On Thu, 2007-03-01 at 11:43 -0300, Daniel K. O. wrote:
> Keith MARSHALL escreveu:
> >> Either way, this is not related to MinGW.
> > Actually, it is! In the paragraph you snipped, the OP asked if it
> > was possible to use the poll function in MinGW. That the answer is
> > `no', or even that the question had already been asked, by another
> > user, and answered as recently as the beginning of this week, does
> > not make it any less relevant.
> I think you are focusing in answering what he asked (poll), instead of
> what he needs to know (mpich).
> Maybe I wasn't clear in what was my goal in the original message, and
> left it open to be misinterpreted.
> > Daniel, please do not dismiss questions like this, on this list!
> I was trying to address his need to compile mpich(2), so I pointed that he:
> 1) probably doesn't need to because of the availability of win32 binaries
> 2) doesn't need sys/poll.h, because mpich2 also works on Windows
> 3) Cygwin provides the emulation you may need for unix; (the possibility
> that the win32 binary actually uses Cygwin)
> So if for some reason he can't / don't want use the win32 binary, he
> just needs to find out how to compile it; that directly depends on the
> mpich2 developers (and maybe Cygwin, or the "other supposed build
> process" developers). So he either doesn't need poll (and thus don't
> need MinGW's list help providing one), or he needs poll from the same
> place the win32 mpich2 binary is already getting it.
> Yes, the discussion IS related to MinGW, but the solution isn't.
Unless, as it was pointed out, that a replacement for poll was written
for MinGW. Possibly using select, which of course would only work on
sockets in Windows, but may be all that's needed to compile mpich under
So could it be worth writing a functionally limited poll for MinGW?