diff options
author | Roger Dingledine <arma@torproject.org> | 2004-05-05 00:59:42 +0000 |
---|---|---|
committer | Roger Dingledine <arma@torproject.org> | 2004-05-05 00:59:42 +0000 |
commit | 608c15682049ee375f81655625b9f85d6a600a1a (patch) | |
tree | eef3cb9c1d379acfde6e897eb4fd925bf2045073 /src/or/main.c | |
parent | efdb356a81c58e2f6a0715acf55c0c9a5ea5d2e6 (diff) | |
download | tor-608c15682049ee375f81655625b9f85d6a600a1a.tar.gz tor-608c15682049ee375f81655625b9f85d6a600a1a.zip |
must close marked connections at the end of run_scheduled_events.
otherwise they might have a socket -1, and if we leave them around
they'll muck up poll/select.
i think this was the cause of our win32 and os x fakepoll crashes,
and probably would cause other errors down the road.
svn:r1786
Diffstat (limited to 'src/or/main.c')
-rw-r--r-- | src/or/main.c | 10 |
1 files changed, 3 insertions, 7 deletions
diff --git a/src/or/main.c b/src/or/main.c index cc3789e85c..122af87bb9 100644 --- a/src/or/main.c +++ b/src/or/main.c @@ -445,16 +445,12 @@ static void run_scheduled_events(time_t now) { last_uploaded_services = now; } -#if 0 - /* 6. and blow away any connections that need to die. can't do this later - * because we might open up a circuit and not realize we're about to cull - * the connection it's running over. - * XXX we can remove this step once we audit circuit-building to make sure - * it doesn't pick a marked-for-close conn. -RD + /* 6. and blow away any connections that need to die. have to do this now, + * because if we marked a conn for close and left its socket -1, then + * we'll pass it to poll/select and bad things will happen. */ for(i=0;i<nfds;i++) conn_close_if_marked(i); -#endif } static int prepare_for_poll(void) { |