Ok, let's start softly, someone has to be the bad guy ;-)

I have a very strong feeling against the patch which has been commited
relative to the use of pselect().

I just say that in its _current_ implementation it needs more work
done. And that in its current implementation it only introduces
problems and at least overhead for platforms that lack poll() and
pselect().

The introduction of this patch changes the 'heart' on which libcurl
builds. If the 'heart' is a common one across many platforms, tests
done for one are mostly valid for others. If we use a different
'heart' for each platform we will get platform dependant runtime
problems. If this is the direction to go, Ok. But in any case that
patch needs a more serious and robust implementation.

-- 
-=[Yang]=-