Re: select/poll bug?

[Roy Marples <roy%marples.name@localhost>]

On Sat, 2008-08-02 at 23:28 +0200, Joerg Sonnenberger wrote:
> On Sat, Aug 02, 2008 at 10:09:54PM +0100, Roy Marples wrote:
> > Lets say that select can handle a greater timeframe them as it's time_t
> > is in seconds, vs an int as milliseconds for poll.
> 
> Doesn't really matter. You have to deal with preliminary wakeups anyway,
> so you just have to round down to the largest possible timeout.

Not entirely sure what you mean here, but some research just put a nail
in using select - well it's coffin anyway. POSIX only says it has to
wait upto 31 days. Linux has a hard limit which is hard to work out, so
I'm now looping on poll. I really really wish the man pages were better
documented about this!

Google indicates that the Mac OSX poll error is with POLLOUT and we're
just using POLLIN so we should be safe there.

Thanks

Roy