Date: Sat, 08 Mar 2003 19:09:08 +0200 From: "Eli Zaretskii" Sender: halo1 AT zahav DOT net DOT il To: rich AT phekda DOT freeserve DOT co DOT uk Message-Id: <3099-Sat08Mar2003190908+0200-eliz@elta.co.il> X-Mailer: emacs 21.3.50 (via feedmail 8 I) and Blat ver 1.8.9 CC: djgpp-workers AT delorie DOT com In-reply-to: <3E69D8A1.F493B542@phekda.freeserve.co.uk> (message from Richard Dawe on Sat, 08 Mar 2003 11:48:49 +0000) Subject: Re: New POSIX: pwrite [PATCH] References: <200303071845 DOT h27Ij5d18334 AT envy DOT delorie DOT com> <3E6938EC DOT D9BFE2E4 AT phekda DOT freeserve DOT co DOT uk> <3028-Sat08Mar2003113243+0200-eliz AT elta DOT co DOT il> <3E69D8A1 DOT F493B542 AT phekda DOT freeserve DOT co DOT uk> Reply-To: Eli Zaretskii Errors-To: nobody AT delorie DOT com X-Mailing-List: djgpp-workers AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk > Date: Sat, 08 Mar 2003 11:48:49 +0000 > From: Richard Dawe > > So we have a choice: > > (1) Always fail the write for redirected stdout, stderr, just in case it's a > pipe. > > (2) Always allow the write for redirected stdout, stderr. > > Even in the case "foo | bar", the handle still refers to a file, even if it is > only a temporary one. So (2) seems good and works. Yes, I think we should take (2). > But will (2) result in > behaviour that is Unixy programs don't expect? I don't think they should be adversely affected. At most, they will get a new feature: pwrite doesn't work on Unix not because Unix doesn't want that, but because it's impossible to implement pwrite efficiently for pipes.