Pierre Joye [Wed, 4 Dec 2002 02:37:21 +0000 (02:37 +0000)]
Fix imagegd crashes when used with truecolor image (from jpeg,png, or imagecreatetruecolor)
may we add parameters to imagegd to let user specify the palette size and dither (set to 256 and true) ?
Fixed a bug that caused a crash in pdf_open_memory_image(), when a truecolor
image was loaded, it now checks if the image is truecolor and performs the
appropriate calculations (fixed bug #20785).
Changed E_WARNING back to E_ERROR.
Harald Radi [Tue, 3 Dec 2002 16:26:57 +0000 (16:26 +0000)]
damn milter api:
"Furthermore, since there is not a one-to-one correspondence between
threads and connections (N connections mapped onto M threads,
M <= N), connection-specific data must be accessed through the handles
provided by the Milter library."
which in other words means php_request_startup() and the corresponding
php_request_shutdown() will never be called for the same thread
Fixed a crash that occurs during the last step of png/jpeg -> gd2 -> png/jpeg
conversion. The same crash still occures with png/jpeg -> gd -> png/jpeg, because
apparently gd format cannot handle truecolor images.
Turned off debug messages inside gd_gd2.c.
Allocating enough memory to hold values.
Fix crash when certan stored procedures was called. This caused the free_result function to free memory not yet allocated.
Fixed flex version check. Some lex scanners like Sun's SGU do not appear
to support -V flag, so by passing them data on stdin we prevent them from
sitting idle waiting for stdin data. I've also added some common version
retrieval flags, that may be supported by various lex scanners.
Repositioned CoInitialize and CoUninitialize that apparetly makes it more
correct according to MS docs. Patch (+5) by Michael Sisolak
<msisolak@yahoo.com>.
Jason Greene [Mon, 2 Dec 2002 05:45:13 +0000 (05:45 +0000)]
Revert patch that should have been reverted a long time ago.
There was a memory leak in the error handling system on win32, that this patch
circumvented (by preventing the errors (EAGAIN mesages) from being generated).
# I must have forgotten to remove this when I fixed the leak
Sterling Hughes [Sun, 1 Dec 2002 22:15:39 +0000 (22:15 +0000)]
TODO: Cleanup elements that have happened, or most likely won't happen. There
is probably more that should be gone, but its still a move in the right
direction.
Shane Caraveo [Sun, 1 Dec 2002 21:37:14 +0000 (21:37 +0000)]
Fix CGI to match cgi spec.
This patch properly fixes support for CGI in PHP. For backwards compatible
broken behaviour, cgi.fix_pathinfo can be set to zero in php.ini.
CGI failed to work under apache at all, either using the cgi-script directive
or as a ScriptAlias setup. Typicaly it would try to parse itself. This will
still happen if you dissable fix_pathinfo, and set DISCARD_PATH.
This also fixes PATH_INFO, and finally we can run pres2 under cgi or fastcgi.
This patch has been tested under Apache 1.3, 2.0, IIS, as both cgi and fastcgi,
on Windows and OSX. A followup patch with build stuff for linux will follow.
Shane Caraveo [Sun, 1 Dec 2002 21:28:27 +0000 (21:28 +0000)]
php_error_docref aborts cgi. This broke using php as cgi under
apache/mod_cgi with the cgi-script directive and bang lines in php scripts.
removing it allows this to work again.
Harald Radi [Sun, 1 Dec 2002 19:50:07 +0000 (19:50 +0000)]
if the api functions are called, check if they're valid in that context
reopen file per request, so milter doesn't hang. PHPA or APC can be used to speed the milter up
added a milter_init callback to init the milter before it is registered
added protos to make sniper happy
Sascha Schumann [Sun, 1 Dec 2002 03:28:21 +0000 (03:28 +0000)]
add a "force HTTP/1.0 response" facility to the SAPI layer
this is necessary, when you want to take over control of a connection
and the web server is doing stupid things by default (like enabling
chunked transfer encoding for no reason).
projects / php / log