Hi,
I don't like the idea to use pthread_mutex directly instaed of our
wrapper. If this change make a panic go away then we doing something
wrong in our wrapper and wrapper should be fixed.
Only and only when it is not fixable (I don't think so because apr
(apache) uses wrapper), we should commit this patch.
Thanks,
Alex
ahh vote: -0
Stipe Tolj schrieb:
Hi list,
as many of you may now, we support the Cygwin 1.2.x (POSIX.1 layer on
Win32) natively, but claimed always to be "unstable" on this platform
due to lacks in the pthreads implementation of Cygwin.
I tried to debug the "manged to lock the mutex twice" bug we face on
higher load of the daemons. It seems that our callback scheme has an
issue, presumably the gwthread_self() way we provide to the openssl core
to identify the associated thread. Our mutex struct is based on pthreads
itself. Now, what I did is to "exchange" our wrapped way by native
pthread calls in the gwlib/conn.c layer for the openssl thread callback
scheme, and voila it works without crashing.
Asking anyone interested in running on Win32 via Cygwin to test and vote.
Thanks,
Stipe
-------------------------------------------------------------------
Kölner Landstrasse 419
40589 Düsseldorf, NRW, Germany
tolj.org system architecture Kannel Software Foundation (KSF)
http://www.tolj.org/ http://www.kannel.org/
mailto:st_{at}_tolj.org mailto:stolj_{at}_kannel.org
-------------------------------------------------------------------
