Re: Re: [ANNOUNCEMENT] pecl_http-2.1.0RC2 (beta) Released.

Pierre Joye in php.pecl.dev (Tue, 12 Aug 2014 13:06:49 +0200): >>>See this patch by Anatol (4 days old) >>>http://git.php.net/?p=php-src.git;a=commitdiff;h=7189039d650de1417e4e580865b1e77b39e3da55 > >is it not what Anatol fixed in core recently? I applied Anatol's fix to <win32/php_stdint.h> and to a number of extensions that had their own copy of it (qb, av, ffmpeg). So this fix was not enough. Anatol's fix seems to aim this loading order: 1. <stdint.h> 2. <win32/php_stdint.h> The fatal errors for php_http.dll were caused by loading the 2 headers the other way around! >Also using the php_stdint.h header should solve that automatically. Any call to <stdint.h> of VC11 after calling <win32/php_stdint.h> leads to fatal errors in PHP 5.5 now. After a long debugging session I tracked it down to FOREACH_HASH_KEYVAL in ext/http/http_misc.h. That macro somehow loads <stdint.h> and was causing the fatal errors. Without changing VC11's stdint.h (which I did for debugging) I could only think of one way to prevent it from loading. stdint.h starts with: /* stdint.h standard header */ #pragma once #ifndef _STDINT #define _STDINT So, by defining _STDINT you can prevent it from loading. I changed <win32/php_stdint.h>: #ifndef _MSC_STDINT_H_ // [ #define _MSC_STDINT_H_ +#define _STDINT #if _MSC_VER > 1000 #pragma once I do not know if this has more consequences. If <win32/php_stdint.h> is a full replacement for <stdint.h> this should not cause troubles. My PHP 5.5.16RC1 worked flawlessly with the exception of php_event.dll. But that was a SSL issue, because I have been trying to get that to work (by compiling the libevent libswith OpenSSL support). New build at https://phpdev.toolsforresearch.com/php-5.5.16RC1-nts-Win32-VC11-x86.zip See the phpinfo.htm inside the zip for the details of the build. Jan NB: I will ask the maintainer of php_qb.ddl and php_av.dll to call <win32/php_stdint.h> if possible in stead of loading his own copy. php_ffmpeg.dll is not maintained anymore and I seem to be the only one that is compiling it for Windows, so that will not be a problem at all.