Hello Steph,
Sunday, October 26, 2003, 6:11:39 AM, you wrote:
> I searched on ssize_t.
> There are 5 files where ssize_t is defined:
> configure.in //irrelevant to win32
> config.w32.h //changed definition to #define ssize_t SSIZE_T because
> <windows.h> is more pervasive in the upper reaches of PHP than
> <stddef.h>
> fcgi_config_win32.h //removed the define, no longer needed
> fcgi_config_x86.h //irrelevant to win32
> php_mnogo.c //irrelevant to win32
> There are 3 files where <stddef.h> is included:
> snprintf.c //removed the include line, no longer needed
> spprintf.c //needs stddef.h, uses ptrdiff_t
This is not correct, both use ptrdiff_t.
Also you need the sizeof define (which works as a detection define, too):
SIZEOF_PTRDIFF_T
On current windows platform this can be assumed to be 4. But when using 64
bit targets or PAE or such then ths may differ. Since we don't care about
those anywhere that shouldn't be a problem in itself.
> cast.c //removed the include line, no longer needed
> There is one file (Sterling's, naturally) where win32 isn't a
> consideration:
> dio.c //irrelevant to win32
> and one file where win32 is irrelevant given the context
> (because PHP_PREAD_64 is only defined in acinclude.m4)
> php.h //irrelevant to win32
> and then there's our friend the gd-breaker:
> gd_jpeg.c //removed the ssize_t define, no longer needed
> Now we don't HAVE to make those changes, but if we do, it's better
> future-proofing for that particular typedef because it provides a more
> generic solution than the one currently in place.
Generally your analyse looks good besides the comment above.
Only i am not sure if we can really nuke the include, you sure about that?
However changing the define to the one you suggest might be a good idea.
--
Best regards,
Marcus mailto:helly@php.net
