I have experienced a core dump with mongoose 2.8 under Scientific Linux 5.1
which is essentially RHEL5. I looked at the core dump and I believe I
understand the problem but do not know what the right way to fix it is. The
core dump happens at line 4530 in the worker_thread when trying to lock the
mutex. Inspecting the core dump reveals that conn is 0, as set with the
memset on line 4508. Appararently the reading loop below 4508 fails for
an unknown reason. Then the code below the loop becomes wrong, as the ctx is
0. My idea to fix this would be to have a local mg_context and assign it to the
one in the conn when available and to the one passed as a parameter else. But
I am am unsure if this is a good solution; there must be a reason why the code
makes a local copy of the mg_context when reading.
Going to 2.10 is not an option due to the API change. In fact, the existence of
a ready made dispatcher was a reason why I decided for mongoose in the first
place.
Original issue reported on code.google.com by mkoenne...@gmail.com on 14 Sep 2010 at 9:00
Original issue reported on code.google.com by
mkoenne...@gmail.com
on 14 Sep 2010 at 9:00