On Tue, 27 Oct 2009, John Engelhart wrote:
> Regardless, I really think that there should be some kind of "minimum" floor
> returned instead of 0/0. Even with a minimum of 1000 microseconds (1
> millisecond), that's 1000 times per second. Is there really a need to make
> an attempt with such frequency? Even trying 100 times per second seems a
> bit excessive, maybe more like 10 times per second.
Well, I agree that having a minimum amount would somewhat work around the
problem you've seen but I'd much rather try to actually address the bug than
trying to fix the results of it.
> Digging a bit more, maybe there's a bug in process_timeouts() ?
Possibly, sure.
> Actually, now that I'm looking at process_timeouts(), it looks like how
> timeouts are managed in general could use a lot of work.
I can only agree that the current system isn't that optimized, but I guess
that in most use cases there are rarely very many queries on the same channel.
-- / daniel.haxx.seReceived on 2009-10-28