[prev] [thread] [next] [lurker] [Date index for 2007/01/10]
> 3. The part I really like: if bk (or whatever it spawns down it's > guts) *successfully proceeds* after reaching this timeout, wouldn't > it be a nifty optimization to go ahead without even *trying* to > lock the "authority" file, whatever that means? Because it's continuing after the timeout to a lock, assuming that some dead process left turds. Worse would be if it didn't try to lock the file at all and stomped all over itself, or if it waited forever for the lock to become available. That said, the bad part is that it leaves turds in the first place, and that if the workarounds suggested (which keep it from communicating with X) work, then clearly it doesn't need to communicate with X at all and should just not try in the first place. And for my third and final hate, if it has determined that the authority file can't be properly locked because of a stale lockfile, then it should *remove* the bad lockfile so that the next process (or even the same process) can do some (relatively) sane locking.
Generated at 00:01 on 12 Jan 2007 by mariachi 0.52