Howard Chu writes: > At any rate, ldap_pvt_thread_initialize appears to be devoted to the > required setup for actually spawning threads, and doesn't have any > effect on setting up mutexes or any other locking primitives. Which > makes sense; if any of those primitives needed elaborate setup they > wouldn't really be "primitives" would they...
No idea. I can imagine them needing to request kernel support first, but since none of the supported variants seem to need that... Actually I think I suggested it because it seemed to fit libldap_r, not because of anything I know about threading. >> but in any case - is there any reason not to always call >> ldap_pvt_thread_initialize() early, maybe just before main() checks >> if argv[0] is a tool name? > > I guess there's no harm in doing it. There's probably more in > slapcommon.c that we should factor out as well. Yes, slapcommon vs. main() is a bit confusing. E.g. slap_tool_destroy does ldap_pvt_tls_destroy() even though main does ldap_pvt_tls_init(). -- Hallvard
