Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!convex!convex.COM
From: tchrist@convex.COM (Tom Christiansen)
Newsgroups: comp.lang.perl
Subject: Re: Need $? for open()
Keywords: new feature request
Message-ID: <100906@convex.convex.com>
Date: 28 Mar 90 13:40:40 GMT
References: <23449@pasteur.Berkeley.EDU>
Sender: news@convex.com
Reply-To: tchrist@convex.COM (Tom Christiansen)
Organization: CONVEX Software Development, Richardson, TX
Lines: 55

In article <23449@pasteur.Berkeley.EDU> grady@postgres.berkeley.edu writes:
>I have a perl program which starts another process and reads
>the input, using open().  Specifially, the relevant part of the
>code looks something like:
>
>	open(RLOG, "rlog $files |") || die "rlog: $@";
>	while (<RLOG>) {
>	    ...
>	}
>
>The problem is that I don't find out if there's an error.  

I addressed this in <5016@convex.convex.com> of very nearly two 
months ago.  It's brief, so I'll include the original here.

--tom

    Date:         31 Jan 90 05:54:28 GMT
    From:         tchrist@convex.com (Tom Christiansen)
    Subject:      Testing successful pipe opens
    Organization: Convex Computer Corp, Richardson, TX
    Message-ID:   <5016@convex.convex.com>
    Newsgroups:   comp.lang.perl


    It's known that opening pipes to non-existent programs doesn't 
    usually trigger useful error diagnostics.  In particular, while
    writing on a badly-opened output pipe will (eventually) trigger a
    SIGPIPE (which you can look for), reading from a badly-opened input 
    pipe just gets you EOF.

    A possible work-around for the bad input pipe follows:

	$kid = open (PIPE, "bogus_command |");  # really should check $kid != 0
	(kill 0, $kid) || die "bogus_command failed";


    This works fine if bogus_command doesn't have shell metas in it, but
    if it does, the shell may well not have exited before the kill 0.

    You can introduce a delay,

	$kid = open (PIPE, "bogus_command </dev/null |");
	sleep 1;
	(kill 0, $kid) || die "bogus_command failed";

    but this is sometimes undesirable and in any event does not guarantee
    correct behavior.  But it seems slightly than nothing.  

    --tom
--

    Tom Christiansen                       {uunet,uiucdcs,sun}!convex!tchrist 
    Convex Computer Corporation                            tchrist@convex.COM
		 "EMACS belongs in <sys/errno.h>: Editor too big!"