Howard Chu wrote:
After all fixing, in my very unrealistic testing (massive modification and accessing of the very same two entries by three clients, plus a number of searches) I'm still seeing some deadlocks; would you consider this a problem, or could we consider it acceptable?Kurt D. Zeilenga wrote:
At 03:40 PM 3/17/2005, Howard Chu wrote:Hm.... OK. We don't need to move lutil_passwd, just the chk_sasl and chk_crypt functions. Since we can register password mechs from outside the library, this is easy enough.
And then move the mutex into the chk_crypt routine itself.
Moving the mutex into -lutil might be a bit problematic. The library is designed not to require -lldap_r (or -lldap or -llber).
You could resolve the problem by moving lutil_passwd into slapd(8).
p.
SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497