masarati@aero.polimi.it writes:
I guess we need to eliminate the unlock/lock of ld_res_mutex and inform ldap_new_connection() about whether ld_req_mutex is locked or not...
How about lock order? Must ld_res_mutex be held when locking ld_req_mutex? Anyway, sounds like it does need an ITS, which the commits should refer to.
By now, I'm just adding thread debugging code, which is harmless in production, and is only meant to track the issue. I'll open an ITS and reference it as soon as I start introducing actual modifications to the code.
p.