Re: [RFC] TSRM patch

Hi!

How shall we continue with this patch?

I think we need to figure out what happens with dynamic loading, especially on Windows. Then having tsrm_ls stored in __thread variable seems to be the best solution, if storing globals in __thread doesn't work with DLLs (which appears to be the case). But I think we need to chose one that works and stick with it - two TS models in one code (and one non-TS) is quite enough, adding one more variant is excessive IMHO.

Dmitry, Lukas, Arnaud and others interested, would it make sense to proceed with the current non-ZTS signals implementation in 5.3 and work on ZTS in HEAD with a goal to improve performance?

With signals, the thing as far as I can see is that this patch does not make PHP 100% robust against random signals - there are a lot of places in the engine where engine can be left in inconsistent state if interrupted in random place (though probability of this happening is low, it's definitely not zero). I also see it still adds a bunch of syscalls per request (at least the version linked from the wiki). So I'm not sure ZEND_SIGNAL patch is mature enough right now - unless I have looked at the wrong one - in which case could somebody update the wiki? -- Stanislav Malyshev, Zend Software Architect stas@zend.com http://www.zend.com/ (408)253-8829 MSN: stas@zend.com