Fix problem where rsync_ctx objects representing connections that had

timed out would get stomped and restarted.  Handle timeouts as
failure, and put repositories on which we've timed out into the
dead_host list.  Rework rsync_mgr() yet again, still trying to juggle
the three-way wait for timeouts, subprocess I/O, and exiting
subprocesses.  Parse the warning message rsyncd sends when we've hit
its per-module connection limit, as a first step towards
per-repository connection limits (not that this helps when a
repository is behind a firewall that starts whacking connections at
the transport layer...).

svn path=/rcynic-ng/rcynic.c; revision=3967

0 files changed, 0 insertions, 0 deletions