Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site cithep.UucP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!princeton!astrovax!cithep!tim
From: tim@cithep.UucP (Tim Smith )
Newsgroups: net.unix-wizards
Subject: Re: Recovery possible from signal aborted write(2) ?
Message-ID: <45@cithep.UucP>
Date: Sun, 23-Sep-84 06:08:54 EDT
Article-I.D.: cithep.45
Posted: Sun Sep 23 06:08:54 1984
Date-Received: Wed, 26-Sep-84 06:42:58 EDT
References: <3253@ecsvax.UUCP>, <138@rlgvax.UUCP>
Organization: Caltech HEP, Pasadena, CA
Lines: 18

****** WARNING: KLUDGE ALERT! KLUDGE ALERT! KLUDGE ALERT! *******

If you have a kernel that leaves u readable by the process,
then you can try looking at u.u_count to see how many characters
were not transfered.  You have to read this directly, though,
because going through /dev/kmem will trash this variable.

Also, this may not work on signals gernerated from the tty, because
in attempting to flush the output buffer, it may screw u_count.

If you have source, you could change the routine that flushes the
output buffer to preserve u_count.

By the way, is there any reason to not have u directly readable?
It's been that way for about two years here at cithep, but I think
only two or three people know it!
-- 
Tim Smith		ihnp4!cithep!tim  or  ihnp4!wlbr!callan!tim