[Webkit-unassigned] [Bug 113168] LayoutTests/fast/js/large-expressions.html crashes on Linux

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=113168





--- Comment #16 from Allan Sandfeld Jensen <allan.jensen at digia.com>  2013-04-07 13:33:31 PST ---
(In reply to comment #15)
> Created an attachment (id=196772)
 --> (https://bugs.webkit.org/attachment.cgi?id=196772&action=review) [details]
> patch after remove platform macro & move the hack to m_bound

Thanks. I may however have been a bit hasty in saying we don't need an ifdef for Linux.There seems to be a difference in opinion on whether guardsize is included or not in stacksize.
http://sourceware.org/ml/libc-help/2008-07/msg00076.html
https://bugzilla.redhat.com/show_bug.cgi?id=435337
http://www.sourceware.org/ml/libc-alpha/2008-05/msg00086.html

More worryingly it may get fixed in new versions of glibc
http://sourceware.org/bugzilla/show_bug.cgi?id=6973
http://sourceware.org/bugzilla/show_bug.cgi?id=11787

None of this matters to setting m_bound which should be correct now, but we may end up setting m_origin too short.

I suggest ifdeffing for GLIBC and in that case subtract guardsize from stacksize like you did the previous patch. m_origin can then be redefined as stackbase + stacksize + guardsize. Btw feel free to typecast with static_cast<char*>, it is shorter than the double reinterpret case.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.