Closed GoogleCodeExporter closed 9 years ago
This really seems to be a problem.
I am trying to trace down a rare bug happening on Windows only, which seems a
typical case of memory invasion/corruption: "An operation was attempted on
something that is not a socket".
Perhaps this issue is the cause of it, but I cannot say for sure.
Original comment by rodrigo...@gmail.com
on 27 Oct 2010 at 12:18
I had this problem too and found the bug reported here.
Under heavy load with ultra small requests this is easy to trigger, and the
above fix seems to have solved the issue.
Is it not still possible that a deadlock happens between WaitForMultipleObjects
and WaitForSingleObject?
Even if it could happen, its much more rare now.
Original comment by term...@gmail.com
on 13 Nov 2010 at 2:01
Just an update: after fixing this API, we have been running without the
mysterious bug for two weeks now. We used to have problems once a day before
the fix.
Original comment by rodrigo...@gmail.com
on 16 Nov 2010 at 11:51
That was fixed in
http://code.google.com/p/mongoose/source/detail?r=995a586a8e313124060baf52a581f9
379a49cb63&path=/mongoose.c, thanks for reporting.
Original comment by valenok
on 21 Jun 2011 at 9:43
Original issue reported on code.google.com by
alex.mac...@gmail.com
on 13 Oct 2010 at 10:37