[webkit-reviews] review denied: [Bug 22614] Need to add Win32 implementation of ThreadSpecific. : [Attachment 25731] Proposed Patch

[bugzilla-daemon at webkit.org]

Alexey Proskuryakov <ap at webkit.org> has denied Jian Li <jianli at chromium.org>'s
request for review:
Bug 22614: Need to add Win32 implementation of ThreadSpecific.
https://bugs.webkit.org/show_bug.cgi?id=22614

Attachment 25731: Proposed Patch
https://bugs.webkit.org/attachment.cgi?id=25731&action=review

------- Additional Comments from Alexey Proskuryakov <ap at webkit.org>
Sorry, I pasted a wrong MSDN link, it was supposed to be
<http://msdn.microsoft.com/en-us/library/ms686997(VS.85).aspx>.

MSDN suggests calling TlsFree from DLL_PROCESS_DETACH, and it is indeed OK to
call it manually, or even not to call it at all - but I'm not talking about
leaking TLS keys. The problem is about leaking TLS data of each thread. In
pthread implementation of this class, destroy() is called automatically for
each ThreadSpecific object on thread exit, because the callback is registered
by pthread_key_create() - but nothing calls destroy() in this patch.