Travis related improvements #654
Conversation
|
This makes Travis only test in debug mode. I think it may be better to run both. |
|
makes sense, I've made that also a build parameter, but this means we will use 4 jobs, instead of 2, but I think it is reasonable. |
|
Do we really need to run 30 minute jobs for 4 different configs across 3 branches for every push? This seems like a very excessive use of Travis. We don't really gain much by running the full ts/dbg matrix here, imho. |
|
Maybe on pulls, but not on regular builds? Not sure if Travis allows this. |
|
the TRAVIS_PULL_REQUEST environment variable could be used to shortcut the compilation step, so that we only do the heavy lifting for pull requests but imo that would be a bit awkward workflow. |
|
I think it should be enough to do a zts-dbg build and an nts-ndbg build. nts-ndbg being the standard production build and zts-dbg being the "nobody forgot a TSRMLS_CC and there are no debug warnings" build. The other combinations don't really offer much additionally. If those two work, then zts-ndbg and nts-dbg will pretty likely work as well. |
|
Debug builds can help us notice leaks as the zend mm warning will make the I'm not against the idea just making sure we are on the same page.
|
|
Yes. Which is why I suggest doing the zts-dbg and nts-ndbg builds, but not nts-dbg and zts-ndbg. |
|
Ok, fine with me.
|
* PHP-5.5: only build and test nts non-debug and ts debug builds as discussed at #654
* PHP-5.5: only build and test nts non-debug and ts debug builds as discussed at #654
…_56_backport * origin/PHP-5.6: (28 commits) /me blushes, really hard only build and test nts non-debug and ts debug builds as discussed at #654 Fix mysqli build when building against libmysqlclient. Fix author name on the #63228 patch. Fixed bug #67169: []= after_array_splice incorrect Revert "show leaks in their own group" using valgrind would be a bit too much for travis I guess. show leaks in their own group Use the right path for the suggested PHP invocation in ext_skel. Changed to more understandable NEWS message one heading is enough add new NEWS block for alpha3 XFAIL the leaking multibyte tests, we should really get to the bottom of this Fixed numbering in UPGRADING file Updated NEWS for fixed phpdbg issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 add missing NEWS entry to the correct release where it was added 5.5.12 was released yesterday add missing NEWS entry to the correct release where it was added Updated NEWS entry for phpdbg ... Conflicts: ext/curl/interface.c
…into str_size_and_int64 * origin/str_size_and_int64_56_backport: (32 commits) fixed ZEND_DEBUG usage updated libmagic.patch added missing stdint header, bug #66907 fix macros /me blushes, really hard only build and test nts non-debug and ts debug builds as discussed at #654 Fix mysqli build when building against libmysqlclient. Fix author name on the #63228 patch. Fixed bug #67169: []= after_array_splice incorrect Revert "show leaks in their own group" using valgrind would be a bit too much for travis I guess. show leaks in their own group Use the right path for the suggested PHP invocation in ext_skel. Changed to more understandable NEWS message one heading is enough add new NEWS block for alpha3 XFAIL the leaking multibyte tests, we should really get to the bottom of this Fixed numbering in UPGRADING file Updated NEWS for fixed phpdbg issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 ...
…into str_size_and_int64 * origin/str_size_and_int64_56_backport: (32 commits) fixed ZEND_DEBUG usage updated libmagic.patch added missing stdint header, bug #66907 fix macros /me blushes, really hard only build and test nts non-debug and ts debug builds as discussed at php/php-src#654 Fix mysqli build when building against libmysqlclient. Fix author name on the #63228 patch. Fixed bug #67169: []= after_array_splice incorrect Revert "show leaks in their own group" using valgrind would be a bit too much for travis I guess. show leaks in their own group Use the right path for the suggested PHP invocation in ext_skel. Changed to more understandable NEWS message one heading is enough add new NEWS block for alpha3 XFAIL the leaking multibyte tests, we should really get to the bottom of this Fixed numbering in UPGRADING file Updated NEWS for fixed phpdbg issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 Added stdion/stdout/stderr constsnts and their php:// wrappers Fixes issue #85 ...
After the recent discussion on the mailing list about our current travis setup, I've did some minor adjustments.
I've