Comment #24 on issue 99 by [email protected]: Memcached 1.4.2 server
segmentation fault
http://code.google.com/p/memcached/issues/detail?id=99
Staring at this stupid bug, again.
Can you point out where in memcached.c (from 1.6) you believe it to be
manually setting the event.ev_base? The update_event() functions look
essentially identical.
The only place it's being manually set in 1.6 (that I can see) is in the
TAP code where it's adding an event to a different base.
I've replicated your -vvv output precisely, and have a script running which
is attempting to set/get/delete a value of all possible sizes on a 32bit
system. No crash.
Can you try the attached patch, and let me know which (if either) of the
assert's your test hits? One thing we're doing wrong is not checking the
output of event_base_set, but it doesn't look like that'd make a difference
here.
Thanks! I don't want to give up on this, but I want to make sure it's done
right.
