[PHP-DEV] Re: fix for 61418 breaks win32 build

2012-03-23 Thread Gustavo Lopes
On Fri, 23 Mar 2012 06:26:22 +0100, Stas Malyshev wrote: Commit 714f1ff4b37c5101b3c61ea108a3d415f41e50df is breaking win32 build and causing bug #61482. I am reverting this commit. Gustavo, please check the code and see if you can improve the fix so it does not break win32. The reproduction d

[PHP-DEV] Git php-src Tags Renamed

2012-03-23 Thread David Soria Parra
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Hi Internals we renamed the tags in the git repository to php-X.Y.Zdev. This avoids an issue with current QA code. git pull origin --tags should give you the new tags. David -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Us

[PHP-DEV] PHP5.4: PHP-FPM + Gettext == invalid pointer

2012-03-23 Thread Steven VAN POECK
Hi, Whenever I activate the gettext extension, PHP-FPM dumps a heap upon starting (see output below) Although the output finishes with "Aborted. Failed", the php processes are working ok, gettext too. PHP 5.4 was compiled with these options: ./configure --sysconfdir=/etc/php5 --with-config-

Re: [PHP-DEV] PHP5.4: PHP-FPM + Gettext == invalid pointer

2012-03-23 Thread Pierre Joye
hi, Please disable APC too to test. btw, do you use APC from current svn's trunk? Cheers, On Fri, Mar 23, 2012 at 12:51 PM, Steven VAN POECK wrote: > Hi, > > Whenever I activate the gettext extension, PHP-FPM dumps a heap upon > starting (see output below) > > Although the output finishes with

Re: [PHP-DEV] PHP5.4: PHP-FPM + Gettext == invalid pointer

2012-03-23 Thread Steven VAN POECK
Hi Pierre, Right. Disabling apc solves the issue, with or without gettext activated. No dump or segfault message anymore. I'm using the standard APC installed by pecl install apc. Version is 3.1.9. It seems to be working fine although it's responsible for the PHP-FPM dump. Should I use the ve

[PHP-DEV] Re: ECDSA support in the openssl extension

2012-03-23 Thread Sergey Ryazanov
Hello all. > Openssl starting from version 0.9.8 supports the ECDSA signature > algorithm. Return value of the openssl_get_md_methods() also includes > "ecdsa-with-SHA1" string. But if I call openssl_sign() with EC key > then I have an error: "key type not supported in this PHP build!". > > After

Re: [PHP-DEV] PHP5.4: PHP-FPM + Gettext == invalid pointer

2012-03-23 Thread Pierre Joye
yes, please do. We have fixed these errors recently. On Fri, Mar 23, 2012 at 1:13 PM, Steven VAN POECK wrote: > Hi Pierre, > > Right. Disabling apc solves the issue, with or without gettext activated. No > dump or segfault message anymore. > > I'm using the standard APC installed by pecl install

Re: [PHP-DEV] PHP5.4: PHP-FPM + Gettext == invalid pointer

2012-03-23 Thread Steven VAN POECK
Hi Pierre, Sorry, I didn't know these issues had been addressed to. I just installed the latest version from trunk. Everything runs smoothly, no segfaults or dumps anymore. Thanks again, Steven On 03/23/2012 01:25 PM, Pierre Joye wrote: yes, please do. We have fixed these errors recently.

[PHP-DEV] Validate complex types

2012-03-23 Thread François Gannaz
Hello I'm wondering if there is a recommended way to validate a multi-type parameter in an internal function. If a parameter is expected to be "long or string" or to be "integer or null", a simple call to `zend_parse_parameters()` won't validate it. I read some extensions' code, and it seems that

[PHP-DEV] Re: fix for 61418 breaks win32 build

2012-03-23 Thread Stas Malyshev
Hi! > Actually, Dmitry and I were discussing this issue yesterday on IRC (check > your logs). He too detected the problem and had reverted the fix for bug > #61418. Sorry, I don't watch IRC lately. > The fix for bug #61482 may be simple and I will both reinstate my fix for > bug #61418 and

Re: [PHP-DEV] Validate complex types

2012-03-23 Thread Stas Malyshev
Hi! > I wanted to use `zend_parse_arg()` to validate a single parameter if the > first call to `zend_parse_parameters_ex()` wasn't successful. But this > function > is never used outside of "Zend/zend_API.c". Is it meant for public use? Are > there any drawbacks in calling it? No, it's currently