https://bugs.openldap.org/show_bug.cgi?id=9340
--- Comment #3 from Howard Chu hyc@openldap.org --- (In reply to stacey.marshall from comment #2)
Hi,
Sorry for the delay. Having compiled without optimization and with -g the stack is as follows:
(dbx) where current thread: t@2 [1] __lwp_sigqueue(0x0, 0xa0037948a68, 0x6, 0x0, 0xffffffffffffffff, 0x0), at 0x7fecce46e17a8 [2] raise(0x6, 0x7feccdfbff0c0, 0x5, 0x5, 0x0, 0x0), at 0x7fecce462af0c [3] abort(0x1, 0x1210, 0x0, 0x1000, 0x7fecce483c278, 0x1a278), at 0x7fecce45fa144 [4] _assert(0x10006ada0, 0x10006adc0, 0x387, 0x0, 0xd3f6300140, 0x7fecce4822000), at 0x7fecce45fb158 =>[5] slapd_add(s = 9, isactive = 0, sl = 0xd3f6300310, id = 1), line 903 in "daemon.c" [6] slapd_daemon_task(ptr = 0xd3f6879f60), line 2392 in "daemon.c"
I compiled up a version with LDAP_THREAD_DEBUG which exited with:
thr_debug.c:434: ldap_pvt_thread_mutex_lock error: usage->magic is 0
You also have to run this under the debugger, to get a stack trace for where that error occurred. (thread_debug.c:434 is simply where that error message is.)
Using truss I'm seeing that some of the initialization comes from cyrus-sasl. Here we have v2.1.26 (+fixes), while I see 2.1.27 is released (with updates provided by Howard Chu).
Let me know if I can provide anything else from the dump.