is defined as:

typedef struct {
   unsigned char data[PAL_SIZE][PAL_SIZE];
} COLOR_MAP;

extern COLOR_MAP *color_map;
   Global pointer to the color mapping table. This must be set before using 
   any translucent or lit drawing functions in a 256 color video mode!

void create_trans_table(COLOR_MAP *table, const PALETTE pal,
                        int r, g, b, void (*callback)(int pos));
   Fills the specified color mapping table with lookup data for doing 
   translucency effects with the specified palette. When combining the 
   colors c1 and c2 with this table, the result will be a color somewhere 
   between the two. The r, g, and b parameters specify the solidity of each 
   color component, ranging from 0 (totally transparent) to 255 (totally 
   solid). For 50% solidity, pass 128. This function treats source color #0 
   as a special case, leaving the destination unchanged whenever a zero 
   source pixel is encountered, so that masked sprites will draw correctly. 
   If the callback function is not NULL, it will be called 256 times during 
   the calculation, allowing you to display a progress indicator.

void create_light_table(COLOR_MAP *table, const PALETTE pal,
                        int r, g, b, void (*callback)(int pos));
   Fills the specified color mapping table with lookup data for doing 
   lighting effects with the specified palette. When combining the colors c1 
   and c2 with this table, c1 is treated as a light level from 0-255. At 
   light level 255 the table will output color c2 unchanged, at light level 
   0 it will output the r, g, b value you specify to this function, and at 
   intermediate light levels it will output a color somewhere between the 
   two extremes. The r, g, and b values are in the range 0-63. If the 
   callback function is not NULL, it will be called 256 times during the 
   calculation, allowing you to display a progress indicator.

void create_color_table(COLOR_MAP *table, const PALETTE pal,
                        void (*blend)(PALETTE pal, int x, int y, RGB *rgb),
                        void (*callback)(int pos));
   Fills the specified color mapping table with lookup data for doing 
   customised effects with the specified palette, calling the blend function 
   to determine the results of each color combination. Your blend routine 
   will be passed a pointer to the palette and the two colors which are to 
   be combined, and should fill in the RGB structure with the desired result 
   in 0-63 format. Allegro will then search the palette for the closest 
   match to the RGB color that you requested, so it doesn't matter if the 
   palette has no exact match for this color. If the callback function is 
   not NULL, it will be called 256 times during the calculation, allowing 
   you to display a progress indicator.

void create_blender_table(COLOR_MAP *table, const PALETTE pal,
                          void (*callback)(int pos));
   Fills the specified color mapping table with lookup data for doing a 
   paletted equivalent of whatever truecolor blender mode is currently 
   selected. After calling set_trans_blender(), set_blender_mode(), or any 
   of the other truecolor blender mode routines, you can use this function 
   to create an 8 bit mapping table that will have the same results as 
   whatever 24 bit blending mode you have enabled.

In truecolor video modes, translucency and lighting are implemented by a 
blender function of the form:

unsigned long (*BLENDER_FUNC)(unsigned long x, y, n);


For each pixel to be drawn, this routine is passed two color parameters x 
and y, decomposes them into their red, green and blue components, combines 
them according to some mathematical transformation involving the 
interpolation factor n, and then merges the result back into a single 
return color value, which will be used to draw the pixel onto 
the destination bitmap.

The parameter x represents the blending modifier color and the parameter y 
represents the base color to be modified. The interpolation factor n is in 
the range [0-255] and controls the solidity of the blending.

When a translucent drawing function is used, x is the color of the source, 
y is the color of the bitmap begin drawn onto and n is the alpha level   
that was passed to the function that sets the blending mode (the RGB triplet 
that was passed to this function is not taken into account).

When a lit sprite drawing function is used, x is the color represented by 
the RGB triplet that was passed to the function that sets the blending mode 
(the alpha level that was passed to this function is not taken into 
account), y is the color of the sprite and n is the alpha level that was 
passed to the drawing function itself.

Since these routines may be used from various different color depths, there 
are three such callbacks, one for use with 15 bit 5.5.5 pixels, one for 16 
bit 5.6.5 pixels, and one for 24 bit 8.8.8 pixels (this can be shared 
between the 24 and 32 bit code since the bit packing is the same).

void set_trans_blender(int r, int g, int b, int a);
   Enables a linear interpolator blender mode for combining translucent 
   or lit truecolor pixels.

void set_alpha_blender();
   Enables the special alpha-channel blending mode, which is used for 
   drawing 32 bit RGBA sprites. After calling this function, you can use 
   draw_trans_sprite() or draw_trans_rle_sprite() to draw a 32 bit source 
   image onto any hicolor or truecolor destination. The alpha values will be 
   taken directly from the source graphic, so you can vary the solidity of 
   each part of the image. You can't use any of the normal translucency 
   functions while this mode is active, though, so you should reset to one 
   of the normal blender modes (eg. set_trans_blender()) before drawing 
   anything other than 32 bit RGBA sprites.

void set_write_alpha_blender();
   Enables the special alpha-channel editing mode, which is used for drawing 
   alpha channels over the top of an existing 32 bit RGB sprite, to turn it 
   into an RGBA format image. After calling this function, you can set the 
   drawing mode to DRAW_MODE_TRANS and then write draw color values (0-255) 
   onto a 32 bit image. This will leave the color values unchanged, but 
   alter the alpha to whatever values you are writing. After enabling this 
   mode you can also use draw_trans_sprite() to superimpose an 8 bit alpha 
   mask over the top of an existing 32 bit sprite.

void set_add_blender(int r, int g, int b, int a);
   Enables an additive blender mode for combining translucent or lit 
   truecolor pixels.

void set_burn_blender(int r, int g, int b, int a);
   Enables a burn blender mode for combining translucent or lit truecolor 
   pixels. Here the lightness values of the colours of the source image 
   reduce the lightness of the destination image, darkening the image.

void set_color_blender(int r, int g, int b, int a);
   Enables a color blender mode for combining translucent or lit truecolor 
   pixels. Applies only the hue and saturation of the source image to the 
   destination image. The luminance of the destination image is not affected.

void set_difference_blender(int r, int g, int b, int a);
   Enables a difference blender mode for combining translucent or lit 
   truecolor pixels. This makes an image which has colours calculated by the 
   difference between the source and destination colours.

void set_dissolve_blender(int r, int g, int b, int a);
   Enables a dissolve blender mode for combining translucent or lit 
   truecolor pixels. Randomly replaces the colours of some pixels in the 
   destination image with those of the source image. The number of pixels 
   replaced depends on the alpha value (higher value, more pixels replaced; 
   you get the idea :).

void set_dodge_blender(int r, int g, int b, int a);
   Enables a dodge blender mode for combining translucent or lit truecolor 
   pixels. The lightness of colours in the source lighten the colours of the 
   destination. White has the most effect; black has none.

void set_hue_blender(int r, int g, int b, int a);
   Enables a hue blender mode for combining translucent or lit truecolor 
   pixels. This applies the hue of the source to the destination.

void set_invert_blender(int r, int g, int b, int a);
   Enables an invert blender mode for combining translucent or lit truecolor 
   pixels. Blends the inverse (or negative) colour of the source with the 
   destination.

void set_luminance_blender(int r, int g, int b, int a);
   Enables a luminance blender mode for combining translucent or lit 
   truecolor pixels. Applies the luminance of the source to the destination.
   The colour of the destination is not affected.

void set_multiply_blender(int r, int g, int b, int a);
   Enables a multiply blender mode for combining translucent or lit 
   truecolor pixels. Combines the source and destination images, multiplying 
   the colours to produce a darker colour. If a colour is multiplied by 
   white it remains unchanged; when multiplied by black it also becomes 
   black.

void set_saturation_blender(int r, int g, int b, int a);
   Enables a saturation blender mode for combining translucent or lit 
   truecolor pixels. Applies the saturation of the source to the destination 
   image.

void set_screen_blender(int r, int g, int b, int a);
   Enables a screen blender mode for combining translucent or lit truecolor 
   pixels. This blender mode lightens the colour of the destination image by 
   multiplying the inverse of the source and destination colours. Sort of 
   like the opposite of the multiply blender mode.

void set_blender_mode(BLENDER_FUNC b15, b16, b24, int r, g, b, a);
   Specifies a custom set of truecolor blender routines, which can be used 
   to implement whatever special interpolation modes you need. This function 
   shares a single blender between the 24 and 32 bit modes.

void set_blender_mode_ex(BLENDER_FUNC b15, b16, b24, b32, b15x, b16x, b24x,
                         int r, g, b, a);
   Like set_blender_mode(), but allows you to specify a more complete set of 
   blender routines. The b15, b16, b24, and b32 routines are used when 
   drawing pixels onto destinations of the same format, while b15x, b16x, 
   and b24x are used by draw_trans_sprite() and draw_trans_rle_sprite() when 
   drawing RGBA images onto destination bitmaps of another format. These 
   blenders will be passed a 32 bit x parameter, along with a y value of a 
   different color depth, and must try to do something sensible in response.



==========================================================
============ Converting between color formats ============
==========================================================

In general, Allegro is designed to be used in only one color depth at a 
time, so you will call set_color_depth() once and then store all your 
bitmaps in the same format. If you want to mix several different pixel 
formats, you can use create_bitmap_ex() in place of create_bitmap(), and 
call bitmap_color_depth() to query the format of a specific image. Most of 
the graphics routines require all their input parameters to be in the same 
format (eg. you cannot stretch a 15 bit source bitmap onto a 24 bit 
destination), but there are some exceptions: blit() and the rotation routines 
can copy between bitmaps of any format, converting the data as required, 
draw_sprite() can draw 256 color source images onto destinations of any 
format, draw_character() _always_ uses a 256 color source bitmap, whatever 
the format of the destination, the draw_trans_sprite() and 
draw_trans_rle_sprite() functions are able to draw 32 bit RGBA images onto 
any hicolor or truecolor destination, as long as you call set_alpha_blender() 
first, and the draw_trans_sprite() function is able to draw an 8 bit alpha 
channel image over the top of an existing 32 bit image, as long as you call 
set_write_alpha_blender() first.

Expanding a 256 color source onto a truecolor destination is fairly fast 
(obviously you must set the correct palette before doing this conversion!). 
Converting between different truecolor formats is slightly slower, and 
reducing truecolor images to a 256 color destination is very slow (it can be 
sped up significantly if you set up the global rgb_map table before doing 
the conversion).

int bestfit_color(const PALETTE pal, int r, int g, int b);
   Searches the specified palette for the closest match to the requested 
   color, which are specified in the VGA hardware 0-63 format. Normally you 
   should call makecol8() instead, but this lower level function may be 
   useful if you need to use a palette other than the currently selected 
   one, or specifically don't want to use the rgb_map lookup table.

extern RGB_MAP *rgb_map;
   To speed up reducing RGB values to 8 bit paletted colors, Allegro uses a 
   32k lookup table (5 bits for each color component). You must set up this 
   table before using the gouraud shading routines, and if present the table 
   will also vastly accelerate the makecol8() function. RGB tables can be 
   precalculated with the rgbmap utility, or generated at runtime. The 
   RGB_MAP structure is defined as:

   typedef struct {
      unsigned char data[32][32][32];
   } RGB_MAP;

void create_rgb_table(RGB_MAP *table, const PALETTE pal,
                       void (*callback)(int pos));
   Fills the specified RGB mapping table with lookup data for the specified 
   palette. If the callback function is not NULL, it will be called 256 
   times during the calculation, allowing you to display a progress 
   indicator.

void hsv_to_rgb(float h, float s, float v, int *r, int *g, int *b);
void rgb_to_hsv(int r, int g, int b, float *h, float *s, float *v);
   Convert color values between the HSV and RGB colorspaces. The RGB values 
   range from 0 to 255, hue is from 0 to 360, and saturation and value are 
   from 0 to 1.



=======================================================
============ Direct access to video memory ============
=======================================================

The bitmap structure looks like:

typedef struct BITMAP
{
   int w, h;               - size of the bitmap in pixels
   int clip;               - non-zero if clipping is turned on
   int cl, cr, ct, cb;     - clip rectangle left, right, top, and bottom
   int seg;                - segment for use with the line pointers
   unsigned char *line[];  - pointers to the start of each line
} BITMAP;

There is some other stuff in the structure as well, but it is liable to 
change and you shouldn't use anything except the above. The clipping 
rectangle is inclusive on the left and top (0 allows drawing to position 0) 
but exclusive on the right and bottom (10 allows drawing to position 9, but 
not to 10). Note this is not the same format as you pass to set_clip(), 
which takes inclusive coordinates for all four corners.

There are several ways to get direct access to the image memory of a bitmap, 
varying in complexity depending on what sort of bitmap you are using.

The simplest approach will only work with memory bitmaps (obtained from 
create_bitmap(), grabber datafiles, and image files) and sub-bitmaps of 
memory bitmaps. This uses a table of char pointers, called 'line', which is 
a part of the bitmap structure and contains pointers to the start of each 
line of the image. For example, a simple memory bitmap putpixel function is:

   void memory_putpixel(BITMAP *bmp, int x, int y, int color)
   {
      bmp->line[y][x] = color;
   }

For truecolor modes you need to cast the line pointer to the appropriate 
type, for example:

   void memory_putpixel_15_or_16_bpp(BITMAP *bmp, int x, int y, int color)
   {
      ((short *)bmp->line[y])[x] = color;
   }

   void memory_putpixel_32(BITMAP *bmp, int x, int y, int color)
   {
      ((long *)bmp->line[y])[x] = color;
   }

If you want to write to the screen as well as to memory bitmaps, you need to 
use some helper macros, because the video memory may not be part of your 
normal address space. This simple routine will work for any linear screen, 
eg. a VESA linear framebuffers:

   void linear_screen_putpixel(BITMAP *bmp, int x, int y, int color)
   {
      bmp_select(bmp);
      bmp_write8((unsigned long)bmp->line[y]+x, color);
   }

For truecolor modes you should replace the bmp_write8() with bmp_write16(), 
bmp_write24(), or bmp_write32(), and multiply the x offset by the number of 
bytes per pixel. There are of course similar functions to read a pixel value
from a bitmap, namely bmp_read8(), bmp_read16(), bmp_read24() and
bmp_read32().

This still won't work in banked SVGA modes, however, or on platforms like 
Windows that do special processing inside the bank switching functions. For 
more flexible access to bitmap memory, you need to call the routines:

unsigned long bmp_write_line(BITMAP *bmp, int line);
   Selects the line of a bitmap that you are going to draw onto.

unsigned long bmp_read_line(BITMAP *bmp, int line);
   Selects the line of a bitmap that you are going to read from.

unsigned long bmp_unwrite_line(BITMAP *bmp);
   Releases the bitmap memory after you are finished with it. You only need 
   to call this once at the end of a drawing operation, even if you have 
   called bmp_write_line() or bmp_read_line() several times before it.

These are implemented as inline assembler routines, so they are not as 
inefficient as they might seem. If the bitmap doesn't require bank switching 
(ie. it is a memory bitmap, mode 13h screen, etc), these functions just 
return bmp->line[line].

Although SVGA bitmaps are banked, Allegro provides linear access to the 
memory within each scanline, so you only need to pass a y coordinate to 
these functions. Various x positions can be obtained by simply adding the x 
coordinate to the returned address. The return value is an unsigned long 
rather than a char pointer because the bitmap memory may not be in your data 
segment, and you need to access it with far pointers. For example, a 
putpixel using the bank switching functions is:

   void banked_putpixel(BITMAP *bmp, int x, int y, int color)
   {
      unsigned long address = bmp_write_line(bmp, y);
      bmp_select(bmp);
      bmp_write8(address+x, color);
      bmp_unwrite_line(bmp);
   }

You will notice that Allegro provides separate functions for setting the 
read and write banks. It is important that you distinguish between these, 
because on some graphics cards the banks can be set individually, and on 
others the video memory is read and written at different addresses. Life is 
never quite as simple as we might wish it to be, though (this is true even 
when we _aren't_ talking about graphics coding :-) and so of course some 
cards only provide a single bank. On these the read and write bank functions 
will behave identically, so you shouldn't assume that you can read from one 
part of video memory and write to another at the same time. You can call 
bmp_read_line(), and read whatever you like from that line, and then call 
bmp_write_line() with the same or a different line number, and write 
whatever you like to this second line, but you mustn't call bmp_read_line() 
and bmp_write_line() together and expect to be able to read one line and 
write the other simultaneously. It would be nice if this was possible, but 
if you do it, your code won't work on single banked SVGA cards.

And then there's mode-X. If you've never done any mode-X graphics coding, 
you probably won't understand this, but for those of you who want to know 
how Allegro sets up the mode-X screen bitmaps, here goes...

The line pointers are still present, and they contain planar addresses, ie. 
the actual location at which you access the first pixel in the line. These 
addresses are guaranteed to be quad aligned, so you can just set the write 
plane, divide your x coordinate by four, and add it to the line pointer. For 
example, a mode-X putpixel is:

   void modex_putpixel(BITMAP *b, int x, int y, int color)
   {
      outportw(0x3C4, (0x100<<(x&3))|2);
      bmp_select(bmp);
      bmp_write8((unsigned long)bmp->line[y]+(x>>2), color);
   }

Oh yeah: the djgpp nearptr hack. Personally I don't like this very much 
because it disables memory protection and isn't portable to other platforms, 
but a lot of people swear by it because it can give you direct access to the 
screen memory via a normal C pointer. Warning: this method will only work 
with the djgpp library, when using VGA 13h or a linear framebuffer modes!

In your setup code:

   #include <sys/nearptr.h>

   unsigned char *screenmemory;
   unsigned long screen_base_addr;

   __djgpp_nearptr_enable();

   __dpmi_get_segment_base_address(screen->seg, &screen_base_addr);

   screenmemory = (unsigned char *)(screen_base_addr + 
                                    screen->line[0] - 
                                    __djgpp_base_address);

Then:

   void nearptr_putpixel(int x, int y, int color)
   {
      screenmemory[x + y*VIRTUAL_W] = color;
   }



=======================================
============ FLIC routines ============
=======================================

There are two high level functions for playing FLI/FLC animations: 
play_fli(), which reads the data directly from disk, and play_memory_fli(), 
which uses data that has already been loaded into RAM. Apart from the 
different sources of the data, these two functions behave identically. They 
draw the animation onto the specified bitmap, which should normally be the 
screen. Frames will be aligned with the top left corner of the bitmap: if 
you want to position them somewhere else you will need to create a 
sub-bitmap for the FLI player to draw onto. If loop is set the player will 
cycle when it reaches the end of the file, otherwise it will play through 
the animation once and then return. If the callback function is not NULL it 
will be called once for each frame, allowing you to perform background tasks 
of your own. This callback should normally return zero: if it returns 
non-zero the player will terminate (this is the only way to stop an 
animation that is playing in looped mode). The FLI player returns FLI_OK if 
it reached the end of the file, FLI_ERROR if something went wrong, and the 
value returned by the callback function if that was what stopped it. If you 
need to distinguish between different return values, your callback should 
return positive integers, since FLI_OK is zero and FLI_ERROR is negative. 
Note that the FLI player will only work when the timer module is installed, 
and that it will alter the palette according to whatever palette data is 
present in the animation file.

Occasionally you may need more detailed control over how an FLI is played, 
for example if you want to superimpose a text scroller on top of the 
animation, or to play it back at a different speed. You could do both of 
these with the lower level functions described below.

int play_fli(const char *filename, BITMAP *bmp, int loop, int (*callback)());
   Plays an Autodesk Animator FLI or FLC animation file, reading the data 
   from disk as it is required.

int play_memory_fli(const void *fli_data, BITMAP *bmp, int loop,
                     int (*callback)());
   Plays an Autodesk Animator FLI or FLC animation, reading the data from a 
   copy of the file which is held in memory. You can obtain the fli_data 
   pointer by mallocing a block of memory and reading an FLI file into it, 
   or by importing an FLI into a grabber datafile. Playing animations from 
   memory is obviously faster than cueing them directly from disk, and is 
   particularly useful with short, looped FLI's. Animations can easily get 
   very large, though, so in most cases you will probably be better just 
   using play_fli().

int open_fli(const char *filename);
int open_memory_fli(const void *fli_data);
   Open FLI files ready for playing, reading the data from disk or memory 
   respectively. Return FLI_OK on success. Information about the current FLI 
   is held in global variables, so you can only have one animation open at a 
   time.

void close_fli();
   Closes an FLI file when you have finished reading from it.

int next_fli_frame(int loop);
   Reads the next frame of the current animation file. If loop is set the 
   player will cycle when it reaches the end of the file, otherwise it will 
   return FLI_EOF. Returns FLI_OK on success, FLI_ERROR or FLI_NOT_OPEN on 
   error, and FLI_EOF on reaching the end of the file. The frame is read 
   into the global variables fli_bitmap and fli_palette.

extern BITMAP *fli_bitmap;
   Contains the current frame of the FLI/FLC animation.

extern PALETTE fli_palette;
   Contains the current FLI palette.

extern int fli_bmp_dirty_from;
extern int fli_bmp_dirty_to;
   These variables are set by next_fli_frame() to indicate which part of the 
   fli_bitmap has changed since the last call to reset_fli_variables(). If 
   fli_bmp_dirty_from is greater than fli_bmp_dirty_to, the bitmap has not 
   changed, otherwise lines fli_bmp_dirty_from to fli_bmp_dirty_to 
   (inclusive) have altered. You can use these when copying the fli_bitmap 
   onto the screen, to avoid moving data unnecessarily.

extern int fli_pal_dirty_from;
extern int fli_pal_dirty_to;
   These variables are set by next_fli_frame() to indicate which part of the 
   fli_palette has changed since the last call to reset_fli_variables(). If 
   fli_pal_dirty_from is greater than fli_pal_dirty_to, the palette has not 
   changed, otherwise colors fli_pal_dirty_from to fli_pal_dirty_to 
   (inclusive) have altered. You can use these when updating the hardware 
   palette, to avoid unnecessary calls to set_palette().

void reset_fli_variables();
   Once you have done whatever you are going to do with the fli_bitmap and 
   fli_palette, call this function to reset the fli_bmp_dirty_* and 
   fli_pal_dirty_* variables.

extern int fli_frame;
   Global variable containing the current frame number in the FLI file. This 
   is useful for synchronising other events with the animation, for instance 
   you could check it in a play_fli() callback function and use it to 
   trigger a sample at a particular point.

extern volatile int fli_timer;
   Global variable for timing FLI playback. When you open an FLI file, a 
   timer interrupt is installed which increments this variable every time a 
   new frame should be displayed. Calling next_fli_frame() decrements it, so 
   you can test it and know that it is time to display a new frame if it is 
   greater than zero.



=============================================
============ Sound init routines ============
=============================================

int detect_digi_driver(int driver_id);
   Detects whether the specified digital sound device is available. Returns 
   the maximum number of voices that the driver can provide, or zero if the 
   hardware is not present. This function must be called _before_ 
   install_sound().

int detect_midi_driver(int driver_id);
   Detects whether the specified MIDI sound device is available. Returns the 
   maximum number of voices that the driver can provide, or zero if the 
   hardware is not present. There are two special-case return values that 
   you should watch out for: if this function returns -1 it is a 
   note-stealing driver (eg. DIGMID) that shares voices with the current 
   digital sound driver, and if it returns 0xFFFF it is an external device 
   like an MPU-401 where there is no way to determine how many voices are 
   available. This function must be called _before_ install_sound(). 

void reserve_voices(int digi_voices, int midi_voices);
   Call this function to specify the number of voices that are to be used by 
   the digital and MIDI sound drivers respectively. This must be done 
   _before_ calling install_sound(). If you reserve too many voices, 
   subsequent calls to install_sound() will fail. How many voices are 
   available depends on the driver, and in some cases you will actually get 
   more than you reserve (eg. the FM synth drivers will always provide 9 
   voices on an OPL2 and 18 on an OPL3, and the SB digital driver will round 
   the number of voices up to the nearest power of two). Pass negative 
   values to restore the default settings. You should be aware that the 
   sound quality is usually inversely related to how many voices you use, so 
   don't reserve any more than you really need.

void set_volume_per_voice(int scale);
   By default, when you reserve more voices for the digital sound driver, 
   Allegro will reduce the volume of each voice to compensate. This is done 
   to avoid too much distortion. The default volume per voice is such that, 
   if you reserve n voices, you can play up to n/2 normalised samples with 
   centre panning without risking distortion. The exception is when you have 
   fewer than 8 voices, where the volume remains the same as for 8 voices.

   If the resultant output is either too loud or too quiet, this function 
   can be used to adjust the volume of each voice. You should first check 
   that your speakers are at a reasonable volume, Allegro's global volume 
   is at maximum (see set_volume() below), and any other mixers such as 
   the Windows Volume Control are set reasonably.

   Once you are sure that Allegro's output level is unsuitable for your 
   application, use this function to adjust it. This must be done _before_ 
   calling install_sound(). Note that this function is currently only 
   relevant for drivers that use the Allegro mixer (which is most of them).

   If you pass 0 to this function, each centred sample will play at the 
   maximum volume possible without distortion, as will all samples played 
   through a mono driver. Samples at the extreme left and right will distort 
   if played at full volume. If you wish to play panned samples at full 
   volume without distortion, you should pass 1 to this function.
   Note: this is different from the function's behaviour in WIPs 3.9.34, 
   3.9.35 and 3.9.36. If you used this function under one of these WIPs, 
   you will have to increase your parameter by one to get the same volume.

   Each time you increase the parameter by one, the volume of each voice 
   will halve. For example, if you pass 4, you can play up to 16 centred 
   samples at maximum volume without distortion.

   Here are the default values, dependent on the number of voices:

   1-8 voices - set_volume_per_voice(2)
    16 voices - set_volume_per_voice(3)
    32 voices - set_volume_per_voice(4)
    64 voices - set_volume_per_voice(5)

   Of course this function does not override the volume you specify with 
   play_sample() or voice_set_volume(). It simply alters the overall output 
   of the program. If you play samples at lower volumes, or if they are not 
   normalised, then you can play more of them without distortion.

   Warning: Allegro uses a clipping table to clip the waveform. The table is 
   big enough to accommodate a total output of up to 4 times the maximum 
   possible without distortion. If your output goes above this limit, the 
   wave will 'wrap around' (peaks become troughs and vice versa), thus 
   distorting much more. You should be careful that this does not happen.

   It is recommended that you hard-code the parameter into your program, 
   rather than offering it to the user. The user can alter the volume with 
   the configuration file instead, or you can provide for this with 
   set_volume().

   To restore volume per voice to its default behaviour, pass -1.

int install_sound(int digi, int midi, const char *cfg_path);
   Initialises the sound module. You should normally pass DIGI_AUTODETECT 
   and MIDI_AUTODETECT as the driver parameters to this function, in which 
   case Allegro will read hardware settings from the current configuration 
   file. This allows the user to select different values with the setup 
   utility: see the config section for details. Alternatively, see the 
   platform specific documentation for a list of the available drivers. The 
   cfg_path parameter is only present for compatibility with previous 
   versions of Allegro, and has no effect on anything. Returns zero if the 
   sound is successfully installed, and -1 on failure. If it fails it will 
   store a description of the problem in allegro_error.

void remove_sound();
   Cleans up after you are finished with the sound routines. You don't 
   normally need to call this, because allegro_exit() will do it for you.

void set_volume(int digi_volume, int midi_volume);
   Alters the global sound output volume. Specify volumes for both digital 
   samples and MIDI playback, as integers from 0 to 255, or pass a negative 
   value to leave one of the settings unchanged. If possible this routine 
   will use a hardware mixer to control the volume, otherwise it will tell 
   the sample and MIDI players to simulate a mixer in software.



=================================================
============ Digital sample routines ============
=================================================

SAMPLE *load_sample(const char *filename);
   Loads a sample from a file, returning a pointer to it, or NULL on error. 
   At present this function supports both mono and stereo WAV and mono VOC 
   files, in 8 or 16 bit formats.

SAMPLE *load_wav(const char *filename);
   Loads a sample from a RIFF WAV file.

SAMPLE *load_voc(const char *filename);
   Loads a sample from a Creative Labs VOC file.

SAMPLE *create_sample(int bits, int stereo, int freq, int len);
   Constructs a new sample structure of the specified type. The data field 
   points to a block of waveform data: see the structure definition in 
   allegro/digi.h for details.

void destroy_sample(SAMPLE *spl);
   Destroys a sample structure when you are done with it. It is safe to call 
   this even when the sample might be playing, because it checks and will 
   kill it off if it is active.

void lock_sample(SAMPLE *spl);
   Under DOS, locks all the memory used by a sample. You don't normally need 
   to call this function because load_sample() and create_sample() do it for
   you.

int play_sample(const SAMPLE *spl, int vol, int pan, int freq, int loop);
   Triggers a sample at the specified volume, pan position, and frequency. 
   The volume and pan range from 0 (min/left) to 255 (max/right). Frequency 
   is relative rather than absolute: 1000 represents the frequency that the 
   sample was recorded at, 2000 is twice this, etc. If the loop flag is set, 
   the sample will repeat until you call stop_sample(), and can be 
   manipulated while it is playing by calling adjust_sample(). Returns the
   voice number that was allocated for the sample (a non-negative number if
   successful).

void adjust_sample(const SAMPLE *spl, int vol, int pan, int freq, int loop);
   Alters the parameters of a sample while it is playing (useful for 
   manipulating looped sounds). You can alter the volume, pan, and 
   frequency, and can also clear the loop flag, which will stop the sample 
   when it next reaches the end of its loop. If there are several copies of 
   the same sample playing, this will adjust the first one it comes across. 
   If the sample is not playing it has no effect.

void stop_sample(const SAMPLE *spl);
   Kills off a sample, which is required if you have set a sample going in 
   looped mode. If there are several copies of the sample playing, it will 
   stop them all.

If you need more detailed control over how samples are played, you can use 
the lower level voice functions rather than just calling play_sample(). This 
is rather more work, because you have to explicitly allocate and free the 
voices rather than them being automatically released when they finish 
playing, but allows far more precise specification of exactly how you want 
everything to sound. You may also want to modify a couple of fields from the 
sample structure:

   int priority;
      Ranging 0-255 (default 128), this controls how voices are
      allocated if you attempt to play more than the driver can handle.
      This may be used to ensure that the less important sounds are
      cut off while the important ones are preserved.

   unsigned long loop_start;
   unsigned long loop_end;
      Loop position in sample units, by default set to the start and end of 
      the sample.

int allocate_voice(const SAMPLE *spl);
   Allocates a soundcard voice and prepares it for playing the specified 
   sample, setting up sensible default parameters (maximum volume, centre 
   pan, no change of pitch, no looping). When you are finished with the 
   voice you must free it by calling deallocate_voice() or release_voice(). 
   Returns the voice number, or -1 if no voices are available.

void deallocate_voice(int voice);
   Frees a soundcard voice, stopping it from playing and releasing whatever 
   resources it is using.

void reallocate_voice(int voice, const SAMPLE *spl);
   Switches an already-allocated voice to use a different sample. Calling 
   reallocate_voice(voice, sample) is equivalent to:

      deallocate_voice(voice);
      voice = allocate_voice(sample);

void release_voice(int voice);
   Releases a soundcard voice, indicating that you are no longer interested 
   in manipulating it. The sound will continue to play, and any resources 
   that it is using will automatically be freed when it finishes. This is 
   essentially the same as deallocate_voice(), but it waits for the sound to 
   stop playing before taking effect.

void voice_start(int voice);
   Activates a voice, using whatever parameters have been set for it.

void voice_stop(int voice);
   Stops a voice, storing the current position and state so that it may 
   later be resumed by calling voice_start().

void voice_set_priority(int voice, int priority);
   Sets the priority of a voice (range 0-255). This is used to decide which 
   voices should be chopped off, if you attempt to play more than the 
   soundcard driver can handle.

SAMPLE *voice_check(int voice);
   Checks whether a voice is currently allocated. It returns a copy of the 
   sample that the voice is using, or NULL if the voice is inactive (ie. it 
   has been deallocated, or the release_voice() function has been called and 
   the sample has then finished playing).

int voice_get_position(int voice);
   Returns the current position of a voice, in sample units, or -1 if it has 
   finished playing.

void voice_set_position(int voice, int position);
   Sets the position of a voice, in sample units.

void voice_set_playmode(int voice, int playmode);
   Adjusts the loop status of the specified voice. This can be done while 
   the voice is playing, so you can start a sample in looped mode (having 
   set the loop start and end positions to the appropriate values), and then 
   clear the loop flag when you want to end the sound, which will cause it 
   to continue past the loop end, play the subsequent part of the sample, 
   and finish in the normal way. The mode parameter is a bitfield containing 
   the following values:
   
   PLAYMODE_PLAY
      Plays the sample a single time. This is the default if you don't set 
      the loop flag.
   
   PLAYMODE_LOOP
      Loops repeatedly through the sample, jumping back to the loop start 
      position upon reaching the loop end.
   
   PLAYMODE_FORWARD
      Plays the sample from beginning to end. This is the default if you 
      don't set the backward flag.
   
   PLAYMODE_BACKWARD
      Reverses the direction of the sample. If you combine this with the 
      loop flag, the sample jumps to the loop end position upon reaching the 
      loop start (ie. you do not need to reverse the loop start and end 
      values when you play the sample in reverse).
   
   PLAYMODE_BIDIR
      When used in combination with the loop flag, causes the sample to 
      change direction each time it reaches one of the loop points, so it 
      alternates between playing forwards and in reverse.
   
int voice_get_volume(int voice);
   Returns the current volume of the voice, range 0-255.

void voice_set_volume(int voice, int volume);
   Sets the volume of the voice, range 0-255.

void voice_ramp_volume(int voice, int time, int endvol);
   Starts a volume ramp (crescendo or diminuendo) from the current volume to 
   the specified ending volume, lasting for time milliseconds.

void voice_stop_volumeramp(int voice);
   Interrupts a volume ramp operation.

int voice_get_frequency(int voice);
   Returns the current pitch of the voice, in Hz.

void voice_set_frequency(int voice, int frequency);
   Sets the pitch of the voice, in Hz.

void voice_sweep_frequency(int voice, int time, int endfreq);
   Starts a frequency sweep (glissando) from the current pitch to the 
   specified ending pitch, lasting for time milliseconds.

void voice_stop_frequency_sweep(int voice);
   Interrupts a frequency sweep operation.

int voice_get_pan(int voice);
   Returns the current pan position, from 0 (left) to 255 (right).

void voice_set_pan(int voice, int pan);
   Sets the pan position, ranging from 0 (left) to 255 (right).

void voice_sweep_pan(int voice, int time, int endpan);
   Starts a pan sweep (left  right movement) from the current position to 
   the specified ending position, lasting for time milliseconds.

void voice_stop_pan_sweep(int voice);
   Interrupts a pan sweep operation.

void voice_set_echo(int voice, int strength, int delay);
   Sets the echo parameters for a voice (not currently implemented).

void voice_set_tremolo(int voice, int rate, int depth);
   Sets the tremolo parameters for a voice (not currently implemented).

void voice_set_vibrato(int voice, int rate, int depth);
   Sets the vibrato parameters for a voice (not currently implemented).



=============================================
============ MIDI music routines ============
=============================================

MIDI *load_midi(const char *filename);
   Loads a MIDI file (handles both format 0 and format 1), returning a 
   pointer to a MIDI structure, or NULL on error.

void destroy_midi(MIDI *midi);
   Destroys a MIDI structure when you are done with it. It is safe to call 
   this even when the MIDI file might be playing, because it checks and will 
   kill it off if it is active.

void lock_midi(MIDI *midi);
   Under DOS, locks all the memory used by a MIDI file. You don't normally 
   need to call this function because load_midi() does it for you.

int play_midi(MIDI *midi, int loop);
   Starts playing the specified MIDI file, first stopping whatever music was 
   previously playing. If the loop flag is set, the data will be repeated 
   until replaced with something else, otherwise it will stop at the end of 
   the file. Passing a NULL pointer will stop whatever music is currently 
   playing. Returns non-zero if an error occurs (this may happen if a 
   patch-caching wavetable driver is unable to load the required samples, or 
   at least it might in the future when somebody writes some patch-caching 
   wavetable drivers :-)

int play_looped_midi(MIDI *midi, int loop_start, int loop_end);
   Starts playing a MIDI file with a user-defined loop position. When the 
   player reaches the loop end position or the end of the file (loop_end may 
   be -1 to only loop at EOF), it will wind back to the loop start point. 
   Both positions are specified in the same beat number format as the 
   midi_pos variable.

void stop_midi();
   Stops whatever music is currently playing. This is the same thing as 
   calling play_midi(NULL, FALSE).

void midi_pause();
   Pauses the MIDI player.

void midi_resume();
   Resumes playback of a paused MIDI file.

int midi_seek(int target);
   Seeks to the given midi_pos in the current MIDI file. If the target is 
   earlier in the file than the current midi_pos it seeks from the 
   beginning; otherwise it seeks from the current position. Returns zero if 
   it could successfully seek to the requested position. Otherwise, a
   return value of 1 means it stopped playing, and midi_pos is set to the
   negative length of the MIDI file (so you can use this function to
   determine the length of a MIDI file). A return value of 2 means the MIDI
   file looped back to the start.

void midi_out(unsigned char *data, int length);
   Streams a block of MIDI commands into the player in realtime, allowing 
   you to trigger notes, jingles, etc, over the top of whatever MIDI file is 
   currently playing.

int load_midi_patches();
   Forces the MIDI driver to load the entire set of patches ready for use. 
   You will not normally need to call this, because Allegro automatically 
   loads whatever data is required for the current MIDI file, but you must 
   call it before sending any program change messages via the midi_out() 
   command. Returns non-zero if an error occurred.

extern volatile long midi_pos;
   Stores the current position (beat number) in the MIDI file, or contains 
   a negative number if no music is currently playing. Useful for 
   synchronising animations with the music, and for checking whether a MIDI 
   file has finished playing.

extern long midi_loop_start;
extern long midi_loop_end;
   The loop start and end points, set by the play_looped_midi() function. 
   These may safely be altered while the music is playing, but you should be 
   sure they are always set to sensible values (start < end). If you are 
   changing them both at the same time, make sure to alter them in the right 
   order in case a MIDI interrupt happens to occur in between your two 
   writes! Setting these values to -1 represents the start and end of the 
   file respectively.

extern void (*midi_msg_callback)(int msg, int byte1, int byte2);
extern void (*midi_meta_callback)(int type, const unsigned char *data, int length);
extern void (*midi_sysex_callback)(const unsigned char *data, int length);
   Hook functions allowing you to intercept MIDI player events. If set to 
   anything other than NULL, these routines will be called for each MIDI 
   message, meta-event, and system exclusive data block respectively. They 
   will execute in an interrupt handler context, so all the code and data 
   they use should be locked, and they must not call any operating system 
   functions. In general you just use these routines to set some flags and 
   respond to them later in your mainline code.

int load_ibk(char *filename, int drums);
   Reads in a .IBK patch definition file for use by the Adlib driver. If 
   drums is set, it will load it as a percussion patch set, otherwise it 
   will use it as a replacement set of General MIDI instruments. You may 
   call this before or after initialising the sound code, or can simply set 
   the ibk_file and ibk_drum_file variables in the configuration file to 
   have the data loaded automatically. Note that this function has no effect 
   on any drivers other than the Adlib one! Returns non-zero on error.



===============================================
============ Audio stream routines ============
===============================================

The audio stream functions are for playing digital sounds that are too big 
to fit in a regular SAMPLE structure, either because they are huge files 
that you want to load in pieces as the data is required, or because you are 
doing something clever like generating the waveform on the fly.

AUDIOSTREAM *play_audio_stream(int len, bits, stereo, freq, vol, pan);
   This function creates a new audio stream and starts it playing. The 
   length is the size of each transfer buffer (in samples), which should 
   normally (but doesn't have to) be a power of two somewhere around 1k in
   size. Larger buffers are more efficient and require fewer updates, but
   result in more latency between you providing the data and it actually
   being played. The bits parameter must be 8 or 16, freq is the sample
   rate of the data in Hertz. The vol and pan values use the same 0-255
   ranges as the regular sample playing functions. The stereo parameter
   should be set to 1 for stereo streams, or 0 otherwise. If you want to
   adjust the pitch, volume, or panning of a stream once it is playing,
   you can use the regular voice_*() functions with stream->voice as a
   parameter. The sample data is always in unsigned format, with stereo
   waveforms consisting of alternating left/right samples, left sample
   first.

void stop_audio_stream(AUDIOSTREAM *stream);
   Destroys an audio stream when it is no longer required.

void *get_audio_stream_buffer(AUDIOSTREAM *stream);
   You must call this function at regular intervals while an audio stream is 
   playing, to provide the next buffer of sample data (the smaller the 
   stream buffer size, the more often it must be called). If it returns 
   NULL, the stream is still playing the previous lot of data, so you don't 
   need to do anything. If it returns a value, that is the location of the 
   next buffer to be played, and you should load the appropriate number of 
   samples (however many you specified when creating the stream) to that 
   address, for example using an fread() from a disk file. After filling the 
   buffer with data, call free_audio_stream_buffer() to indicate that the 
   new data is now valid. Note that this function should not be called from 
   a timer handler...

void free_audio_stream_buffer(AUDIOSTREAM *stream);
   Call this function after get_audio_stream_buffer() returns a non-NULL 
   address, to indicate that you have loaded a new block of samples to that 
   location and the data is now ready to be played.



============================================
============ Recording routines ============
============================================

int install_sound_input(int digi, int midi);
   Initialises the sound recorder module, returning zero on success. You 
   must install the normal sound playback system before calling this 
   routine. The two card parameters should use the same constants as 
   install_sound(), including DIGI_NONE and MIDI_NONE to disable parts of 
   the module, or DIGI_AUTODETECT and MIDI_AUTODETECT to guess the hardware.

void remove_sound_input();
   Cleans up after you are finished with the sound input routines. You don't 
   normally need to call this, because remove_sound() and/or allegro_exit() 
   will do it for you.

int get_sound_input_cap_bits();
   Checks which sample formats are supported by the current audio input 
   driver, returning one of the bitfield values:

      0 = audio input not supported
      8 = eight bit audio input is supported
      16 = sixteen bit audio input is supported
      24 = both eight and sixteen bit audio input are supported

int get_sound_input_cap_stereo();
   Checks whether the current audio input driver is capable of stereo 
   recording.

int get_sound_input_cap_rate(int bits, int stereo);
   Returns the maximum possible sample frequency for recording in the 
   specified format, or zero if these settings are not supported.

int get_sound_input_cap_parm(int rate, int bits, int stereo);
   Checks whether the specified recording frequency, number of bits, and 
   mono/stereo mode are supported by the current audio driver, returning one 
   of the values:

      0 = it is impossible to record in this format
      1 = recording is possible, but audio output will be suspended
      2 = recording is possible at the same time as playing other sounds
      -n = sampling rate not supported, but rate 'n' would work instead

int set_sound_input_source(int source);
   Selects the audio input source, returning zero on success or -1 if the 
   hardware does not provide an input select register. The parameter should 
   be one of the values:

      SOUND_INPUT_MIC
      SOUND_INPUT_LINE
      SOUND_INPUT_CD

int start_sound_input(int rate, int bits, int stereo);
   Starts recording in the specified format, suspending audio playback as 
   necessary (this will always happen with the current drivers). Returns the 
   buffer size in bytes if successful, or zero on error.

void stop_sound_input();
   Stops audio recording, switching the card back into the normal playback 
   mode.

int read_sound_input(void *buffer);
   Retrieves the most recently recorded audio buffer into the specified 
   location, returning non-zero if a buffer has been copied or zero if no 
   new data is yet available. The buffer size can be obtained by checking 
   the return value from start_sound_input(). You must be sure to call this 
   function at regular intervals during the recording (typically around 100 
   times a second), or some data will be lost. If you are unable to do this 
   often enough from the mainline code, use the digi_recorder() callback to 
   store the waveform into a larger buffer of your own. Note: many cards 
   produce a click or popping sound when switching between record and 
   playback modes, so it is often a good idea to discard the first buffer 
   after you start a recording. The waveform is always stored in unsigned 
   format, with stereo data consisting of alternate left/right samples.

extern void (*digi_recorder)();
   If set, this function is called by the input driver whenever a new sample 
   buffer becomes available, at which point you can use read_sound_input() 
   to copy the data into a more permenent location. This routine runs in an 
   interrupt context, so it must execute very quickly, the code and all 
   memory that it touches must be locked, and you cannot call any operating 
   system routines or access disk files.

extern void (*midi_recorder)(unsigned char data);
   If set, this function is called by the MIDI input driver whenever a new 
   byte of MIDI data becomes available. It runs in an interrupt context, so 
   it must execute very quickly and all the code/data must be locked.



=======================================================
============ File and compression routines ============
=======================================================

The following routines implement a fast buffered file I/O system, which 
supports the reading and writing of compressed files using a ring buffer 
algorithm based on the LZSS compressor by Haruhiko Okumura. This does not 
achieve quite such good compression as programs like zip and lha, but 
unpacking is very fast and it does not require much memory. Packed files 
always begin with the 32 bit value F_PACK_MAGIC, and autodetect files with 
the value F_NOPACK_MAGIC.

The following FA_* flags are guaranteed to work: FA_RDONLY, FA_HIDDEN,
FA_SYSTEM, FA_LABEL, FA_DIREC, FA_ARCH. Do not use any other flags from 
DOS/Windows or your code will not compile on another platform. Flags
FA_SYSTEM, FA_LABEL and FA_ARCH are valuable only on DOS/Windows (entries
with system flag, volume labels and archive flag). FA_RDONLY is for 
directory entries with read-only flag on DOS-like systems or unwritable 
by current user on Unix-like systems. FA_HIDDEN is for entries with hidden 
flag on DOS-like systems or starting with '.' on Unix (dotted files - 
excluding '.' and '..'). FA_DIREC represents directories. Flags can be 
combined using '|' (binary OR operator).

When passed to the functions as the 'attrib' parameter, these flags 
represent an upper set in which the actual flag set of a matching file must 
be included. That is, in order for a file to be matching, its attributes 
may contain any of the specified flags but must not contain any of the 
unspecified flags. Thus, if you pass 'FA_DIREC | FA_RDONLY', normal files 
and directories will be included as well as read-only files and 
directories, but not hidden files and directories. Similarly, if you pass 
'FA_ARCH' then both archived and non-archived files will be included.

void get_executable_name(char *buf, int size);
   Fills buf with the full path to the current executable, writing at most 
   size bytes.  This generally comes from argv[0], but on Unix systems if
   argv[0] does not specify the path, we search for our file in $PATH.

char *fix_filename_case(char *path);
   Converts a filename to a standardised case. On DOS platforms, they will 
   be entirely uppercase. Returns a copy of the path parameter.

char *fix_filename_slashes(char *path);
   Converts all the directory separators in a filename to a standard 
   character. On DOS platforms, this is a backslash. Returns a copy of the 
   path parameter.

char *fix_filename_path(char *dest, const char *path, int size);
   Converts a partial filename into a full path, storing at most size bytes 
   into the dest buffer. Returns a copy of the dest parameter.

char *replace_filename(char *dest, const char *path, 
                        const char *filename, int size);
   Replaces the specified path+filename with a new filename tail, storing 
   at most size bytes into the dest buffer. Returns a copy of the dest 
   parameter.

char *replace_extension(char *dest, const char *filename, 
                         const char *ext, int size);
   Replaces the specified filename+extension with a new extension tail, 
   storing at most size bytes into the dest buffer. Returns a copy of the 
   dest parameter.

char *append_filename(char *dest, const char *path, 
                       const char *filename, int size);
   Concatenates the specified filename onto the end of the specified path, 
   storing at most size bytes into the dest buffer. Returns a copy of the 
   dest parameter.

char *get_filename(const char *path);
   When passed a completely specified file path, this returns a pointer to 
   the filename portion. Both '\' and '/' are recognized as directory 
   separators.

char *get_extension(const char *filename);
   When passed a complete filename (with or without path information) this 
   returns a pointer to the file extension.

void put_backslash(char *filename);
   If the last character of the filename is not a '\', '/', '#' or a device
   separator (ie. ':' under DOS), this routine will concatenate either a '\'
   or '/' on to it (depending on the platform). Note: ignore the function
   name, it's out of date.

int file_exists(const char *filename, int attrib, int *aret);
   Checks whether a file matching the given name and attributes (see above) 
   exists, returning non-zero if it does. If aret is not NULL, it will be 
   set to the attributes of the matching file. If an error occurs the system 
   error code will be stored in errno.

int exists(const char *filename);
   Shortcut version of file_exists(), which checks for normal files, which 
   may have the archive or read-only bits set, but are not hidden, 
   directories, system files, etc.

long file_size(const char *filename);
   Returns the size of a file, in bytes. If the file does not exist or an 
   error occurs, it will return zero and store the system error code in 
   errno.

time_t file_time(const char *filename);
   Returns the modification time (number of seconds since 00:00:00 GMT 
   1/1/1970) of a file. If the file does not exist or an error occurs, it 
   will return zero and store the system error code in errno.

int delete_file(const char *filename);
   Removes a file from the disk.

int for_each_file(const char *name, int attrib,
                  void (*callback)(const char *filename, int attrib, int param),
                  int param);
   Finds all the files on the disk which match the given wildcard 
   specification and file attributes (see above), and executes callback() 
   once for each. callback() will be passed three arguments, the first a 
   string which contains the completed filename, the second being the 
   attributes of the file, and the third an int which is simply a copy of 
   param (you can use this for whatever you like). If an error occurs an 
   error code will be stored in errno, and callback() can cause 
   for_each_file() to abort by setting errno itself. Returns the number of 
   successful calls made to callback().

int al_findfirst(const char *pattern, struct al_ffblk *info, int attrib);
   Low-level function for searching files. This function finds the first 
   file which matches the given wildcard specification and file attributes 
   (see above). The information about the file (if any) will be put in the 
   al_ffblk structure which you have to provide. The function returns zero 
   if a match is found, nonzero if none is found or if an error occured  
   and, in the latter case, sets errno accordingly. The al_ffblk structure 
   looks like:

  struct al_ffblk
  {
      int attrib;       - actual attributes of the file found
      time_t time;      - modification time of file
      long size;        - size of file
      char name[512];   - name of file
  };

   There is some other stuff in the structure as well, but it is there for 
   internal use only.

int al_findnext(struct al_ffblk *info);
   This finds the next file in a search started by al_findfirst(). Returns 
   zero if a match is found, nonzero if none is found or if an error 
   occured and, in the latter case, sets errno accordingly.

void al_findclose(struct al_ffblk *info);
   This closes a previously opened search with al_findfirst().

int find_allegro_resource(char *dest, const char *resource, 
                          const char *ext, const char *datafile, 
                          const char *objectname, const char *envvar, 
                          const char *subdir, int size);
   Searches for a support file, eg. allegro.cfg or language.dat. Passed a 
   resource string describing what you are looking for, along with extra 
   optional information such as the default extension, what datafile to look 
   inside, what the datafile object name is likely to be, any special 
   environment variable to check, and any subdirectory that you would like 
   to check as well as the default location, this function looks in a hell 
   of a lot of different places :-) Returns zero on success, and stores a 
   full path to the file (at most size bytes) into the dest buffer.

void packfile_password(const char *password);
   Sets the encryption password to be used for all read/write operations
   on files opened in future using Allegro's packfile functions (whether
   they are compressed or not), including all the save, load and config
   routines. Files written with an encryption password cannot be read
   unless the same password is selected, so be careful: if you forget the
   key, I can't make your data come back again! Pass NULL or an empty
   string to return to the normal, non-encrypted mode. If you are using
   this function to prevent people getting access to your datafiles, be
   careful not to store an obvious copy of the password in your executable:
   if there are any strings like "I'm the password for the datafile", it
   would be fairly easy to get access to your data :-)

   Note #1: when writing a packfile, you can change the password to whatever
   you want after opening the file, without affecting the write operation.
   On the contrary, when writing a sub-chunk of a packfile, you must make
   sure that the password that was active at the time the sub-chunk was
   opened is still active before closing the sub-chunk. This is guaranteed
   to be true if you didn't call the packfile_password() routine in the
   meantime. Read operations, either on packfiles or sub-chunks, have no
   such restriction.

   Note #2: as explained above, the password is used for all read/write
   operations on files, including for several functions of the library that
   operate on files without explicitly using packfiles, e.g load_bitmap().
   The unencrypted mode is mandatory in order for those functions to work.
   Therefore remember to call packfile_password(NULL) before using them if
   you previously changed the password. As a rule of thumb, always call
   packfile_password(NULL) when you are done with operations on packfiles.

PACKFILE *pack_fopen(const char *filename, const char *mode);
   Opens a file according to mode, which may contain any of the flags:

      'r' - open file for reading.

      'w' - open file for writing, overwriting any existing data.

      'p' - open file in packed mode. Data will be compressed as it is 
            written to the file, and automatically uncompressed during read 
            operations. Files created in this mode will produce garbage if 
            they are read without this flag being set. 

      '!' - open file for writing in normal, unpacked mode, but add the 
            value F_NOPACK_MAGIC to the start of the file, so that it can 
            later be opened in packed mode and Allegro will automatically 
            detect that the data does not need to be decompressed.

   Instead of these flags, one of the constants F_READ, F_WRITE, 
   F_READ_PACKED, F_WRITE_PACKED or F_WRITE_NOPACK may be used as the mode 
   parameter. On success, pack_fopen() returns a pointer to a file 
   structure, and on error it returns NULL and stores an error code in 
   errno. An attempt to read a normal file in packed mode will cause errno 
   to be set to EDOM.

   The packfile functions also understand several "magic" filenames that are 
   used for special purposes. These are:

      "#" - read data that has been appended to your executable file with 
      the exedat utility, as if it was a regular independent disk file.

      'filename.dat#object_name' - open a specific object from a datafile, 
      and read from it as if it was a regular file. You can treat nested 
      datafiles exactly like a normal directory structure, for example 
      you could open 'filename.dat#graphics/level1/mapdata'.

      '#object_name' - combination of the above, reading an object from a 
      datafile that has been appended onto your executable.

   With these special filenames, the contents of a datafile object or 
   appended file can be read in an identical way to a normal disk file, so 
   any of the file access functions in Allegro (eg. load_pcx() and 
   set_config_file()) can be used to read from them. Note that you can't 
   write to these special files, though: the fake file is read only. Also, 
   you must save your datafile uncompressed or with per-object compression 
   if you are planning on loading individual objects from it (otherwise 
   there will be an excessive amount of seeking when it is read). Finally, 
   be aware that the special Allegro object types aren't the same format as 
   the files you import the data from. When you import data like bitmaps or 
   samples into the grabber, they are converted into a special 
   Allegro-specific format, but the '#' marker file syntax reads the objects 
   as raw binary chunks. This means that if, for example, you want to use 
   load_pcx to read an image from a datafile, you should import it as a 
   binary block rather than as a BITMAP object.

int  pack_fclose(PACKFILE *f);
int  pack_fseek(PACKFILE *f, int offset);
int  pack_feof(PACKFILE *f);
int  pack_ferror(PACKFILE *f);
int  pack_getc(PACKFILE *f);
int  pack_putc(int c, PACKFILE *f);
int  pack_igetw(PACKFILE *f);
long pack_igetl(PACKFILE *f);
int  pack_iputw(int w, PACKFILE *f);
long pack_iputl(long l, PACKFILE *f);
int  pack_mgetw(PACKFILE *f);
long pack_mgetl(PACKFILE *f);
int  pack_mputw(int w, PACKFILE *f);
long pack_mputl(long l, PACKFILE *f);
long pack_fread(void *p, long n, PACKFILE *f);
long pack_fwrite(const void *p, long n, PACKFILE *f);
char *pack_fgets(char *p, int max, PACKFILE *f);
int  pack_fputs(const char *p, PACKFILE *f);

   These work like the equivalent stdio functions. There are some
   differences, however:

   Seeking only supports forward movement relative to the current position.
   Note that seeking is very slow when reading compressed files, and so
   should be avoided unless you are sure that the file is not compressed.

   The pack_i* and pack_m* routines read and write 16 and 32 bit values using
   the Intel and Motorola byte ordering systems (endianness) respectively.
   Intel is least significant byte first (little-endian); Motorola is most
   significant byte first (big-endian).

   pack_fread() and pack_fwrite() take a single size parameter instead of
   that silly size and num_elements system.

   The pack_fgets() function does not include a trailing carriage return in
   the returned string.

   pack_fputs() always writes in the UTF-8 text encoding format, converting
   from the current text encoding. Newlines (\n) are written as \r\n on
   DOS/Windows. If you do not want either of these things to happen,
   use pack_fwrite() and/or pack_putc() instead.

   pack_feof() returns nonzero as soon as you reach the end of the file. It 
   does not wait for you to attempt to read beyond the end of the file,
   contrary to the ISO C feof() function. The only way to know whether you
   have read beyond the end of the file is to check the return value of the
   read operation you use (and be wary of pack_*getl() as EOF is also a valid
   return value with these functions).

PACKFILE *pack_fopen_chunk(PACKFILE *f, int pack);
   Opens a sub-chunk of a file. Chunks are primarily intended for use by the 
   datafile code, but they may also be useful for your own file routines. A 
   chunk provides a logical view of part of a file, which can be compressed 
   as an individual entity and will automatically insert and check length 
   counts to prevent reading past the end of the chunk. To write a chunk to 
   the file f, use the code:

      /* assumes f is a PACKFILE * which has been opened */
      f = pack_fopen_chunk(f, pack);    /* in write mode */
      write some data to f
      f = pack_fclose_chunk(f);

   The data written to the chunk will be prefixed with two length counts (32 
   bit, big-endian). For uncompressed chunks these will both be set to the 
   size of the data in the chunk. For compressed chunks (created by setting 
   the pack flag), the first length will be the raw size of the chunk, and 
   the second will be the negative size of the uncompressed data.

   To read the chunk, use the code:

      /* assumes f is a PACKFILE * which has been opened */
      f = pack_fopen_chunk(f, FALSE);    */ in read mode */
      read data from f
      f = pack_fclose_chunk(f);

   This sequence will read the length counts created when the chunk was 
   written, and automatically decompress the contents of the chunk if it 
   was compressed. The length will also be used to prevent reading past the 
   end of the chunk (Allegro will return EOF if you attempt this), and to 
   automatically skip past any unread chunk data when you call 
   pack_fclose_chunk().

   Chunks can be nested inside each other by making repeated calls to 
   pack_fopen_chunk(). When writing a file, the compression status is 
   inherited from the parent file, so you only need to set the pack flag if 
   the parent is not compressed but you want to pack the chunk data. If the 
   parent file is already open in packed mode, setting the pack flag will 
   result in data being compressed twice: once as it is written to the 
   chunk, and again as the chunk passes it on to the parent file.

PACKFILE *pack_fclose_chunk(PACKFILE *f);
   Closes a sub-chunk of a file, previously obtained by calling 
   pack_fopen_chunk().



===========================================
============ Datafile routines ============
===========================================

Datafiles are created by the grabber utility, and have a .dat extension. 
They can contain bitmaps, palettes, fonts, samples, MIDI music, FLI/FLC 
animations, and any other binary data that you import.

Warning: when using truecolor images, you should always set the graphics 
mode before loading any bitmap data! Otherwise the pixel format (RGB or BGR)
will not be known, so the file may be converted wrongly.

See the documentation for pack_fopen() for information about how to read 
directly from a specific datafile object.

DATAFILE *load_datafile(const char *filename);
   Loads a datafile into memory, and returns a pointer to it, or NULL on 
   error. If the datafile has been encrypted, you must first use the 
   packfile_password() function to set the appropriate key. See grabber.txt 
   for more information. If the datafile contains truecolor graphics, you 
   must set the video mode or call set_color_conversion() before loading it.

DATAFILE *load_datafile_callback(const char *filename,
                                 void (*callback)(DATAFILE *d));
   Loads a datafile into memory, calling the specified hook function once 
   for each object in the file, passing it a pointer to the object just read.

void unload_datafile(DATAFILE *dat);
   Frees all the objects in a datafile.

DATAFILE *load_datafile_object(const char *filename, 
                                const char *objectname);
   Loads a specific object from a datafile. This won't work if you strip the 
   object names from the file, and it will be very slow if you save the file 
   with global compression. See grabber.txt for more information.

void unload_datafile_object(DATAFILE *dat);
   Frees an object previously loaded by load_datafile_object().

DATAFILE *find_datafile_object(const DATAFILE *dat, const char *objectname);
   Searches an already loaded datafile for an object with the specified 
   name, returning a pointer to it, or NULL if the object cannot be found. 
   It understands '/' and '#' separators for nested datafile paths.

const char *get_datafile_property(const DATAFILE *dat, int type);
   Returns the specified property string for the object, or an empty string 
   if the property isn't present. See grabber.txt for more information.

void register_datafile_object(int id, void *(*load)(PACKFILE *f, long size),
                                      void (*destroy)(void *data));
   Used to add custom object types, specifying functions to load and destroy 
   objects of this type. See grabber.txt for more information.

void fixup_datafile(DATAFILE *data);
   If you are using compiled datafiles (produced by the dat2s utility) on a
   platform that doesn't support constructors, or on a platform that does
   support constructors and the datafiles contain truecolor images, you
   must call this function once after your set the video mode that you will
   be using. This will ensure the datafiles are properly initialised in the
   first case and convert the color values into the appropriate format in
   the second case. It handles flipping between RGB and BGR formats, and
   converting between different color depths whenever that can be done
   without changing the size of the image (ie. changing between 15<->16
   bit hicolor for both bitmaps and RLE sprites, and 24<->32 bit
   truecolor for RLE sprites).

When you load a datafile, you will obtain a pointer to an array of DATAFILE 
structures:

typedef struct DATAFILE
{
   void *dat;     - pointer to the actual data
   int type;      - type of the data
   long size;     - size of the data in bytes
   void *prop;    - list of object properties
} DATAFILE;

The type field will be one of the values:
   DAT_FILE       - dat points to a nested datafile
   DAT_DATA       - dat points to a block of binary data
   DAT_FONT       - dat points to a font object
   DAT_SAMPLE     - dat points to a sample structure
   DAT_MIDI       - dat points to a MIDI file
   DAT_PATCH      - dat points to a GUS patch file
   DAT_FLI        - dat points to an FLI/FLC animation
   DAT_BITMAP     - dat points to a BITMAP structure
   DAT_RLE_SPRITE - dat points to a RLE_SPRITE structure
   DAT_C_SPRITE   - dat points to a linear compiled sprite
   DAT_XC_SPRITE  - dat points to a mode-X compiled sprite
   DAT_PALETTE    - dat points to an array of 256 RGB structures
   DAT_END        - special flag to mark the end of the data list

The grabber program can also produce a header file defining the index of 
each object within the file as a series of #defined constants, using the 
names you gave the objects in the grabber. So, for example, if you have made 
a datafile called foo.dat which contains a bitmap called THE_IMAGE, you 
could display it with the code fragment:

   #include "foo.h"

   DATAFILE *data = load_datafile("foo.dat");
   draw_sprite(screen, data[THE_IMAGE].dat, x, y);

If you are programming in C++ you will get an error because the dat field is 
a void pointer and draw_sprite() expects a BITMAP pointer. You can get 
around this with a cast, eg:

   draw_sprite(screen, (BITMAP *)data[THE_IMAGE].dat, x, y);

When you load a single datafile object, you will obtain a pointer to a 
single DATAFILE structure. This means that you don't access it any more like 
an array, and it doesn't have any DAT_END object. Example:

   music_object = load_datafile_object("datafile.dat", "MUSIC");
   play_midi(music_object->dat);




===================================================
============ Fixed point math routines ============
===================================================

Allegro provides some routines for working with fixed point numbers, and 
defines the type 'fixed' to be a signed 32 bit integer. The high word is 
used for the integer part and the low word for the fraction, giving a range 
of -32768 to 32767 and an accuracy of about four or five decimal places. 
Fixed point numbers can be assigned, compared, added, subtracted, negated 
and shifted (for multiplying or dividing by powers of two) using the normal 
integer operators, but you should take care to use the appropriate 
conversion routines when mixing fixed point with integer or floating point 
values. Writing 'fixed_point_1 + fixed_point_2' is ok, but 'fixed_point + 
integer' is not.

fixed itofix(int x);
   Converts an integer to fixed point. This is the same thing as x<<16.

int fixtoi(fixed x);
   Converts fixed point to integer, rounding as required.

int fixfloor(fixed x);
   Returns the greatest integer not greater than x. That is, it rounds 
   towards negative infinity.

int fixceil(fixed x);
   Returns the smallest integer not less than x. That is, it rounds towards
   positive infinity.

fixed ftofix(double x);
   Converts a floating point value to fixed point.

double fixtof(fixed x);
   Converts fixed point to floating point.

fixed fixmul(fixed x, fixed y);
   A fixed point value can be multiplied or divided by an integer with the 
   normal '*' and '/' operators. To multiply two fixed point values, though, 
   you must use this function.

   If an overflow or division by zero occurs, errno will be set and the 
   maximum possible value will be returned, but errno is not cleared if the 
   operation is successful. This means that if you are going to test for 
   overflow you should set errno=0 before calling fixmul().

fixed fixdiv(fixed x, fixed y);
   Fixed point division: see comments about fixmul().

fixed fixadd(fixed x, fixed y);
   Although fixed point numbers can be added with the normal '+' integer 
   operator, that doesn't provide any protection against overflow. If 
   overflow is a problem, you should use this function instead. It is slower 
   than using integer operators, but if an overflow occurs it will clamp the 
   result, rather than just letting it wrap, and set errno.

fixed fixsub(fixed x, fixed y);
   Fixed point subtraction: see comments about fixadd().

The fixed point square root, sin, cos, tan, inverse sin, and inverse cos 
functions are implemented using lookup tables, which are very fast but not 
particularly accurate. At the moment the inverse tan uses an iterative 
search on the tan table, so it is a lot slower than the others.

Angles are represented in a binary format with 256 equal to a full circle, 
64 being a right angle and so on. This has the advantage that a simple 
bitwise 'and' can be used to keep the angle within the range zero to a full 
circle, eliminating all those tiresome 'if (angle >= 360)' checks.

fixed fixsin(fixed x);
   Lookup table sine.

fixed fixcos(fixed x);
   Lookup table cosine.

fixed fixtan(fixed x);
   Lookup table tangent.

fixed fixasin(fixed x);
   Lookup table inverse sine.

fixed fixacos(fixed x);
   Lookup table inverse cosine.

fixed fixatan(fixed x);
   Fixed point inverse tangent.

fixed fixatan2(fixed y, fixed x);
   Fixed point version of the libc atan2() routine.

fixed fixsqrt(fixed x);
   Fixed point square root.

fixed fixhypot(fixed x, fixed y);
   Fixed point hypotenuse (returns the square root of x*x + y*y).


If you are programming in C++ you can ignore all the above and use the fix 
class instead, which overloads a lot of operators to provide automatic 
conversion to and from integer and floating point values, and calls the 
above routines as they are required. You should not mix the fix class with 
the fixed typedef though, because the compiler will mistake the fixed values 
for regular integers and insert unnecessary conversions. For example, if x 
is an object of class fix, calling fixsqrt(x) will return the wrong result. 
You should use the overloaded sqrt(x) or x.sqrt() instead.


The fixed point functions used to be named with an "f" prefix instead of
"fix", eg. fixsqrt() used to be fsqrt(), but were renamed due to conflicts
with some libc implementations. This should not affect most existing code as
there are backwards compatibility aliases. These aliases are static inline
functions which map the old names to the new names, eg. fsqrt() calls
fixsqrt(). You can disable the aliases by defining the preprocessor macro
ALLEGRO_NO_FIX_ALIASES before including allegro.h.



==========================================
============ 3D math routines ============
==========================================

Allegro contains some 3d helper functions for manipulating vectors, 
constructing and using transformation matrices, and doing perspective 
projections from 3d space onto the screen. It is not, and never will be, a 
fully fledged 3d library (my goal is to supply generic support routines, 
not shrink-wrapped graphics code :-) but these functions may be useful for 
developing your own 3d code.

Allegro uses a right-handed coordinate system, i.e. if you point the thumb 
of your right hand along the x axis, and the index finger along the y axis, 
your middle finger points in the direction of the z axis. This also means, 
for any rotation, if you point the thumb of your right hand along the axis 
of rotation, then the fingers curl in the positive direction of rotation.

All the 3d math functions are available in two versions: one which uses 
fixed point arithmetic, and another which uses floating point. The syntax 
for these is identical, but the floating point functions and structures are 
postfixed with '_f', eg. the fixed point function cross_product() has a 
floating point equivalent cross_product_f(). If you are programming in C++, 
Allegro also overloads these functions for use with the 'fix' class.

3d transformations are accomplished by the use of a modelling matrix. This 
is a 4x4 array of numbers that can be multiplied with a 3d point to produce 
a different 3d point. By putting the right values into the matrix, it can be 
made to do various operations like translation, rotation, and scaling. The 
clever bit is that you can multiply two matrices together to produce a third 
matrix, and this will have the same effect on points as applying the 
original two matrices one after the other. For example, if you have one 
matrix that rotates a point and another that shifts it sideways, you can 
combine them to produce a matrix that will do the rotation and the shift in 
a single step. You can build up extremely complex transformations in this 
way, while only ever having to multiply each point by a single matrix.

Allegro actually cheats in the way it implements the matrix structure. 
Rotation and scaling of a 3d point can be done with a simple 3x3 matrix, but 
in order to translate it and project it onto the screen, the matrix must be 
extended to 4x4, and the point extended into 4d space by the addition of an 
extra coordinate, w=1. This is a bad thing in terms of efficiency, but 
fortunately an optimisation is possible. Given the 4x4 matrix:

   ( a, b, c, d )
   ( e, f, g, h )
   ( i, j, k, l )
   ( m, n, o, p )

a pattern can be observed in which parts of it do what. The top left 3x3 
grid implements rotation and scaling. The three values in the top right 
column (d, h, and l) implement translation, and as long as the matrix is 
only used for affine transformations, m, n and o will always be zero and p 
will always be 1. If you don't know what affine means, read Foley & Van 
Damme: basically it covers scaling, translation, and rotation, but not 
projection. Since Allegro uses a separate function for projection, the 
matrix functions only need to support affine transformations, which means 
that there is no need to store the bottom row of the matrix. Allegro 
implicitly assumes that it contains (0,0,0,1), and optimises the matrix 
manipulation functions accordingly.

Matrices are stored in the structures:

typedef struct MATRIX            - fixed point matrix structure
{
   fixed v[3][3];                - 3x3 scaling and rotation component
   fixed t[3];                   - x/y/z translation component
} MATRIX;

typedef struct MATRIX_f          - floating point matrix structure
{
   float v[3][3];                - 3x3 scaling and rotation component
   float t[3];                   - x/y/z translation component
} MATRIX_f

extern MATRIX identity_matrix;
extern MATRIX_f identity_matrix_f;
   Global variables containing the 'do nothing' identity matrix. Multiplying 
   by the identity matrix has no effect.

void get_translation_matrix(MATRIX *m, fixed x, fixed y, fixed z);
void get_translation_matrix_f(MATRIX_f *m, float x, float y, float z);
   Constructs a translation matrix, storing it in m. When applied to the 
   point (px, py, pz), this matrix will produce the point (px+x, py+y, 
   pz+z). In other words, it moves things sideways.

void get_scaling_matrix(MATRIX *m, fixed x, fixed y, fixed z);
void get_scaling_matrix_f(MATRIX_f *m, float x, float y, float z);
   Constructs a scaling matrix, storing it in m. When applied to the point 
   (px, py, pz), this matrix will produce the point (px*x, py*y, pz*z). In 
   other words, it stretches or shrinks things.

void get_x_rotate_matrix(MATRIX *m, fixed r);
void get_x_rotate_matrix_f(MATRIX_f *m, float r);
   Construct X axis rotation matrices, storing them in m. When applied to a 
   point, these matrices will rotate it about the X axis by the specified 
   angle (given in binary, 256 degrees to a circle format).

void get_y_rotate_matrix(MATRIX *m, fixed r);
void get_y_rotate_matrix_f(MATRIX_f *m, float r);
   Construct Y axis rotation matrices, storing them in m. When applied to a 
   point, these matrices will rotate it about the Y axis by the specified 
   angle (given in binary, 256 degrees to a circle format).

void get_z_rotate_matrix(MATRIX *m, fixed r);
void get_z_rotate_matrix_f(MATRIX_f *m, float r);
   Construct Z axis rotation matrices, storing them in m. When applied to a 
   point, these matrices will rotate it about the Z axis by the specified 
   angle (given in binary, 256 degrees to a circle format).

void get_rotation_matrix(MATRIX *m, fixed x, fixed y, fixed z);
void get_rotation_matrix_f(MATRIX_f *m, float x, float y, float z);
   Constructs a transformation matrix which will rotate points around all 
   three axis by the specified amounts (given in binary, 256 degrees to a 
   circle format).

void get_align_matrix(MATRIX *m, fixed xfront, yfront, zfront, 
                                 fixed xup, fixed yup, fixed zup);
   Rotates a matrix so that it is aligned along the specified coordinate 
   vectors (they need not be normalized or perpendicular, but the up and 
   front must not be equal). A front vector of 1,0,0 and up vector of 0,1,0 
   will return the identity matrix.

void get_align_matrix_f(MATRIX *m, float xfront, yfront, zfront, 
                                   float xup, yup, zup);
   Floating point version of get_align_matrix().

void get_vector_rotation_matrix(MATRIX *m, fixed x, y, z, fixed a);
void get_vector_rotation_matrix_f(MATRIX_f *m, float x, y, z, float a);
   Constructs a transformation matrix which will rotate points around the 
   specified x,y,z vector by the specified angle (given in binary, 256 
   degrees to a circle format).

void get_transformation_matrix(MATRIX *m, fixed scale,
                               fixed xrot, yrot, zrot, x, y, z);
   Constructs a transformation matrix which will rotate points around all 
   three axis by the specified amounts (given in binary, 256 degrees to a 
   circle format), scale the result by the specified amount (pass 1 for no 
   change of scale), and then translate to the requested x, y, z position.

void get_transformation_matrix_f(MATRIX_f *m, float scale,
                                 float xrot, yrot, zrot, x, y, z);
   Floating point version of get_transformation_matrix().

void get_camera_matrix(MATRIX *m, fixed x, y, z, xfront, yfront, zfront,
                       fixed xup, yup, zup, fov, aspect);
   Constructs a camera matrix for translating world-space objects into a 
   normalised view space, ready for the perspective projection. The x, y, 
   and z parameters specify the camera position, xfront, yfront, and zfront 
   are the 'in front' vector specifying which way the camera is facing (this 
   can be any length: normalisation is not required), and xup, yup, and zup 
   are the 'up' direction vector. The fov parameter specifies the field of 
   view (ie. width of the camera focus) in binary, 256 degrees to the circle 
   format. For typical projections, a field of view in the region 32-48 will 
   work well. Finally, the aspect ratio is used to scale the Y dimensions of 
   the image relative to the X axis, so you can use it to adjust the 
   proportions of the output image (set it to 1 for no scaling).

void get_camera_matrix_f(MATRIX_f *m, float x, y, z, xfront, yfront, zfront,
                         float xup, yup, zup, fov, aspect);
   Floating point version of get_camera_matrix().

void qtranslate_matrix(MATRIX *m, fixed x, fixed y, fixed z);
void qtranslate_matrix_f(MATRIX_f *m, float x, float y, float z);
   Optimised routine for translating an already generated matrix: this 
   simply adds in the translation offset, so there is no need to build two 
   temporary matrices and then multiply them together.

void qscale_matrix(MATRIX *m, fixed scale);
void qscale_matrix_f(MATRIX_f *m, float scale);
   Optimised routine for scaling an already generated matrix: this simply 
   adds in the scale factor, so there is no need to build two temporary 
   matrices and then multiply them together.

void matrix_mul(const MATRIX *m1, *m2, MATRIX *out);
void matrix_mul_f(const MATRIX_f *m1, *m2, MATRIX_f *out);
   Multiplies two matrices, storing the result in out (this may be a 
   duplicate of one of the input matrices, but it is faster when the inputs 
   and output are all different). The resulting matrix will have the same 
   effect as the combination of m1 and m2, ie. when applied to a point p, (p 
   * out) = ((p * m1) * m2). Any number of transformations can be 
   concatenated in this way. Note that matrix multiplication is not 
   commutative, ie. matrix_mul(m1, m2) != matrix_mul(m2, m1).

fixed vector_length(fixed x, fixed y, fixed z);
float vector_length_f(float x, float y, float z);
   Calculates the length of the vector (x, y, z), using that good 'ole 
   Pythagoras theorem.

void normalize_vector(fixed *x, fixed *y, fixed *z);
void normalize_vector_f(float *x, float *y, float *z);
   Converts the vector (*x, *y, *z) to a unit vector. This points in the 
   same direction as the original vector, but has a length of one.

fixed dot_product(fixed x1, y1, z1, x2, y2, z2);
float dot_product_f(float x1, y1, z1, x2, y2, z2);
   Calculates the dot product (x1, y1, z1) . (x2, y2, z2), returning the 
   result.

void cross_product(fixed x1, y1, z1, x2, y2, z2, *xout, *yout, *zout);
void cross_product_f(float x1, y1, z1, x2, y2, z2, *xout, *yout, *zout);
   Calculates the cross product (x1, y1, z1) x (x2, y2, z2), storing the 
   result in (*xout, *yout, *zout). The cross product is perpendicular to 
   both of the input vectors, so it can be used to generate polygon normals.

fixed polygon_z_normal(const V3D *v1, *v2, *v3);
float polygon_z_normal_f(const V3D_f *v1, *v2, *v3);
   Finds the Z component of the normal vector to the specified three 
   vertices (which must be part of a convex polygon). This is used mainly in 
   back-face culling. The back-faces of closed polyhedra are never visible 
   to the viewer, therefore they never need to be drawn. This can cull on 
   average half the polygons from a scene. If the normal is negative the 
   polygon can safely be culled. If it is zero, the polygon is perpendicular 
   to the screen.

void apply_matrix(const MATRIX *m, fixed x, y, z, *xout, *yout, *zout);
void apply_matrix_f(const MATRIX_f *m, float x, y, z, *xout, *yout, *zout);
   Multiplies the point (x, y, z) by the transformation matrix m, storing 
   the result in (*xout, *yout, *zout).

void set_projection_viewport(int x, int y, int w, int h);
   Sets the viewport used to scale the output of the persp_project() 
   function. Pass the dimensions of the screen area you want to draw onto, 
   which will typically be 0, 0, SCREEN_W, and SCREEN_H.

void persp_project(fixed x, y, z, *xout, *yout);
void persp_project_f(float x, y, z, *xout, *yout);
   Projects the 3d point (x, y, z) into 2d screen space, storing the result 
   in (*xout, *yout) and using the scaling parameters previously set by 
   calling set_projection_viewport(). This function projects from the 
   normalized viewing pyramid, which has a camera at the origin and facing 
   along the positive z axis. The x axis runs left/right, y runs up/down, 
   and z increases with depth into the screen. The camera has a 90 degree 
   field of view, ie. points on the planes x=z and -x=z will map onto the 
   left and right edges of the screen, and the planes y=z and -y=z map to 
   the top and bottom of the screen. If you want a different field of view 
   or camera location, you should transform all your objects with an 
   appropriate viewing matrix, eg. to get the effect of panning the camera 
   10 degrees to the left, rotate all your objects 10 degrees to the right.



==================================================
============ Quaternion math routines ============
==================================================

Quaternions are an alternate way to represent the rotation part of a
transformation, and can be easier to manipulate than matrices. As with a 
matrix, you can encode a geometric transformations in one, concatenate 
several of them to merge multiple transformations, and apply them to a 
vector, but they can only store pure rotations. The big advantage is that 
you can accurately interpolate between two quaternions to get a part-way 
rotation, avoiding the gimbal problems of the more conventional euler angle 
interpolation.

Quaternions only have floating point versions, without any _f suffix. Other 
than that, most of the quaternion functions correspond with a matrix 
function that performs a similar operation.

Quaternion means 'of four parts', and that's exactly what it is. Here is the 
structure:

typedef struct QUAT
{
   float w, x, y, z;
}

You will have lots of fun figuring out what these numbers actually mean, but 
that is beyond the scope of this documentation. Quaternions do work -- trust 
me.

extern QUAT identity_quat;
   Global variable containing the 'do nothing' identity quaternion. 
   Multiplying by the identity quaternion has no effect.

void get_x_rotate_quat(QUAT *q, float r);
void get_y_rotate_quat(QUAT *q, float r);
void get_z_rotate_quat(QUAT *q, float r);
   Construct axis rotation quaternions, storing them in q. When applied to a 
   point, these quaternions will rotate it about the relevant axis by the 
   specified angle (given in binary, 256 degrees to a circle format).

void get_rotation_quat(QUAT *q, float x, float y, float z);
   Constructs a quaternion that will rotate points around all three axis by 
   the specified amounts (given in binary, 256 degrees to a circle format).

void get_vector_rotation_quat(QUAT *q, float x, y, z, float a);
   Constructs a quaternion that will rotate points around the specified 
   x,y,z vector by the specified angle (given in binary, 256 degrees to a 
   circle format).

void quat_to_matrix(const QUAT *q, MATRIX_f *m);
   Constructs a rotation matrix from a quaternion.

void matrix_to_quat(const MATRIX_f *m, QUAT *q);
   Constructs a quaternion from a rotation matrix. Translation is discarded 
   during the conversion. Use get_align_matrix_f() if the matrix is not 
   orthonormalized, because strange things may happen otherwise.

void quat_mul(const QUAT *p, const QUAT *q, QUAT *out);
   Multiplies two quaternions, storing the result in out. The resulting 
   quaternion will have the same effect as the combination of p and q, ie. 
   when applied to a point, (point * out) = ((point * p) * q). Any number of 
   rotations can be concatenated in this way. Note that quaternion 
   multiplication is not commutative, ie. quat_mul(p, q) != quat_mul(q, p). 

void apply_quat(const QUAT *q, float x, y, z, *xout, *yout, *zout);
   Multiplies the point (x, y, z) by the quaternion q, storing the result in 
   (*xout, *yout, *zout). This is quite a bit slower than apply_matrix_f(), 
   so only use it to translate a few points. If you have many points, it is 
   much more efficient to call quat_to_matrix() and then use 
   apply_matrix_f().

void quat_interpolate(const QUAT *from, *to, float t, QUAT *out);
   Constructs a quaternion that represents a rotation between from and to. 
   The argument t can be anything between 0 and 1 and represents where 
   between from and to the result will be. 0 returns from, 1 returns to, and 
   0.5 will return a rotation exactly in between. The result is copied to 
   out. This function will create the short rotation (less than 180 degrees) 
   between from and to.

void quat_slerp(const QUAT *from, *to, float t, QUAT *out, int how);
   The same as quat_interpolate(), but allows more control over how the 
   rotation is done. The how parameter can be any one of the values:

   QUAT_SHORT  - like quat_interpolate(), use shortest path
   QUAT_LONG   - rotation will be greater than 180 degrees
   QUAT_CW     - rotate clockwise when viewed from above
   QUAT_CCW    - rotate counterclockwise when viewed from above
   QUAT_USER   - the quaternions are interpolated exactly as given



======================================
============ GUI routines ============
======================================

Allegro contains an object-oriented dialog manager, which was originally 
based on the Atari GEM system (form_do(), objc_draw(), etc: old ST 
programmers will know what I'm talking about :-) You can use the GUI as-is 
to knock out simple interfaces for things like the test program and grabber 
utility, or you can use it as a basis for more complicated systems of your 
own. Allegro lets you define your own object types by writing new dialog 
procedures, so you can take complete control over the visual aspects of the 
interface while still using Allegro to handle input from the mouse, 
keyboard, joystick, etc.

A GUI dialog is stored as an array of DIALOG objects, each one containing 
the fields:

typedef struct DIALOG
{
   int (*proc)(int, DIALOG *, int); - dialog procedure (message handler)
   int x, y, w, h;                  - position and size of the object
   int fg, bg;                      - foreground and background colors
   int key;                         - ASCII keyboard shortcut
   int flags;                       - flags about the status of
                                      the object
   int d1, d2;                      - whatever you want to use them for
   void *dp, *dp2, *dp3;            - pointers to more
                                      object-specific data
} DIALOG;

The array should end with an object which has the proc pointer set to NULL.

The flags field may contain any combination of the bit flags:
   D_EXIT          - this object should close the dialog when it is clicked
   D_SELECTED      - this object is selected
   D_GOTFOCUS      - this object has got the input focus
   D_GOTMOUSE      - the mouse is currently on top of this object
   D_HIDDEN        - this object is hidden and inactive
   D_DISABLED      - this object is greyed-out and inactive
   D_DIRTY         - this object needs to be redrawn
   D_INTERNAL      - don't use this! It is for internal use by the
                     library...
   D_USER          - any powers of two above this are free for your own use

Each object is controlled by a dialog procedure, which is stored in the proc 
pointer. This will be called by the dialog manager whenever any action 
concerning the object is required, or you can call it directly with the 
object_message() function. The dialog procedure should follow the form:

   int foo(int msg, DIALOG *d, int c);

It will be passed a flag (msg) indicating what action it should perform, a 
pointer to the object concerned (d), and if msg is MSG_CHAR or MSG_XCHAR, 
the key that was pressed (c). Note that d is a pointer to a specific object, 
and not to the entire dialog.

The dialog procedure should return one of the values:
   D_O_K          - normal return status
   D_CLOSE        - tells the dialog manager to close the dialog
   D_REDRAW       - tells the dialog manager to redraw the entire dialog
   D_REDRAWME     - tells the dialog manager to redraw the current object
   D_WANTFOCUS    - requests that the input focus be given to this object
   D_USED_CHAR    - MSG_CHAR and MSG_XCHAR return this if they used the key

Dialog procedures may be called with any of the messages:

MSG_START:
   Tells the object to initialise itself. The dialog manager sends this to 
   all the objects in a dialog just before it displays the dialog.

MSG_END:
   Sent to all objects when closing a dialog, allowing them to perform 
   whatever cleanup operations they require.

MSG_DRAW:
   Tells the object to draw itself onto the screen. The mouse pointer will 
   be turned off when this message is sent, so the drawing code does not 
   need to worry about it.

MSG_CLICK:
   Informs the object that a mouse button has been clicked while the mouse 
   was on top of the object. Typically an object will perform its own mouse 
   tracking as long as the button is held down, and only return from this 
   message handler when it is released.

MSG_DCLICK:
   Sent when the user double-clicks on an object. A MSG_CLICK will be sent 
   when the button is first pressed, then MSG_DCLICK if it is released and 
   pressed again within a short space of time.

MSG_KEY:
   Sent when the keyboard shortcut for the object is pressed, or if enter, 
   space, or a joystick button is pressed while it has the input focus.

MSG_CHAR:
   When a key is pressed, this message is sent to the object that has the 
   input focus, with a readkey() format character code (ASCII value in the 
   low byte, scancode in the high byte) as the c parameter. If the object 
   deals with the keypress it should return D_USED_CHAR, otherwise it should 
   return D_O_K to allow the default keyboard interface to operate. If you 
   need to access Unicode character input, you should use MSG_UCHAR instead 
   of MSG_CHAR.

MSG_UCHAR:
   If an object ignores the MSG_CHAR input, this message will be sent 
   immediately after it, passed the full Unicode key value as the c 
   parameter. This enables you to read character codes greater than 255, but 
   cannot tell you anything about the scancode: if you need to know that, 
   use MSG_CHAR instead. This handler should return D_USED_CHAR if it 
   processed the input, or D_O_K otherwise.

MSG_XCHAR:
   When a key is pressed, Allegro will send a MSG_CHAR and MSG_UCHAR to the 
   object with the input focus. If this object doesn't process the key (ie. 
   it returns D_O_K rather than D_USED_CHAR), the dialog manager will look 
   for an object with a matching keyboard shortcut in the key field, and 
   send it a MSG_KEY. If this fails, it broadcasts a MSG_XCHAR to all 
   objects in the dialog, allowing them to respond to special keypresses 
   even when they don't have the input focus. Normally you should ignore 
   this message (return D_O_K rather than D_USED_CHAR), in which case 
   Allegro will perform default actions such as moving the focus in response 
   to the arrow keys and closing the dialog if ESC is pressed.

MSG_WANTFOCUS:
   Queries whether an object is willing to accept the input focus. It should 
   return D_WANTFOCUS if it does, or D_O_K if it isn't interested in getting 
   user input.

MSG_GOTFOCUS:
MSG_LOSTFOCUS:
   Sent whenever an object gains or loses the input focus. These messages 
   will always be followed by a MSG_DRAW, to let objects display themselves 
   differently when they have the input focus. If you return D_WANTFOCUS in 
   response to a MSG_LOSTFOCUS event, this will prevent your object from 
   losing the focus when the mouse moves off it onto the screen background 
   or some inert object, so it will only lose the input focus when some 
   other object is ready to take over the focus (this trick is used by the 
   d_edit_proc() object).

MSG_GOTMOUSE:
MSG_LOSTMOUSE:
   Sent when the mouse moves on top of or away from an object. Unlike the 
   focus messages, these are not followed by a MSG_DRAW, so if the object is 
   displayed differently when the mouse is on top of it, it is responsible 
   for redrawing itself in response to these messages.

MSG_IDLE:
   Sent whenever the dialog manager has nothing better to do.

MSG_RADIO:
   Sent by radio button objects to deselect other buttons in the same group 
   when they are clicked. The group number is passed in the c message 
   parameter.

MSG_WHEEL:
   Sent to the focused object whenever the mouse wheel moves. The c message 
   parameter contains the number of clicks.

MSG_LPRESS, MSG_MPRESS, MSG_RPRESS:
   Sent when the corresponding mouse button is pressed.

MSG_LRELEASE, MSG_MRELEASE, MSG_RRELEASE:
   Sent when the corresponding mouse button is released.

MSG_USER:
   The first free message value. Any numbers from here on (MSG_USER, 
   MSG_USER+1, MSG_USER+2, ... MSG_USER+n) are free to use for whatever you 
   like.

Allegro provides several standard dialog procedures. You can use these as 
they are to provide simple user interface objects, or you can call them from 
within your own dialog procedures, resulting in a kind of OOP inheritance. 
For instance, you could make an object which calls d_button_proc to draw 
itself, but handles the click message in a different way, or an object which 
calls d_button_proc for everything except drawing itself, so it would behave 
like a normal button but could look completely different.

Since the release of Allegro version 3.9.33 (CVS), some GUI objects and 
menus are being drawn differently unlike in previous Allegro versions. The 
changes are the following:

   Shadows under d_shadow_box_proc and d_button_proc are always black.

   The most important (and immediately visible) change is, that some objects 
   are being drawn smaller. The difference is exactly one pixel in both 
   height and width, when comparing to previous versions. The reason is, 
   that in previous versions these objects were too large on the screen - 
   their size was d->w+1 and d->h+1 pixels (and not d->w and d->h, as it 
   should be). This change affects the following objects :

          d_box_proc,
          d_shadow_box_proc,
          d_button_proc,
          d_check_proc,
          d_radio_proc,
          d_list_proc,
          d_text_list_proc and
          d_textbox_proc.

   When you want to convert old dialogs to look equally when compiling with 
   the new Allegro version, just increase the size of the mentioned objects 
   by one pixel in both width and height fields. 

   When a GUI menu item (not in a bar menu) has a child menu, there is a 
   small arrow next to the child menu name, pointing to the right - so the 
   user can immediately see that this menu item has a child menu - and 
   there is no need to use such menu item names as for example "New...", 
   to show that it has a child menu. The submenu will be drawn to the right 
   of the parent menu, trying not to overlap it.


int d_clear_proc(int msg, DIALOG *d, int c);
   This just clears the screen when it is drawn. Useful as the first object 
   in a dialog.

int d_box_proc(int msg, DIALOG *d, int c);
int d_shadow_box_proc(int msg, DIALOG *d, int c);
   These draw boxes onto the screen, with or without a shadow.

int d_bitmap_proc(int msg, DIALOG *d, int c);
   This draws a bitmap onto the screen, which should be pointed to by the 
   dp field.

int d_text_proc(int msg, DIALOG *d, int c);
int d_ctext_proc(int msg, DIALOG *d, int c);
int d_rtext_proc(int msg, DIALOG *d, int c);
   These draw text onto the screen. The dp field should point to the string 
   to display. d_ctext_proc() centres the string around the x coordinate, 
   and d_rtext_proc() right aligns it. Any '&' characters in the string will 
   be replaced with lines underneath the following character, for displaying 
   keyboard shortcuts (as in MS Windows). To display a single ampersand, put 
   "&&". To draw the text in something other than the default font, set the 
   dp2 field to point to your custom font data.

int d_button_proc(int msg, DIALOG *d, int c);
   A button object (the dp field points to the text string). This object can 
   be selected by clicking on it with the mouse or by pressing its keyboard 
   shortcut. If the D_EXIT flag is set, selecting it will close the dialog, 
   otherwise it will toggle on and off. Like d_text_proc(), ampersands can 
   be used to display the keyboard shortcut of the button.

int d_check_proc(int msg, DIALOG *d, int c);
   This is an example of how you can derive objects from other objects. Most 
   of the functionality comes from d_button_proc(), but it displays itself 
   as a check box. If the d1 field is non-zero, the text will be printed to 
   the right of the check, otherwise it will be on the left.

   Note: the object width should allow space for the text as well as the
   check box (which is square, with sides equal to the object height).

int d_radio_proc(int msg, DIALOG *d, int c);
   A radio button object. A dialog can contain any number of radio button 
   groups: selecting a radio button causes other buttons within the same 
   group to be deselected. The dp field points to the text string, d1 
   specifies the group number, and d2 is the button style (0=circle, 
   1=square).

int d_icon_proc(int msg, DIALOG *d, int c);
   A bitmap button. The fg color is used for the dotted line showing focus, 
   and the bg color for the shadow used to fill in the top and left sides of 
   the button when "pressed". d1 is the "push depth", ie. the number of 
   pixels the icon will be shifted to the right and down when selected 
   (default 2) if there is no "selected" image. d2 is the distance by which 
   the dotted line showing focus is indented (default 2). dp points to a 
   bitmap for the icon, while dp2 and dp3 are the selected and disabled 
   images respectively (optional, may be NULL).

int d_keyboard_proc(int msg, DIALOG *d, int c);
   This is an invisible object for implementing keyboard shortcuts. You can 
   put an ASCII code in the key field of the dialog object (a character such 
   as 'a' to respond to a simple keypress, or a number 1-26 to respond to a 
   control key a-z), or you can put a keyboard scancode in the d1 and/or d2 
   fields. When one of these keys is pressed, the object will call the 
   function pointed to by dp. This should return an int, which will be 
   passed back to the dialog manager, so it can return D_O_K, D_REDRAW, 
   D_CLOSE, etc.

int d_edit_proc(int msg, DIALOG *d, int c);
   An editable text object (the dp field points to the string). When it has 
   the input focus (obtained by clicking on it with the mouse), text can be 
   typed into this object. The d1 field specifies the maximum number of 
   characters that it will accept, and d2 is the text cursor position within 
   the string.

int d_list_proc(int msg, DIALOG *d, int c);
   A list box object. This will allow the user to scroll through a list of 
   items and to select one by clicking or with the arrow keys. If the D_EXIT 
   flag is set, double clicking on a list item will close the dialog. The 
   index of the selected item is held in the d1 field, and d2 is used to 
   store how far it has scrolled through the list. The dp field points to a 
   function which will be called to obtain information about the contents of 
   the list. This should follow the form:

      char *foobar(int index, int *list_size);

   If index is zero or positive, the function should return a pointer to the 
   string which is to be displayed at position index in the list. If index 
   is negative, it should return NULL and list_size should be set to the 
   number of items in the list. 

   To create a multiple selection listbox, set the dp2 field to an array of 
   byte flags indicating the selection state of each list item (non-zero for 
   selected entries). This table must be at least as big as the number of 
   objects in the list!

int d_text_list_proc(int msg, DIALOG *d, int c);
   Like d_list_proc, but allows the user to type in the first few characters 
   of a listbox entry in order to select it. Uses dp3 internally, so you 
   mustn't store anything important there yourself.

int d_textbox_proc(int msg, DIALOG *d, int c);
   A text box object. The dp field points to the text which is to be 
   displayed in the box. If the text is long, there will be a vertical 
   scrollbar on the right hand side of the object which can be used to 
   scroll through the text. The default is to print the text with word 
   wrapping, but if the D_SELECTED flag is set, the text will be printed 
   with character wrapping. The d1 field is used internally to store the 
   number of lines of text, and d2 is used to store how far it has scrolled 
   through the text.

int d_slider_proc(int msg, DIALOG *d, int c);
   A slider control object. This object holds a value in d2, in the range 
   from 0 to d1. It will display as a vertical slider if h is greater than 
   or equal to w, otherwise it will display as a horizontal slider. The dp 
   field can contain an optional bitmap to use for the slider handle, and 
   dp2 can contain an optional callback function, which is called each time 
   d2 changes. The callback function should have the following prototype:

      int function(void *dp3, int d2);

   The d_slider_proc object will return the value of the callback function.

int d_menu_proc(int msg, DIALOG *d, int c);
   This object is a menu bar which will drop down child menus when it is 
   clicked or if an alt+key corresponding to one of the shortcuts in the 
   menu is pressed. It ignores a lot of the fields in the dialog structure, 
   in particular the color is taken from the gui_*_color variables, and the 
   width and height are calculated automatically (the w and h fields from 
   the DIALOG are only used as a minimum size.) The dp field points to an 
   array of menu structures: see do_menu() for more information. The top 
   level menu will be displayed as a horizontal bar, but when child menus 
   drop down from it they will be in the normal vertical format used by 
   do_menu(). When a menu item is selected, the return value from the menu 
   callback function is passed back to the dialog manager, so your callbacks 
   should return D_O_K, D_REDRAW, or D_CLOSE.

int d_yield_proc(int msg, DIALOG *d, int c);
   An invisible helper object that yields timeslices for the scheduler (if
   the system supports it) when the gui has nothing to do but waiting for
   user actions. You should put one instance of this object in each dialog 
   array because it may be needed on systems with an unusual scheduling 
   algorithm (for instance QNX) in order to make the GUI fully responsive.

The behaviour of the dialog manager can be controlled by the variables:

extern int gui_mouse_focus;
   If set, the input focus follows the mouse pointer around the dialog, 
   otherwise a click is required to move it.

extern int gui_fg_color;
extern int gui_bg_color;
   The foreground and background colors for the standard dialogs (alerts, 
   menus, and the file selector). They default to 255 and 0.

extern int gui_mg_color;
   The color used for displaying greyed-out dialog objects (with the 
   D_DISABLED flag set). Defaults to 8.

extern int gui_font_baseline;
   If set to a non-zero value, adjusts the keyboard shortcut underscores to 
   account for the height of the descenders in your font.

extern int (*gui_mouse_x)();
extern int (*gui_mouse_y)();
extern int (*gui_mouse_z)();
extern int (*gui_mouse_b)();
   Hook functions, used by the GUI routines whenever they need to access the 
   mouse state. By default these just return copies of the mouse_x, mouse_y, 
   mouse_z, and mouse_b variables, but they could be used to offset or scale 
   the mouse position, or read input from a different source entirely.

You can change the global 'font' pointer to make the GUI objects use 
something other than the standard 8x8 font. The standard dialog procedures, 
menus, and alert boxes, will work with fonts of any size, but the 
gfx_mode_select() dialog will look wrong with anything other than 8x8 fonts.

int gui_textout(BITMAP *bmp, const char *s, int x, y, color, centre);
   Helper function for use by the GUI routines. Draws a text string onto the 
   screen, interpreting the '&' character as an underbar for displaying 
   keyboard shortcuts. Returns the width of the output string in pixels.

int gui_strlen(const char *s);
   Helper function for use by the GUI routines. Returns the length of a 
   string in pixels, ignoring '&' characters.

void position_dialog(DIALOG *dialog, int x, int y);
   Moves an array of dialog objects to the specified screen position 
   (specified as the top left corner of the dialog).

void centre_dialog(DIALOG *dialog);
   Moves an array of dialog objects so that it is centered in the screen.

void set_dialog_color(DIALOG *dialog, int fg, int bg);
   Sets the foreground and background colors of an array of dialog objects.

int find_dialog_focus(DIALOG *dialog);
   Searches the dialog for the object which has the input focus, returning 
   an index or -1 if the focus is not set. This is useful if you are calling 
   do_dialog() several times in a row and want to leave the focus in the 
   same place it was when the dialog was last displayed, as you can call 
   do_dialog(dlg, find_dialog_focus(dlg));

int offer_focus(DIALOG *d, int obj, int *focus_obj, int force);
   Offers the input focus to a particular object. Normally the function sends
   the MSG_WANTFOCUS message to query whether the object is willing to accept
   the focus. However, passing any non zero value as force argument instructs
   the function to authoritatively set the focus to the object.

int object_message(DIALOG *dialog, int msg, int c);
   Sends a message to an object and returns the answer it has generated.
   Remember that the first parameter is the dialog object (not a whole
   array) that you wish to send the message to. For example, to make the
   second object in a dialog draw itself, you might write:

      object_message(&dialog[1], MSG_DRAW, 0);

int dialog_message(DIALOG *dialog, int msg, int c, int *obj);
   Sends a message to all the objects in an array. If any of the dialog 
   procedures return values other than D_O_K, it returns the value and sets 
   obj to the index of the object which produced it.

int broadcast_dialog_message(int msg, int c);
   Broadcasts a message to all the objects in the active dialog. If any of 
   the dialog procedures return values other than D_O_K, it returns that 
   value.

int do_dialog(DIALOG *dialog, int focus_obj);
   The basic dialog manager function. This displays a dialog (an array of 
   dialog objects, terminated by one with a NULL dialog procedure), and sets 
   the input focus to the focus_obj (-1 if you don't want anything to have 
   the focus). It interprets user input and dispatches messages as they are 
   required, until one of the dialog procedures tells it to close the 
   dialog, at which point it returns the index of the object that caused it 
   to exit.

int popup_dialog(DIALOG *dialog, int focus_obj);
   Like do_dialog(), but it stores the data on the screen before drawing the 
   dialog and restores it when the dialog is closed. The screen area to be 
   stored is calculated from the dimensions of the first object in the 
   dialog, so all the other objects should lie within this one.

DIALOG_PLAYER *init_dialog(DIALOG *dialog, int focus_obj);
   This function provides lower level access to the same functionality as 
   do_dialog(), but allows you to combine a dialog box with your own program 
   control structures. It initialises a dialog, returning a pointer to a 
   player object that can be used with update_dialog() and 
   shutdown_dialog(). With these functions, you could implement your own 
   version of do_dialog() with the lines:

      DIALOG_PLAYER *player = init_dialog(dialog, focus_obj);

      while (update_dialog(player))
         ;

      return shutdown_dialog(player);

int update_dialog(DIALOG_PLAYER *player);
   Updates the status of a dialog object returned by init_dialog(). Returns 
   TRUE if the dialog is still active, or FALSE if it has terminated. Upon a 
   return value of FALSE, it is up to you whether to call shutdown_dialog() 
   or to continue execution. The object that requested the exit can be 
   determined from the player->obj field.

int shutdown_dialog(DIALOG_PLAYER *player);
   Destroys a dialog player object returned by init_dialog(), returning the 
   object that caused it to exit (this is the same as the return value from 
   do_dialog()).

extern DIALOG *active_dialog;
   Global pointer to the most recent activated dialog. This may be useful if 
   an object needs to iterate through a list of all its siblings.

Popup or pulldown menus are created as an array of the structures:

typedef struct MENU
{
   char *text;                   - the text to display for the menu item
   int (*proc)(void);            - called when the menu item is clicked
   struct MENU *child;           - nested child menu
   int flags;                    - disabled or checked state
   void *dp;                     - pointer to any data you need
} MENU;

Each menu item contains a text string. This can use the '&' character to 
indicate keyboard shortcuts, or can be an zero-length string to display the 
item as a non-selectable splitter bar. If the string contains a "\t" tab 
character, any text after this will be right-justified, eg. for displaying 
keyboard shortcut information. The proc pointer is a function which will be 
called when the menu item is selected, and child points to another menu, 
allowing you to create nested menus. Both proc and child may be NULL. The 
proc function returns an integer which is ignored if the menu was brought up 
by calling do_menu(), but which is passed back to the dialog manager if it 
was created by a d_menu_proc() object. The array of menu items is terminated 
by an entry with a NULL text pointer.

Menu items can be disabled (greyed-out) by setting the D_DISABLED bit in the 
flags field, and a check mark can be displayed next to them by setting the 
D_SELECTED bit. With the default alignment and font this will usually 
overlap the menu text, so if you are going to use checked menu items it 
would be a good idea to prefix all your options with a space or two, to 
ensure there is room for the check.

int do_menu(MENU *menu, int x, int y)
   Displays and animates a popup menu at the specified screen coordinates 
   (these will be adjusted if the menu does not entirely fit on the screen). 
   Returns the index of the menu item that was selected, or -1 if the menu 
   was cancelled. Note that the return value cannot indicate selection from 
   child menus, so you will have to use the callback functions if you want 
   multi-level menus.

extern MENU *active_menu;
   When a menu callback procedure is triggered, this will be set to the menu 
   item that was selected, so your routine can determine where it was called 
   from.

extern void (*gui_menu_draw_menu)(int x, int y, int w, int h);
extern void (*gui_menu_draw_menu_item)(MENU *m, int x, int y, int w,
                                        int h, int bar, int sel);
   If set, these functions will be called whenever a menu needs to be
   drawn, so you can change how menus look.

   gui_menu_draw_menu() is passed the position and size of the
   menu. It should draw the background of the menu onto screen.

   gui_menu_draw_menu_item() is called once for each menu item that is
   to be drawn. bar will be set if the item is part of a top-level
   horizontal menu bar, and sel will be set if the menu item is
   selected. It should also draw onto screen.

int alert(const char *s1, *s2, *s3, const char *b1, *b2, int c1, c2);
   Displays a popup alert box, containing three lines of text (s1-s3), and 
   with either one or two buttons. The text for these buttons is passed in 
   b1 and b2 (b2 may be NULL), and the keyboard shortcuts in c1 and c2. 
   Returns 1 or 2 depending on which button was clicked. If the alert is 
   dismissed by pressing ESC when ESC is not one of the keyboard shortcuts, 
   it treats it as a click on the second button (this is consistent with the 
   common "Ok", "Cancel" alert).

int alert3(const char *s1, *s2, *s3, const char *b1, *b2, *b3, int c1, c2, c3);
   Like alert(), but with three buttons. Returns 1, 2, or 3.

int file_select(const char *message, char *path, const char *ext);
   Deprecated. Use file_select_ex() instead, passing the two constants 
   OLD_FILESEL_WIDTH and OLD_FILESEL_HEIGHT if you want the file selector 
   to be displayed with the dimensions of the old file selector.

int file_select_ex(const char *message, char *path, const char *ext,
                    int size, int w, int h);
   Displays the Allegro file selector, with the message as caption. The path 
   parameter contains the initial filename to display (this can be used to 
   set the starting directory, or to provide a default filename for a 
   save-as operation). The user selection is returned by altering the path 
   buffer, whose maximum capacity in bytes is specified by the size parameter.
   Note that it should have room for at least 80 characters (not bytes),
   so you should reserve 6x that amount, just to be sure. The list of files
   is filtered according to the file extensions in the ext parameter.
   Passing NULL includes all files; "PCX;BMP" includes only files with
   .PCX or .BMP extensions. If you wish to control files by their attributes,
   one of the fields in the extension list can begin with a slash, followed
   by a set of attribute characters. Any attributes written on their own,
   or with a + before them, indicate to include only files which have
   that attribute set. Any attributes with a '-' before them indicate to
   leave out any files with that attribute. The flag characters are
   'r' (read-only), 'h' (hidden), 's' (system), 'd' (directory),
   and 'a' (archive). For example, an extension string of "PCX;BMP;/+r-d"
   will display only PCX or BMP files that are read-only, and no
   directories. The file selector is stretched to the width and height
   specified in the w and h parameters, and to the size of the standard
   Allegro font. If either the width or height argument is set to zero, it is
   stretched to the corresponding screen dimension. This function returns
   zero if it was closed with the Cancel button or non-zero if it was OK'd.

int gfx_mode_select(int *card, int *w, int *h);
   Displays the Allegro graphics mode selection dialog, which allows the 
   user to select a screen mode and graphics card. Stores the selection in 
   the three variables, and returns zero if it was closed with the Cancel 
   button or non-zero if it was OK'd.

int gfx_mode_select_ex(int *card, int *w, int *h, int *color_depth);
   Extended version of the graphics mode selection dialog, which allows the 
   user to select the color depth as well as the resolution and hardware 
   driver. This version of the function reads the initial values from the 
   parameters when it activates, so you can specify the default values.

extern int (*gui_shadow_box_proc)(int msg, struct DIALOG *d, int c);
extern int (*gui_ctext_proc)(int msg, struct DIALOG *d, int c);
extern int (*gui_button_proc)(int msg, struct DIALOG *d, int c);
extern int (*gui_edit_proc)(int msg, struct DIALOG *d, int c);
extern int (*gui_list_proc)(int msg, struct DIALOG *d, int c);
extern int (*gui_text_list_proc)(int msg, struct DIALOG *d, int c);
   If set, these functions will be used by the standard Allegro dialogs.
   This allows you to customise the look and feel, much like gui_fg_color
   and gui_bg_color, but much more flexiblely.



=======================================
============ DOS specifics ============
=======================================

Drivers JOY_TYPE_*/DOS
   The DOS library supports the following type parameters for the 
   install_joystick() function:

   JOY_TYPE_AUTODETECT
      Attempts to autodetect your joystick hardware. It isn't possible to 
      reliably distinguish between all the possible input setups, so this 
      routine can only ever choose the standard joystick, Sidewider, GamePad 
      Pro, or GrIP drivers, but it will use information from the 
      configuration file if one is available (this can be created using the 
      setup utility or by calling the save_joystick_data() function), so you 
      can always use JOY_TYPE_AUTODETECT in your code and then select the 
      exact hardware type from the setup program.

   JOY_TYPE_NONE
      Dummy driver for machines without any joystick.

   JOY_TYPE_STANDARD
      A normal two button stick.

   JOY_TYPE_2PADS
      Dual joystick mode (two sticks, each with two buttons).

   JOY_TYPE_4BUTTON
      Enable the extra buttons on a 4-button joystick.

   JOY_TYPE_6BUTTON
      Enable the extra buttons on a 6-button joystick.

   JOY_TYPE_8BUTTON
      Enable the extra buttons on an 8-button joystick.

   JOY_TYPE_FSPRO
      CH Flightstick Pro or compatible stick, which provides four buttons, 
      an analogue throttle control, and a 4-direction coolie hat.

   JOY_TYPE_WINGEX
      A Logitech Wingman Extreme, which should also work with any 
      Thrustmaster Mk.I compatible joystick. It provides support for four 
      buttons and a coolie hat. This also works with the Wingman Warrior, if 
      you plug in the 15 pin plug (remember to unplug the 9-pin plug!) and 
      set the tiny switch in front to the "H" position (you will not be able 
      to use the throttle or the spinner though).

   JOY_TYPE_SIDEWINDER
      The Microsoft Sidewinder digital pad (supports up to four controllers, 
      each with ten buttons and a digital direction control).

   JOY_TYPE_SIDEWINDER_AG
      An alternative driver to JOY_TYPE_SIDEWINDER.
      Try this if your Sidewinder isn't recognized with JOY_TYPE_SIDEWINDER.

   JOY_TYPE_GAMEPAD_PRO
      The Gravis GamePad Pro (supports up to two controllers, each with ten 
      buttons and a digital direction control).

   JOY_TYPE_GRIP
      Gravis GrIP driver, using the grip.gll driver file.

   JOY_TYPE_GRIP4
      Version of the Gravis GrIP driver that is constrained to only move 
      along the four main axis.

   JOY_TYPE_SNESPAD_LPT1
   JOY_TYPE_SNESPAD_LPT2
   JOY_TYPE_SNESPAD_LPT3
      SNES joypads connected to LPT1, LPT2, and LPT3 respectively.

   JOY_TYPE_PSXPAD_LPT1
   JOY_TYPE_PSXPAD_LPT2
   JOY_TYPE_PSXPAD_LPT3
      PSX joypads connected to LPT1, LPT2, and LPT3 respectively. See 
      http://www.ziplabel.com/dpadpro/index.html for information 
      about the parallel cable required. The driver automagically detects 
      which types of PSX pads are connected out of digital, analog (red or 
      green mode), NegCon, multi taps, Namco light guns, Jogcons (force 
      feedback steering wheel) and the mouse. If the controller isn't 
      recognised it is treated as an analog controller, meaning the driver 
      should work with just about anything. You can connect controllers in 
      any way you see fit, but only the first 8 will be used.

      The Sony Dual Shock or Namco Jogcon will reset themselves (to digital 
      mode) after not being polled for 5 seconds. This is normal, the same 
      thing happens on a Playstation, it's designed to stop any vibration in 
      case the host machine crashes. Other mode switching controllers may 
      have similar quirks. However, if this happens to a Jogcon controller 
      the mode button is disabled. To reenable the mode button on the Jogcon 
      you need to hold down the Start and Select buttons at the same time.

      The G-con45 needs to be connected to (and pointed at) a TV type monitor
      connected to your computer. The composite video out on my video card 
      works fine for this (a Hercules Stingray 128/3D 8Mb). The TV video 
      modes in Mame should work too.

   JOY_TYPE_N64PAD_LPT1
   JOY_TYPE_N64PAD_LPT2
   JOY_TYPE_N64PAD_LPT3
      N64 joypads connected to LPT1, LPT2, and LPT3 respectively. See 
      http://www.st-hans.de/N64.htm for information about the 
      necessary hardware adaptor. It supports up to four controllers on a 
      single parallel port. There is no need to calibrate the analog stick, 
      as this is done by the controller itself when powered up. This means 
      that the stick has to be centred when the controller is initialised. 
      One possible issue people may have with this driver is that it is 
      physically impossible to move the analog stick fully diagonal, but I 
      can't see this causing any major problems. This is because of the 
      shape of the rim that the analog stick rests against. Like the Gravis 
      Game Pad Pro, this driver briefly needs to disable hardware interrupts 
      while polling. This causes a noticable performance hit on my machine 
      in both drivers, but there is no way around it. At a (very) rough 
      guess I'd say it slows down Mame 5% - 10%.

   JOY_TYPE_DB9_LPT1
   JOY_TYPE_DB9_LPT2
   JOY_TYPE_DB9_LPT3
      A pair of two-button joysticks connected to LPT1, LPT2, and LPT3 
      respectively. Port 1 is compatible with Linux joy-db9 driver 
      (multisystem 2-button), and port 2 is compatible with Atari interface 
      for DirectPad Pro. See the source file (src/dos/multijoy.c) for pinout 
      information.

   JOY_TYPE_TURBOGRAFIX_LPT1
   JOY_TYPE_TURBOGRAFIX_LPT2
   JOY_TYPE_TURBOGRAFIX_LPT3
      These drivers support up to 7 joysticks, each one with up to 5 
      buttons, connected to LPT1, LPT2, and LPT3 respectively. They use the 
      TurboGraFX interface by Steffen Schwenke: see 
      http://www.burg-halle.de/~schwenke/parport.html for details 
      on how to build this.

   JOY_TYPE_WINGWARRIOR
      A Wingman Warrior joystick.

   JOY_TYPE_IFSEGA_ISA
   JOY_TYPE_IFSEGA_PCI
   JOY_TYPE_IFSEGA_PCI_FAST
      Drivers for the IF-SEGA joystick interface cards by the IO-DATA 
      company (these come in PCI, PCI2, and ISA variants).

Drivers GFX_*/DOS
   The DOS library supports the following card parameters for the 
   set_gfx_mode() function:

   GFX_TEXT
      Return to text mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers. This will always fail
      under DOS.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution or color depth you get. See the
      set_gfx_mode() documentation for details.

   GFX_VGA
      The standard 256 color VGA mode 13h, using the GFX_VGA driver. This is 
      normally sized 320x200, which will work on any VGA but doesn't support 
      large virtual screens and hardware scrolling. Allegro also provides 
      some tweaked variants of the mode which are able to scroll, sized 
      320x100 (with a 200 pixel high virtual screen), 160x120 (with a 409 
      pixel high virtual screen), 256x256 (no scrolling), and 80x80 (with a 
      819 pixel high virtual screen).

   GFX_MODEX
      Mode-X will work on any VGA card, and provides a range of different 
      256 color tweaked resolutions.

      Stable mode-X resolutions:

         Square aspect ratio: 320x240

         Skewed aspect ratio: 256x224, 256x240, 320x200, 320x400,
                              320x480, 320x600, 360x200, 360x240,
                              360x360, 360x400, 360x480

         These have worked on every card/monitor that I've tested.

      Unstable mode-X resolutions:

         Square aspect ratio: 360x270, 376x282, 400x300

         Skewed aspect ratio: 256x200, 256x256, 320x350, 360x600,
                              376x308, 376x564, 400x150, 400x600

         These only work on some monitors. They were fine on my old machine, 
         but don't get on very well with my new monitor. If you are worried 
         about the possibility of damaging your monitor by using these 
         modes, don't be. Of course I'm not providing any warranty with any 
         of this, and if your hardware does blow up that is tough, but I 
         don't think this sort of tweaking can do any damage. From the 
         documentation of Robert Schmidt's TWEAK program:

            "Some time ago, putting illegal or unsupported values or 
            combinations of such into the video card registers might prove 
            hazardous to both your monitor and your health. I have *never* 
            claimed that bad things can't happen if you use TWEAK, although 
            I'm pretty sure it never will. I've never heard of any damage 
            arising from trying out TWEAK, or from general VGA tweaking in 
            any case."

      Most of the mode-X drawing functions are slower than in mode 13h, due 
      to the complexity of the planar bitmap organisation, but solid area 
      fills and plane-aligned blits from one part of video memory to another 
      can be significantly faster, particularly on older hardware. Mode-X 
      can address the full 256k of VGA RAM, so hardware scrolling and page 
      flipping are possible, and it is possible to split the screen in order 
      to scroll the top part of the display but have a static status 
      indicator at the bottom.

   GFX_VESA1
      Use the VESA 1.x driver.

   GFX_VESA2B
      Use the VBE 2.0 banked mode driver.

   GFX_VESA2L
      Use the VBE 2.0 linear framebuffer driver.

   GFX_VESA3
      Use the VBE 3.0 driver. This is the only VESA driver that supports the 
      request_refresh_rate() function.

      The standard VESA modes are 640x480, 800x600, and 1024x768. These 
      ought to work with any SVGA card: if they don't, get a copy of the 
      SciTech Display Doctor and see if that fixes it. What color depths are 
      available will depend on your hardware. Most cards support both 15 and 
      16 bit resolutions, but if at all possible I would advise you to 
      support both (it's not hard...) in case one is not available. Some 
      cards provide both 24 and 32 bit truecolor, in which case it is a 
      choice between 24 (saves memory) or 32 (faster), but many older cards 
      have no 32 bit mode and some newer ones don't support 24 bit 
      resolutions. Use the vesainfo test program to see what modes your VESA 
      driver provides.

      Many cards also support 640x400, 1280x1024, and 1600x1200, but these 
      aren't available on everything, for example the S3 chipset has no 
      640x400 mode. Other weird resolution may be possible, eg. some Tseng 
      boards can do 640x350, and the Avance Logic has a 512x512 mode.

      The SciTech Display Doctor provides several scrollable low resolution 
      modes in a range of different color depths (320x200, 320x240, 320x400, 
      320x480, 360x200, 360x240, 360x400, and 360x480 all work on my ET4000 
      with 8, 15, or 16 bits per pixel). These are lovely, allowing 
      scrolling and page flipping without the complexity of the mode-X 
      planar setup, but unfortunately they aren't standard so you will need 
      Display Doctor in order to use them.

   GFX_VBEAF
      VBE/AF is a superset of the VBE 2.0 standard, which provides an API 
      for accessing hardware accelerator features. VBE/AF drivers are 
      currently only available from the FreeBE/AF project or as part of the 
      SciTech Display Doctor package, but they can give dramatic speed 
      improvements when used with suitable hardware. For a detailed 
      discussion of hardware acceleration issues, refer to the documentation 
      for the gfx_capabilities flag.

      You can use the afinfo test program to check whether you have a VBE/AF 
      driver, and to see what resolutions it supports.

      The SciTech VBE/AF drivers require nearptr access to be enabled, so 
      any stray pointers are likely to crash your machine while their 
      drivers are in use. This means it may be a good idea to use VESA while 
      debugging your program, and only switch to VBE/AF once the code is 
      working correctly. The FreeBE/AF drivers do not have this problem.

   GFX_XTENDED
      An unchained 640x400 mode, as described by Mark Feldman in the PCGPE. 
      This uses VESA to select an SVGA mode (so it will only work on cards 
      supporting the VESA 640x400 resolution), and then unchains the VGA 
      hardware as for mode-X. This allows the entire screen to be addressed 
      without the need for bank switching, but hardware scrolling and page 
      flipping are not possible. This driver will never be autodetected (the 
      normal VESA 640x400 mode will be chosen instead), so if you want to 
      use it you will have to explicitly pass GFX_XTENDED to set_gfx_mode().

Drivers DIGI_*/DOS
   The DOS sound functions support the following digital soundcards:

      DIGI_AUTODETECT      - let Allegro pick a digital sound driver
      DIGI_NONE            - no digital sound
      DIGI_SB              - Sound Blaster (autodetect type)
      DIGI_SB10            - SB 1.0 (8 bit mono single shot dma)
      DIGI_SB15            - SB 1.5 (8 bit mono single shot dma)
      DIGI_SB20            - SB 2.0 (8 bit mono auto-initialised
                             dma)
      DIGI_SBPRO           - SB Pro (8 bit stereo)
      DIGI_SB16            - SB16 (16 bit stereo)
      DIGI_AUDIODRIVE      - ESS AudioDrive
      DIGI_SOUNDSCAPE      - Ensoniq Soundscape
      DIGI_WINSOUNDSYS     - Windows Sound System

Drivers MIDI_*/DOS
   The DOS sound functions support the following MIDI soundcards:

      MIDI_AUTODETECT      - let Allegro pick a MIDI sound driver
      MIDI_NONE            - no MIDI sound
      MIDI_ADLIB           - Adlib or SB FM synth (autodetect type)
      MIDI_OPL2            - OPL2 synth (mono, used in Adlib and SB)
      MIDI_2XOPL2          - dual OPL2 synths (stereo, used in
                             SB Pro-I)
      MIDI_OPL3            - OPL3 synth (stereo, SB Pro-II
                             and above)
      MIDI_SB_OUT          - SB MIDI interface
      MIDI_MPU             - MPU-401 MIDI interface
      MIDI_DIGMID          - sample-based software wavetable player
      MIDI_AWE32           - AWE32 (EMU8000 chip)

void split_modex_screen(int line);
   This function is only available in mode-X. It splits the VGA display into 
   two parts at the specified line. The top half of the screen can be 
   scrolled to any part of video memory with the scroll_screen() function, 
   but the part below the specified line number will remain fixed and will 
   display from position (0, 0) of the screen bitmap. After splitting the 
   screen you will generally want to scroll so that the top part of the 
   display starts lower down in video memory, and then create two 
   sub-bitmaps to access the two sections (see examples/exscroll.c for a 
   demonstration of how this could be done). To disable the split, call 
   split_modex_screen(0).

extern int i_love_bill;
   When running in clean DOS mode, the timer handler dynamically reprograms 
   the clock chip to generate interrupts at exactly the right times, which 
   gives an extremely high accuracy. Unfortunately, this constant speed 
   adjustment doesn't work under most multitasking systems (notably 
   Windows), so there is an alternative mode that just locks the hardware 
   timer interrupt to a speed of 200 ticks per second. This reduces the 
   accuracy of the timer (for instance, rest() will round the delay time to 
   the nearest 5 milliseconds), and prevents the vertical retrace simulator 
   from working, but on the plus side, it makes Allegro programs work under 
   Windows. This flag is set by allegro_init() if it detects the presence of
   a multitasking OS, and enables the fixed rate timer mode.



===========================================
============ Windows specifics ============
===========================================

A Windows program that uses the Allegro library is only required to include
one or more header files from the include/allegro tree, or allegro.h; however,
if it also needs to directly call non portable Win32 API functions, it must
include the Windows-specific header file winalleg.h after the Allegro headers,
and before any Win32 API header file. By default winalleg.h includes the main
Win32 C API header file windows.h. If instead you want to use the C++
interface to the Win32 API (a.k.a. the Microsoft Foundation Classes), define
the preprocessor symbol ALLEGRO_AND_MFC before including any Allegro header
so that afxwin.h will be included. Note that, in this latter case, the Allegro
debugging macros ASSERT() and TRACE() are renamed AL_ASSERT() and AL_TRACE()
respectively.

Windows GUI applications start with a WinMain() entry point, rather than the
standard main() entry point. Allegro is configured to build GUI applications
by default and to do some magic in order to make a regular main() work with
them, but you have to help it out a bit by writing END_OF_MAIN() right after
your main() function. If you don't want to do that, you can just include
winalleg.h and write a WinMain() function. Note that this magic may bring
about conflicts with a few programs using direct calls to Win32 API
functions; for these programs, the regular WinMain() is required and the
magic must be disabled by defining the preprocessor symbol
ALLEGRO_NO_MAGIC_MAIN before including Allegro headers.

If you want to build a console application using Allegro, you have to define
the preprocessor symbol USE_CONSOLE before including Allegro headers; it will
instruct the library to use console features and also to disable the special
processing of the main() function described above.

When creating the main window, Allegro searches the executable for an ICON 
resource named "allegro_icon". If it is present, Allegro automatically 
loads it and uses it as its application icon; otherwise, Allegro uses the 
default IDI_APPLICATION icon. See the manual of your compiler for a method 
to create an ICON resource, or use the wfixicon utility from the tools/win 
directory.

DirectX requires that system and video bitmaps (including the screen) be 
locked before you can draw onto them. This will be done automatically, but 
you can usually get much better performance by doing it yourself: see the 
acquire_bitmap() function for details.

Due to a major oversight in the design of DirectX, there is no way to 
preserve the contents of video memory when the user switches away from your 
program. You need to be prepared for the fact that your screen contents, and 
the contents of any video memory bitmaps, may be destroyed at any point. You 
can use the set_display_switch_callback() function to find out when this 
happens.

On the Windows platform, the only return values for the desktop_color_depth()
function are 8, 16, 24 and 32. This means that 15-bit and 16-bit desktops 
cannot be differentiated and are both reported as 16-bit desktops. See
below for the consequences for windowed and overlay DirectX drivers.

Drivers GFX_*/Windows
   The Windows library supports the following card parameters for the 
   set_gfx_mode() function:

   GFX_TEXT
      This closes any graphic mode previously opened with set_gfx_mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution or color depth you get. See the
      set_gfx_mode() documentation for details.

   GFX_DIRECTX
      Alias for GFX_DIRECTX_ACCEL.

   GFX_DIRECTX_ACCEL
      The regular fullscreen DirectX driver, running with hardware 
      acceleration enabled.

   GFX_DIRECTX_SOFT
      DirectX fullscreen driver that only uses software drawing, rather than 
      any hardware accelerated features.

   GFX_DIRECTX_SAFE
      Simplified fullscreen DirectX driver that doesn't support any hardware 
      acceleration, video or system bitmaps, etc.

   GFX_DIRECTX_WIN
      The regular windowed DirectX driver, running in color conversion mode 
      when the color depth doesn't match that of the Windows desktop. Color 
      conversion is much slower than direct drawing and is not supported 
      between 15-bit and 16-bit color depths. This limitation is needed to 
      work around that of desktop_color_depth() (see above) and allows to 
      select the direct drawing mode in a reliable way on desktops reported 
      as 16-bit:

   if (desktop_color_depth() == 16) {
      set_color_depth(16);
      if (set_gfx_mode(GFX_DIRECTX_WIN, 640, 480, 0, 0) != 0) {
         set_color_depth(15);
         if (set_gfx_mode(GFX_DIRECTX_WIN, 640, 480, 0, 0) != 0) {
            /* 640x480 direct drawing mode not supported */
            goto Error;
         }
      }
      /* ok, we are in direct drawing mode */
   }

      Note that, mainly for performance reasons, this driver requires the
      width of the screen to be a multiple of 4.

   GFX_DIRECTX_OVL
      The DirectX overlay driver. It uses special hardware features to run 
      your program in a windowed mode: it doesn't work on all hardware, but 
      performance is excellent on cards that are capable of it. It requires 
      the color depth to be the same as that of the Windows desktop. In light 
      of the limitation of desktop_color_depth() (see above), the reliable 
      way of setting the overlay driver on desktops reported as 16-bit is:

   if (desktop_color_depth() == 16) {
      set_color_depth(16);
      if (set_gfx_mode(GFX_DIRECTX_OVL, 640, 480, 0, 0) != 0) {
         set_color_depth(15);
         if (set_gfx_mode(GFX_DIRECTX_OVL, 640, 480, 0, 0) != 0) {
            /* 640x480 overlay driver not supported */
            goto Error;
         }
      }
      /* ok, the 640x480 overlay driver is running */
   }


   GFX_GDI
      The windowed GDI driver. It is extremely slow, but is guaranteed to 
      work on all hardware, so it can be useful for situations where you 
      want to run in a window and don't care about performance. Note that 
      this driver features a hardware mouse cursor emulation in order to 
      speed up basic mouse operations (like GUI operations).

Drivers DIGI_*/Windows
   The Windows sound functions support the following digital soundcards:

      DIGI_AUTODETECT      - let Allegro pick a digital sound driver
      DIGI_NONE            - no digital sound
      DIGI_DIRECTX(n)      - use DirectSound device #n (zero-based) with
                             direct mixing
      DIGI_DIRECTAMX(n)    - use DirectSound device #n (zero-based) with
                             Allegro mixing
      DIGI_WAVOUTID(n)     - high (n=0) or low (n=1) quality WaveOut device

Drivers MIDI_*/Windows
   The Windows sound functions support the following MIDI soundcards:

      MIDI_AUTODETECT      - let Allegro pick a MIDI sound driver
      MIDI_NONE            - no MIDI sound
      MIDI_WIN32MAPPER     - use win32 MIDI mapper
      MIDI_WIN32(n)        - use win32 device #n (zero-based)
      MIDI_DIGMID          - sample-based software wavetable player


The following functions provide a platform specific interface to seamlessly 
integrate Allegro into general purpose Win32 programs. To use these routines, 
you must include winalleg.h after other Allegro headers.

HWND win_get_window(void);
   Retrieves a handle to the window used by Allegro. Note that Allegro
   uses an underlying window even though you don't set any graphics mode,
   unless you have installed the neutral system driver (SYSTEM_NONE).

void win_set_window(HWND wnd);
   Registers an user-created window to be used by Allegro. This function is
   meant to be called before initialising the library with allegro_init()
   or installing the autodetected system driver (SYSTEM_AUTODETECT). It
   lets you attach Allegro to any already existing window and prevents the
   library from creating its own, thus leaving you total control over the
   window; in particular, you are responsible for processing the events as
   usual (Allegro will automatically monitor a few of them, but will not
   filter out any of them). You can then use every component of the library 
   (mouse, keyboard, sound, timers and so on) except the graphics subsystem,
   bearing in mind that some Allegro functions are blocking (e.g readkey()
   if the key buffer is empty) and thus must be carefully manipulated by the
   window thread.

   However you can also call it after the library has been initialised,
   provided that no graphics mode is set. In this case the keyboard, mouse,
   sound and sound recording modules will be restarted.

   Passing NULL instructs Allegro to switch back to its built-in window if
   an user-created window was registered, or to request a new handle from
   Windows for its built-in window if this was already in use.

void win_set_wnd_create_proc(HWND (*proc)(WNDPROC));
   Registers an user-defined procedure to be used by Allegro for creating
   its window. This function must be called *before* initializing the
   library with allegro_init() or installing the autodetected system
   driver (SYSTEM_AUTODETECT). It lets you customize Allegro's window but
   only by its creation: unlike with win_set_window(), you have no control
   over the window once it has been created (in particular, you are not
   responsible for processing the events). The registered function will be
   passed a window procedure (WNDPROC object) that it must make the 
   procedure of the new window of and it must return a handle to the new 
   window. You can then use the full-featured library in the regular way.

HDC win_get_dc(BITMAP *bmp);
   Retrieves a handle to the device context of a DirectX video or system
   bitmap.

void win_release_dc(BITMAP *bmp, HDC dc);
   Releases a handle to the device context of the bitmap that was
   previously retrieved with win_get_dc().


The following GDI routines are a very platform specific thing, to allow 
drawing Allegro memory bitmaps onto a Windows device context. When you want 
to use this, you'll have to install the neutral system driver (SYSTEM_NONE) 
or attach Allegro to an external window with win_set_window().

There are two ways to draw your Allegro bitmaps to the Windows GDI. When you 
are using static bitmaps (for example just some pictures loaded from a 
datafile), you can convert them to DDB (device-dependent bitmaps) with 
convert_bitmap_to_hbitmap() and then just use Win32's BitBlt() to draw it.

When you are using dynamic bitmaps (for example some things which react to 
user input), it's better to use set_palette_to_hdc() and blit_to_hdc() 
functions, which work with DIB (device-independent bitmaps).

There are also functions to blit from a device context into an Allegro 
BITMAP, so you can do things like screen capture.

All the drawing and conversion functions use the current palette as a color 
conversion table. You can alter the current palette with the 
set_palette_to_hdc() or select_palette() functions. Warning: when the GDI 
system color palette is explicitly changed, (by another application, for 
example) the current Allegro palette is not updated along with it!

To use these routines, you must include winalleg.h after Allegro headers.

void set_gdi_color_format(void);
   Tells Allegro to use the GDI color layout for truecolor images. This is 
   optional, but it will make the conversions work faster. If you are going 
   to call this, you should do it right after initialising Allegro and 
   before creating any graphics.

void set_palette_to_hdc(HDC dc, PALETTE pal);
   Selects and realizes an Allegro palette on the specified device context.

HPALETTE convert_palette_to_hpalette(PALETTE pal);
   Converts an Allegro palette to a Windows palette and returns a handle to 
   it. You should call DeleteObject() when you no longer need it.

void convert_hpalette_to_palette(HPALETTE hpal, PALETTE pal);
   Converts a Windows palette to an Allegro palette.

HBITMAP convert_bitmap_to_hbitmap(BITMAP *bitmap);
   Converts an Allegro memory bitmap to a Windows DDB and returns a handle 
   to it. This bitmap uses its own memory, so you can destroy the original 
   bitmap without affecting the converted one. You should call 
   DeleteObject() when you no longer need this bitmap.

BITMAP *convert_hbitmap_to_bitmap(HBITMAP bitmap);
   Creates an Allegro memory bitmap from a Windows DDB.

void draw_to_hdc(HDC dc, BITMAP *bitmap, int x, int y);
   Draws an entire Allegro bitmap to a Windows device context, using the 
   same parameters as the draw_sprite() function.

void blit_to_hdc(BITMAP *bitmap, HDC dc, int sx, sy, dx, dy, w, h);
   Blits an Allegro memory bitmap to a Windows device context, using the 
   same parameters as the blit() function.

void stretch_blit_to_hdc(BITMAP *bitmap, HDC dc, int sx, sy, sw, sh,
                                                 int dx, dy, dw, dh);
   Blits an Allegro memory bitmap to a Windows device context, using the 
   same parameters as the stretch_blit() function.

void blit_from_hdc(HDC hdc, BITMAP *bitmap, int sx, sy, dx, dy, w, h);
   Blits from a Windows device context to an Allegro memory bitmap, using 
   the same parameters as the blit() function. See stretch_blit_from_hdc() 
   for details.

void stretch_blit_from_hdc(HDC hcd, BITMAP *bitmap, int sx, sy, sw, sh,
                                                    int dx, dy, dw, dh);
   Blits from a Windows device context to an Allegro memory bitmap, using 
   the same parameters as the stretch_blit() function. It uses the current 
   Allegro palette and does conversion to this palette, regardless of the 
   current DC palette. So if you are blitting from 8 bit mode, you should 
   first set the DC palette with the set_palette_to_hdc() function.



========================================
============ Unix specifics ============
========================================

In order to locate things like the config and translation files, Allegro 
needs to know the path to your executable. Since there is no standard way to 
find that, it needs to capture a copy of your argv[] parameter, and it does 
this with some preprocessor macro trickery. Unfortunately it can't quite 
pull this off without a little bit of your help, so you will have to write 
END_OF_MAIN() right after your main() function. Pretty easy, really, and if 
you forget, you'll get a nice linker error about a missing _mangled_main 
function to remind you :-)

Drivers GFX_*/Linux
   When running in Linux console mode, Allegro supports the following card 
   parameters for the set_gfx_mode() function:

   GFX_TEXT
      Return to text mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers. This will always fail
      under Linux console mode.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution or color depth you get. See the
      set_gfx_mode() documentation for details.

   GFX_FBCON
      Use the framebuffer device (eg. /dev/fb0). This requires you to have 
      framebuffer support compiled into your kernel, and correctly 
      configured for your hardware. It is currently the only console mode 
      driver that will work without root permissions, unless you are using
      a development version of SVGAlib.

   GFX_VBEAF
      Use a VBE/AF driver (vbeaf.drv), assuming that you have installed one 
      which works under Linux (currently only two of the FreeBE/AF project 
      drivers are capable of this: I don't know about the SciTech ones). 
      VBE/AF requires root permissions, but is currently the only Linux 
      driver which supports hardware accelerated graphics.

   GFX_SVGALIB
      Use the SVGAlib library for graphics output. This requires root
      permissions if your version of SVGAlib requires them.

   GFX_VGA
   GFX_MODEX
      Use direct hardware access to set standard VGA or mode-X resolutions, 
      supporting the same modes as in the DOS versions of these drivers. 
      Requires root permissions.

Drivers GFX_*/X
   When running in X mode, Allegro supports the following card parameters 
   for the set_gfx_mode() function:

   GFX_TEXT
      This closes any graphic mode previously opened with set_gfx_mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution or color depth you get. See the
      set_gfx_mode() documentation for details.

   GFX_XWINDOWS
      The standard X graphics driver. This should work on any Unix system, 
      and can operate remotely. It does not require root permissions.

   GFX_XWINDOWS_FULLSCREEN
      The same as above, but while GFX_XWINDOWS runs windowed, this one uses
      the XF86VidMode extension to make it run in fullscreen mode even
      without root permissions. You're still using the standard X protocol
      though, so expect the same low performances as with the windowed
      driver version.

   GFX_XDGA
      Use the XFree86 DGA 1.0 extension to write directly to the screen
      surface. DGA is normally much faster than the standard X mode, but does
      not produce such well behaved windowed programs, and will not work
      remotely. This driver requires root permissions.

   GFX_XDGA_FULLSCREEN
      Like GFX_XDGA, but also changes the screen resolution so that it will 
      run fullscreen. This driver requires root permissions.

   GFX_XDGA2
      Use new DGA 2.0 extension provided by XFree86 4.0.x. This will work
      in fullscreen mode, and it will support hardware acceleration if
      available. This driver requires root permissions.

   GFX_XDGA2_SOFT
      The same as GFX_XDGA2, but turns off hardware acceleration support.
      This driver requires root permissions.

Drivers DIGI_*/Unix
   The Unix sound functions support the following digital soundcards:

      DIGI_AUTODETECT      - let Allegro pick a digital sound driver
      DIGI_NONE            - no digital sound
      DIGI_OSS             - Open Sound System
      DIGI_ESD             - Enlightened Sound Daemon
      DIGI_ARTS            - aRts (Analog Real-Time Synthesizer)
      DIGI_ALSA            - ALSA sound driver

Drivers MIDI_*/Unix
   The Unix sound functions support the following MIDI soundcards:

      MIDI_AUTODETECT      - let Allegro pick a MIDI sound driver
      MIDI_NONE            - no MIDI sound
      MIDI_OSS             - Open Sound System
      MIDI_DIGMID          - sample-based software wavetable player
      MIDI_ALSA            - ALSA RawMIDI driver




========================================
============ BeOS specifics ============
========================================

Drivers GFX_*/BeOS
   BeOS Allegro supports the following card parameters for the
   set_gfx_mode() function:

   GFX_TEXT
      This closes any graphic mode previously opened with set_gfx_mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution. See the set_gfx_mode() 
      documentation for details.

   GFX_BEOS_FULLSCREEN
      Fullscreen exclusive mode. Supports only resolutions higher or equal
      to 640x480, and uses hardware acceleration if available.

   GFX_BEOS_FULLSCREEN_SAFE
      Works the same as GFX_FULLSCREEN, but disables acceleration.

   GFX_BEOS_WINDOWED
      Fast windowed mode using the BDirectWindow class. Not all graphics
      cards support this.

Drivers DIGI_*/BeOS
   The BeOS sound functions support the following digital soundcards:

      DIGI_AUTODETECT      - let Allegro pick a digital sound driver
      DIGI_NONE            - no digital sound
      DIGI_BEOS            - BeOS digital output

Drivers MIDI_*/BeOS
   The BeOS sound functions support the following MIDI soundcards:

      MIDI_AUTODETECT      - let Allegro pick a MIDI sound driver
      MIDI_NONE            - no MIDI sound
      MIDI_BEOS            - BeOS MIDI output
      MIDI_DIGMID          - sample-based software wavetable player




=======================================
============ QNX specifics ============
=======================================

Drivers GFX_*/QNX
   QNX Allegro supports the following card parameters for the
   set_gfx_mode() function:

   GFX_TEXT
      This closes any graphic mode previously opened with set_gfx_mode.

   GFX_AUTODETECT
      Let Allegro pick an appropriate graphics driver.

   GFX_AUTODETECT_FULLSCREEN
      Autodetects a graphics driver, but will only use fullscreen drivers,
      failing if these are not available on current platform.

   GFX_AUTODETECT_WINDOWED
      Same as above, but uses only windowed drivers.

   GFX_SAFE
      Special driver for when you want to reliably set a graphics mode and 
      don't really care what resolution. See the set_gfx_mode() 
      documentation for details.

   GFX_PHOTON_DIRECT
      Fullscreen exclusive mode through Photon.

   GFX_PHOTON
      Windowed mode in a Photon window. Note that, mainly for performance
      reasons, this driver requires the width of the screen to be a multiple
      of 4.

Drivers DIGI_*/QNX
   The QNX sound functions support the following digital soundcards:

      DIGI_AUTODETECT      - let Allegro pick a digital sound driver
      DIGI_NONE            - no digital sound
      DIGI_ALSA            - ALSA sound driver

Drivers MIDI_*/QNX
   The QNX sound functions support the following MIDI soundcards:

      MIDI_AUTODETECT      - let Allegro pick a MIDI sound driver
      MIDI_NONE            - no MIDI sound
      MIDI_ALSA            - ALSA RawMIDI driver
      MIDI_DIGMID          - sample-based software wavetable player


The following functions provide a platform specific interface to seamlessly 
integrate Allegro into general purpose QNX programs. To use these routines, 
you must include qnxalleg.h after other Allegro headers.

PtWidget_t qnx_get_window(void);
   Retrieves a handle to the window used by Allegro. Note that Allegro
   uses an underlying window even though you don't set any graphics mode,
   unless you have installed the neutral system driver (SYSTEM_NONE).



=======================================================
============ Differences between platforms ============
=======================================================

Here's a quick summary of things that may cause problems when moving your 
code from one platform to another (you can find a more detailed version of 
this in the docs section of the Allegro website).

The Windows and Unix versions require you to write END_OF_MAIN() after your 
main() function, which is used to magically turn an ANSI C style main() 
into a Windows style WinMain(), and so that the Unix code can grab a copy 
of your argv[] parameter.

On many platforms Allegro runs very slowly if you rely on it in order to 
automatically lock bitmaps when drawing onto them. For good performance, 
you need to call acquire_bitmap() and release_bitmap() yourself, and try 
to keep the amount of locking to a minimum.

The Windows version may lose the contents of video memory if the user 
switches away from your program, so you need to deal with that.

None of the currently supported platforms require input polling, but it is 
possible that some future ones might, so if you want to ensure 100% 
portability of your program, you should call poll_mouse() and 
poll_keyboard() in all the relevant places.

Allegro defines a number of standard macros that can be used to check 
various attributes of the current platform:

ALLEGRO_PLATFORM_STR
   Text string containing the name of the current platform.

ALLEGRO_DOS
ALLEGRO_DJGPP
ALLEGRO_WATCOM
ALLEGRO_WINDOWS
ALLEGRO_MSVC
ALLEGRO_MINGW32
ALLEGRO_BCC32
ALLEGRO_UNIX
ALLEGRO_LINUX
ALLEGRO_BEOS
ALLEGRO_QNX
ALLEGRO_GCC
   Defined if you are building for a relevant system. Often several of these 
   will apply, eg. DOS+Watcom, or Windows+GCC+MinGW32.

ALLEGRO_I386
ALLEGRO_BIG_ENDIAN
ALLEGRO_LITTLE_ENDIAN
   Defined if you are building for a processor of the relevant type.

ALLEGRO_VRAM_SINGLE_SURFACE
   Defined if the screen is a single large surface that is then partitioned 
   into multiple video sub-bitmaps (eg. DOS), rather than each video bitmap 
   being a totally unique entity (eg. Windows).

ALLEGRO_CONSOLE_OK
   Defined if when you are not in a graphics mode, there is a text mode 
   console that you can printf() to, and from which the user could 
   potentially redirect stdout to capture it even while you are in a 
   graphics mode. If this define is absent, you are running in an 
   environment like Windows that has no stdout at all.

ALLEGRO_MAGIC_MAIN
   Defined if Allegro uses a magic main, i.e takes over the main() entry 
   point and turns it into a secondary entry point suited to its needs.

ALLEGRO_LFN
   Non-zero if long filenames are supported, or zero if you are limited to 
   8.3 format (in the djgpp version, this is a variable depending on the 
   runtime environment).

LONG_LONG
   Defined to whatever represents a 64-bit "long long" integer for the 
   current compiler, or not defined if that isn't supported.

OTHER_PATH_SEPARATOR
   Defined to a path separator character other than a forward slash for 
   platforms that use one (eg. a backslash under DOS and Windows), or 
   defined to a forward slash if there is no other separator character.

DEVICE_SEPARATOR
   Defined to the filename device separator character (a colon for DOS and 
   Windows), or to zero if there are no explicit devices in paths (Unix).

Allegro also defines a number of standard macros that can be used to 
insulate you from some of the differences between systems:

USE_CONSOLE
   If you define this prior to including Allegro headers, Allegro will be 
   set up for building a console application rather than the default GUI 
   program on some platforms (especially Windows).

INLINE
   Use this in place of the regular "inline" function modifier keyword, and 
   your code will work correctly on any of the supported compilers.

ZERO_SIZE_ARRAY(type, name)
   Use this to declare zero-sized arrays in terminal position inside 
   structures, like in the BITMAP structure. These arrays are effectively 
   equivalent to the flexible array members of ISO C99.



=======================================================
============ Reducing your executable size ============
=======================================================

Some people complain that Allegro produces very large executables. This is 
certainly true: with the djgpp version, a simple "hello world" program will 
be about 200k, although the per-executable overhead is much less for 
platforms that support dynamic linking. But don't worry, Allegro takes up a 
relatively fixed amount of space, and won't increase as your program gets 
larger. As George Foot so succinctly put it, anyone who is concerned about 
the ratio between library and program code should just get to work and write 
more program code to catch up :-)

Having said that, there are several things you can do to make your programs 
smaller:

For all platforms, you can use an executable compressor called UPX, which
is available at http://upx.tsx.org/ . This usually manages a compression
ratio of about 40%.

When using djgpp: for starters, read the djgpp FAQ section 8.14, and take
note of the -s switch. And don't forget to compile your program with
optimisation enabled!

If a DOS program is only going to run in a limited number of graphics modes, 
you can specify which graphics drivers you would like to include with the 
code:

   BEGIN_GFX_DRIVER_LIST
      driver1
      driver2
      etc...
   END_GFX_DRIVER_LIST

where the driver names are any of the defines:

   GFX_DRIVER_VBEAF
   GFX_DRIVER_VGA
   GFX_DRIVER_MODEX
   GFX_DRIVER_VESA3
   GFX_DRIVER_VESA2L
   GFX_DRIVER_VESA2B
   GFX_DRIVER_XTENDED
   GFX_DRIVER_VESA1

This construct must be included in only one of your C source files. The 
ordering of the names is important, because the autodetection routine works 
down from the top of the list until it finds the first driver that is able 
to support the requested mode. I suggest you stick to the default ordering 
given above, and simply delete whatever entries you aren't going to use.

If your DOS program doesn't need to use all the possible color depths, you 
can specify which pixel formats you want to support with the code:

   BEGIN_COLOR_DEPTH_LIST
      depth1
      depth2
      etc...
   END_COLOR_DEPTH_LIST

where the color depth names are any of the defines:

   COLOR_DEPTH_8
   COLOR_DEPTH_15
   COLOR_DEPTH_16
   COLOR_DEPTH_24
   COLOR_DEPTH_32

Removing any of the color depths will save quite a bit of space, with the 
exception of the 15 and 16 bit modes: these share a great deal of code, so 
if you are including one of them, there is no reason not to use both. Be 
warned that if you try to use a color depth which isn't in this list, your 
program will crash horribly!

In the same way as the above, you can specify which DOS sound drivers you 
want to support with the code:

   BEGIN_DIGI_DRIVER_LIST
      driver1
      driver2
      etc...
   END_DIGI_DRIVER_LIST

using the digital sound driver defines:

   DIGI_DRIVER_SOUNDSCAPE
   DIGI_DRIVER_AUDIODRIVE
   DIGI_DRIVER_WINSOUNDSYS
   DIGI_DRIVER_SB

and for the MIDI music:

   BEGIN_MIDI_DRIVER_LIST
      driver1
      driver2
      etc...
   END_MIDI_DRIVER_LIST

using the MIDI driver defines:

   MIDI_DRIVER_AWE32
   MIDI_DRIVER_DIGMID
   MIDI_DRIVER_ADLIB
   MIDI_DRIVER_MPU
   MIDI_DRIVER_SB_OUT

If you are going to use either of these sound driver constructs, you must 
include both.

Likewise for the DOS joystick drivers, you can declare an inclusion list:

   BEGIN_JOYSTICK_DRIVER_LIST
      driver1
      driver2
      etc...
   END_JOYSTICK_DRIVER_LIST

using the joystick driver defines:

   JOYSTICK_DRIVER_WINGWARRIOR
   JOYSTICK_DRIVER_SIDEWINDER
   JOYSTICK_DRIVER_GAMEPAD_PRO
   JOYSTICK_DRIVER_GRIP
   JOYSTICK_DRIVER_STANDARD
   JOYSTICK_DRIVER_SNESPAD
   JOYSTICK_DRIVER_PSXPAD
   JOYSTICK_DRIVER_N64PAD
   JOYSTICK_DRIVER_DB9
   JOYSTICK_DRIVER_TURBOGRAFX
   JOYSTICK_DRIVER_IFSEGA_ISA
   JOYSTICK_DRIVER_IFSEGA_PCI
   JOYSTICK_DRIVER_IFSEGA_PCI_FAST

The standard driver includes support for the dual joysticks, increased 
numbers of buttons, Flightstick Pro, and Wingman Extreme, because these are 
all quite minor variations on the basic code.

If you are _really_ serious about this size, thing, have a look at the top 
of include/allegro/alconfig.h and you will see the lines:

   #define ALLEGRO_COLOR8
   #define ALLEGRO_COLOR16
   #define ALLEGRO_COLOR24
   #define ALLEGRO_COLOR32

If you comment out any of these definitions and then rebuild the library, 
you will get a version without any support for the absent color depths, 
which will be even smaller than using the DECLARE_COLOR_DEPTH_LIST() macro. 
Removing the ALLEGRO_COLOR16 define will get rid of the support for both 15 
and 16 bit hicolor modes, since these share a lot of the same code.

Note: the aforementioned methods for removing unused hardware drivers only 
apply to statically linked versions of the library, eg. DOS. On Windows and 
Unix platforms, you can build Allegro as a DLL or shared library, which 
prevents these methods from working, but saves so much space that you 
probably won't care about that. Removing unused color depths from alconfig.h 
will work on any platform, though.

If you are distributing a copy of the setup program along with your game, 
you may be able to get a dramatic size reduction by merging the setup code 
into your main program, so that only one copy of the Allegro routines will 
need to be linked. See setup.txt for details. In the djgpp version, after 
compressing the executable, this will probably save you about 200k compared 
to having two separate programs for the setup and the game itself.



===================================
============ Debugging ============
===================================

There are three versions of the Allegro library: the normal optimised code, 
one with extra debugging support, and a profiling version. See the platform 
specific readme files for information about how to install and link with 
these alternative libs. Although you will obviously want to use the 
optimised library for the final version of your program, it can be very 
useful to link with the debug lib while you are working on it, because this 
will make debugging much easier, and includes assert tests that will help to 
locate errors in your code at an earlier stage. Allegro also contains some 
debugging helper functions:

void al_assert(const char *file, int line);
   Raises an assert for an error at the specified file and line number. The 
   file parameter is always given in ASCII format. If you have installed a 
   custom assert handler it uses that, or if the environment variable 
   ALLEGRO_ASSERT is set it writes a message into the file specified by the 
   environment, otherwise it aborts the program with an error message. You 
   will usually want to use the ASSERT() macro instead of calling this 
   function directly.

void al_trace(const char *msg, ...);
   Outputs a debugging trace message, using a printf() format string given 
   in ASCII. If you have installed a custom trace handler it uses that, or 
   if the environment variable ALLEGRO_TRACE is set it writes into the file 
   specified by the environment, otherwise it writes the message to 
   "allegro.log" in the current directory. You will usually want to use the 
   TRACE() macro instead of calling this function directly.

void ASSERT(condition);
   Debugging helper macro. Normally compiles away to nothing, but if you 
   defined the preprocessor symbol DEBUGMODE before including Allegro headers,
   it will check the supplied condition and call al_assert() if it fails.

void TRACE(char *msg, ...);
   Debugging helper macro. Normally compiles away to nothing, but if you 
   defined the preprocessor symbol DEBUGMODE before including Allegro headers,
   it passes the supplied message given in ASCII format to al_trace().

void register_assert_handler(int (*handler)(const char *msg));
   Supplies a custom handler function for dealing with assert failures. Your 
   callback will be passed a formatted error message in ASCII, and should 
   return non-zero if it has processed the error, or zero to continue with 
   the default actions. You could use this to ignore assert failures, or to 
   display the error messages on a graphics mode screen without aborting the 
   program.

void register_trace_handler(int (*handler)(const char *msg));
   Supplies a custom handler function for dealing with trace output. Your 
   callback will be passed a formatted error message in ASCII, and should 
   return non-zero if it has processed the message, or zero to continue with 
   the default actions. You could use this to ignore trace output, or to 
   display the messages on a second monochrome monitor, etc.



==========================================
============ Makefile targets ============
==========================================

There are a number of options that you can use to control exactly how 
Allegro will be compiled. On Unix platforms you do this by passing arguments 
to the configure script (run "configure --help" for a list), but on other 
platforms you can set the environment variables:

   DEBUGMODE=1
   Selects a debug build, rather than the normal optimised version.

   DEBUGMODE=2
   Selects a build intended to debug Allegro itself, rather than the
   normal optimised version.

   PROFILEMODE=1
   Selects a profiling build, rather than the normal optimised version.

   WARNMODE=1
   Selects strict compiler warnings. If you are planning to work on Allegro 
   yourself, rather than just using it in your programs, you should be sure 
   to have this mode enabled.

   STATICLINK=1 (MSVC and Mingw32 only)
   Link as a static library, rather than the default DLL.

   TARGET_ARCH_COMPAT=[cpu] (implemented for most GNU platforms)
   This option will optimize for the given processor while maintaining
   compatibility with older processors.
   Example: set TARGET_ARCH_COMPAT=i586

   TARGET_ARCH_EXCL=[cpu] (implemented for most GNU platforms)
   This option will optimize for the given processor. Please note that
   using it will cause the code to *NOT* run on older processors.
   Example: set TARGET_ARCH_EXCL=i586

   TARGET_OPTS=[opts] (implemented for most GNU platforms)
   This option allows you to customize general compiler optimisations.

   CROSSCOMPILE=1 (djgpp only)
   Allows you to build the djgpp library under Linux, using djgpp as a 
   cross-compiler.

   ALLEGRO_USE_C=1 (djgpp only)
   Allows you to build the djgpp library using C drawing code instead of the 
   usual asm routines. This is only really useful for testing, since the asm 
   version is faster.

If you only want to recompile a specific test program or utility, you can 
specify it as an argument to make, eg. "make demo" or "make grabber". The 
makefiles also provide some special pseudo-targets:

   'default'
   The normal build process. Compiles the current library version (one of 
   optimised, debugging, or profiling, selected by the above environment 
   variables), builds the test and example programs, and converts the 
   documentation files.

   'all'
   Compiles all three library versions (optimised, debugging, and 
   profiling), builds the test and example programs, and converts the 
   documentation files.

   'lib'
   Compiles the current library version (one of optimised, debugging, or 
   profiling, selected by the above environment variables).

   'install'
   Copies the current library version (one of optimised, debugging, or 
   profiling, selected by the above environment variables), into your 
   compiler lib directory, recompiling it as required, and installs the 
   Allegro headers.

   'installall'
   Copies all three library versions (optimised, debugging, and profiling), 
   into your compiler lib directory, recompiling them as required, and 
   installs the Allegro headers.

   'uninstall'
   Removes the Allegro library and headers from your compiler directories. 

   'docs'
   Converts the documentation files from the ._tx sources.

   'docs-dvi' (Unix only)
   Creates the allegro.dvi device independent documentation file. This is
   not a default target, since you need the texi2dvi tool to create it. The
   generated file is especially prepared to be printed on paper.

  'docs-ps' or 'docs-gzipped-ps' (Unix only)
   Creates a Postcript file from the previously generated DVI file. This is
   not a default target, since you need the texi2dvi and dvips tools to
   create it. The second target compresses the generated Postscript file.
   The generated file is especially prepared to be printed on paper.

   'install-man' or 'install-gzipped-man' (Unix only)
   This generates Unix man pages for each Allegro function or variable and
   installs them. The second target compresses the manual pages after
   installing them.

   'install-info' or 'install-gzipped-info' (Unix only)
   Converts the documentation to Info format and installs it. The second
   target compresses the info file after installing it.

   'clean'
   Removes generated object and library files, either to recover disk space 
   or to force a complete rebuild the next time you run make. This target is 
   designed so that if you run a "make install" followed by "make clean", 
   you will still have a functional version of Allegro.

   'distclean'
   Like "make clean", but more so. This removes all the executable files and 
   the documentation, leaving you with only the same files that are included
   when you unzip a new Allegro distribution.

   'veryclean'
   Use with extreme caution! This target deletes absolutely all generated 
   files, including some that may be non-trivial to recreate. After you run 
   "make veryclean", a simple rebuild will not work: at the very least you 
   will have to run "make depend", and perhaps also fixdll.bat if you are 
   using the Windows library. These targets make use of non-standard tools 
   like SED, so unless you know what you are doing and have all this stuff 
   installed, you should not use them.

   'depend'
   Regenerates the dependency files (obj/*/makefile.dep). You need to run 
   this after "make veryclean", or whenever you add new headers to the 
   Allegro sources.

   'compress' (djgpp, Mingw32 and MSVC only)
   Uses the DJP or UPX executable compressors (whichever you have installed) 
   to compress the example executables and utility programs, which can 
   recover a significant amount of disk space.



====================================
============ Conclusion ============
====================================

All good things must come to an end. Writing documentation is not a good 
thing, though, and that means it goes on for ever. There is always something 
I've forgotten to explain, or some essential detail I've left out, but for 
now you will have to make do with this. Feel free to ask if you can't figure 
something out.

Enjoy. I hope you find some of this stuff useful.


By Shawn Hargreaves.

http://alleg.sourceforge.net/




