Fixed bug #50333 Improving multi-threaded scalability by using emalloc/efree/estrdup #500
Conversation
* PHP-5.5: Assume the free space is correct on Travis CI.
… PHP-5.5 * 'PHP-5.5' of https://git.php.net/repository/php-src: Assume the free space is correct on Travis CI.
* 'master' of https://git.php.net/repository/php-src: Assume the free space is correct on Travis CI.
Conflicts: ext/openssl/xp_ssl.c
* PHP-5.5: Added support for TLSv1.1 and TLSv1.2 Conflicts: ext/openssl/xp_ssl.c
* PHP-5.5: TLS news
This reverts commit 2aaa3d5.
This reverts commit 62be976.
* PHP-5.5: Revert "TLS news" Revert "Added support for TLSv1.1 and TLSv1.2"
…-src * 'updated_tls_support' of https://github.com/rdlowrey/php-src: Added support for TLSv1.1 and TLSv1.2 Conflicts: ext/openssl/xp_ssl.c
… PHP-5.5 * 'PHP-5.5' of https://git.php.net/repository/php-src: Revert "TLS news" Revert "Added support for TLSv1.1 and TLSv1.2" TLS news Added support for TLSv1.1 and TLSv1.2
* 'master' of https://git.php.net/repository/php-src: TLS news previous revert killed that file Revert "TLS news" Revert "Added support for TLSv1.1 and TLSv1.2" TLS news fix ws Added support for TLSv1.1 and TLSv1.2 Added support for TLSv1.1 and TLSv1.2
* 'ldap_escape-dev' of https://github.com/DaveRandom/php-src: Add ldap_escape()
…a/php-src * 'openssl_compile_warning_fix' of https://github.com/bukka/php-src: Fix compiler warnings in openssl.c
|
The Travis build failed because tsrm_virtual_cwd.h is used in ext/phar. Log: https://s3.amazonaws.com/archive.travis-ci.org/jobs/12722496/log.txt |
|
Thanks, fixed. |
|
After Pierre's idea I'm going to convert this PR to less intrusive, without moving the virtual cwd stuff into zend and doing that much renames. |
|
hm, now the build succeeded but it still shows it'l failed https://api.travis-ci.org/jobs/12727144/log.txt?deansi=true |
|
The build still shows ton of failed unit tests as far as I can see, and memory errors: Also, we need some benchmarks to ensure we are getting performance improvement before introducing this big a patch. |
|
Yep, that's what Dmitry has pointed out in this mail http://news.php.net/php.internals/69730 . So I'll follow his suggestion with do_alloca() first, to see if it makes sense at all to make the less intrusive variant. Thanks. |
| memcpy((d)->cwd, (s)->cwd, (s)->cwd_length+1); | ||
|
|
||
| #define CWD_STATE_FREE(s) \ | ||
| free((s)->cwd); | ||
| efree((s)->cwd); \ | ||
| (s)->cwd = NULL; |
There was a problem hiding this comment.
As I understood, this new assignment makes sense only for call from virtual_cwd_deactivate(). May be it makes sense to move it there.
|
The perf test results after the recent changes http://windows.php.net/downloads/snaps/ostc/pftt/perf/results-20131101-MasterVanilla-Master50333Patch2-2742.html |
|
Looks like this one is ready to merge as Dmitry gave his OK http://news.php.net/php.internals/70009 |
|
merged into master |
Things done