On Wed, Oct 19, 2011 at 9:58 AM, Leonid Evdokimov <l...@darkk.net.ru> wrote: > Here are some patches. > > 0001 fires without following fixes
I like it; applying with a few comments and style tweaks. (The style tweak is that I generally do "struct foo *bar;" rather than "struct foo* bar;" because in C syntax, the * really does apply to the declared variable rather than to the type. Consider how if you want to declare bunch of pointers, you can't say "struct foo* bar, baz, quux;") > aff6ba1 Fix request_finished memory leak with debugging turned on. > 9b724b2 Fix evsig_dealloc memory leak with debugging turned on. > > 0002 is just a preparation for 0003 I like 0002; merged that too. > 0003 fires with current codebase but, seems, it should not. I'm sorry, > I can't produce a nice fix right now, so I'm just sharing the testcase > at this moment. :) Hm. I'll look too, but if we don't find anything, this should probably turn into an issue on one of the trackers so that I don't forget about it later. event_reinit is tricky and fragile code; I am sad but not shocked to see a bug there. I tried poking at it a little bit with gdb to see where the failure was happening, but the bug seemed to evaporate when I added breakpoints: ouch. I'll try to look more. Have you tried running the test under valgrind? If there's really a double-unassign happening, it might notice something wrong. > And the last question: should I send further patches to ML or via > github pull request? Both are fine! I used to be very picky about how I got patches, but right now I am trying to accomodate as many different workflows as possible in order to learn more about which ones work for me. cheers, -- Nick *********************************************************************** To unsubscribe, send an e-mail to majord...@freehaven.net with unsubscribe libevent-users in the body.
