Closed GoogleCodeExporter closed 9 years ago
Interesting -- I don't have a win2k machine to test on, so I don't know what
might be
going on here. It looks like something is being set to NULL that ought not be,
but I
don't know what it is. In theory tcmalloc should work fine on win2k.
I'm planning on making a new tcmalloc release this week (fingers crossed!) It
has a
few windows fixes, which might magically fix the problem here. I know you've
already
upgraded the OS, but if you will happen to be able to try the new tcmalloc and
see if
it works better on win2k, that would be great to see. Otherwise, I don't know
how
much I can look into this.
Original comment by csilv...@gmail.com
on 15 Apr 2009 at 7:01
I will try it out.
Original comment by weidenri...@gmx.de
on 15 Apr 2009 at 7:16
If I make GetStackTrace in stacktrace_win32-inl.h return 0 everytime, there is
no
crash. But this is surely not a valid fix.
Original comment by weidenri...@gmx.de
on 15 Apr 2009 at 8:52
Actually, that fix is fine. GetStackTrace is still rather experimental on
windows.
It isn't required for correctness, it just gives access to some debugging info.
So
feel free to use that for now. This narrows down where the crash might be!
That's
very helpful.
Original comment by csilv...@gmail.com
on 15 Apr 2009 at 9:06
We recently discovered that RtlCaptureStackBackTrace is present in ntdll on
Win2k,
but it crashes if you call it. I don't know if it crashes on every input, but
it
seems to be unusable for anything serious.
Original comment by dvi...@gmail.com
on 15 Apr 2009 at 10:49
That's good to know. I'll update the docs for Win2K.
Original comment by csilv...@gmail.com
on 15 Apr 2009 at 11:12
(Not really fixed, but at least documented!)
Original comment by csilv...@gmail.com
on 17 Apr 2009 at 5:23
Original issue reported on code.google.com by
weidenri...@gmx.de
on 15 Apr 2009 at 5:52