NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/sys/next/programm/5679

Text File | 1992-08-18 | 1.8 KB | 42 lines

Newsgroups: comp.sys.next.programmer
Path: sparky!uunet!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!destroyer!ncar!csn!boulder!alumni.cs.colorado.edu!rademach
From: [e-mail] (SIMON RADEMACHER)
Subject: Re: select() broken?
Message-ID: <[e-mail]>
Sender: [e-mail] (The Daily Planet)
Nntp-Posting-Host: alumni.cs.colorado.edu
Organization: University of Colorado, Boulder
References: <[e-mail]>
Date: Tue, 18 Aug 1992 23:42:07 GMT
Lines: 29

In article <[e-mail]> [e-mail] (SIMON RADEMACHER) writes:
[summary: select() did not see events on file descriptors after one
 disconnected.]
>
>5) When one of the accepted fd's disconnects, the read fd_set shows an
>   event on the fd.  A read is done resulting in EOF.  The fd is closed()
>   and removed from the read fd_set.
OOPS----^^^^^^^^^^^^^^^^^^^^^^^^^^^^
I lied. The array used to mark active file descriptors was cleared of the
disconnected fd. Hence, in the loop, the bit was not being set for that
fd, but it remained set from the previous go through the loop. Actually
clearing the bit fixes the problem.

6) From then on, the select() always waits for the duration of the time-
>   out, and returns 0, eg. no events in any of the sets.

Hmm, that fixes the problem, but is this a bug? When the bit for
a just close()d file is in the read fd_set, select() does not see any
events on any file descriptors. I would guess that if any invalid fd were
in the set the same thing would happen.  So, whether select() is broken
or not depends on if you think select() should be smart enough to ignore
invalid/closed fds.

In any case, thanks to those who sent mail with help.

--
Simon Rademacher
[e-mail]