On Wed, 08 Sep 2004 09:14:56 -0600, Jean-Jacques Clar <[EMAIL PROTECTED]> wrote: > > >It should not be possibe for two threads to atomically decrement the refcount on > >the same object to 0. > I think there is a small window in there where it is possible to have the decs > happening on both CPUs one after the other making that bug possible in > decrement_refcount() and memcache_cache_free(). > > >Sounds > >like a bug in netware's apr_atomic_dec() function. > > Sorry but when running the same test on SLES9 with 2.0.51rc1, > my error_log is full of threads segfaulting.
I loaded the latest APACHE_2_0_BRANCH cache code into a 2.0.47-based server on SLES8/PPC. Not hard to make segfault. I changed MCacheMaxObjectCount to 1 from what you posted and had clients iterate through a list of 20 or so /index.html.LANGUAGE files. I changed the worker MPM configuration to use exactly 1 child process with 800 threads. My backtrace was (sorry, no line numbers): pthread_mutex_lock() free() mod_mem_cache.so mod_mem_cache.so libapr-0.so.0 apr_pool_destroy() httpd ap_run_process_connection() Next I'll try to replicate with 2.0.51 rc2 on the same box.
