Hrm... alright... any ideas/workarounds? I'm pretty handy with C
and PostgreSQL if you want me to test anything or hack something into
place. -sc
Are you sure you limited your syslog trace to a single process? I
don't want to end up chasing the wrong trail here.
Yup. I can guarantee that there is only one user on the backend. I've
got perdition in front and have only enabled it for my account.
And what imap client are you triggering this with? I haven't been able
to reproduce this bug with just the search command, so apparently
there's something else involved as well.
OS-X's Mail.app.
Other than that, _ic_search is already under scrutiny because of poor
performance.
It'd require a change in the database structure, but did you see my
other post about breaking things down by header? That'd get rid of the
need for any LIKE/REGEXP's when looking for a message, which should be
quite quick.
Perhaps some of the other guys has some input on what might be causing
this. I've been suffering from what may well be this very problem on
one of my main production installations. All outlook-office clients
there.
Yuk. At least I'm not the only one experiencing the problem though (I
almost feel a tad better). You haven't found a workaround, have you?
Not advertise the ability to search? *shrug*
I hope I can get this resolved in the next week.
This bug and the pop3 loop should be considered *grave* and must be
resolved asap.
Alright, here's what I know:
*) _ic_search() never terminates it's do{} while(); loop starting at
2033.
*) result is always 1, and i never grows beyond 10 (sounds like a stack
corruption error).
*) Tried setting MAX_RETRIES to 10 to see if that'd help since that was
the max value of i while diving through some hand placed debugging
statements... but it didn't, loop continues even though i is ten,
MAX_RETRIES is ten, and the second part of the test is i < MAX_RETRIES.
*) I can manually set result to zero at the end of the do{} while();
loop in _ic_search() and the loop does terminate, however the value of
i is 14. Definitely feels like stack corruption someplace.
Any ideas? -sc
--
Sean Chittenden
