Front page | perl.perl5.porters |
Postings from February 2004
On Sat, Jan 31, 2004 at 10:33:42PM +0000, Dave Mitchell <davem@fdisolutions.com> wrote:
> On Tue, Jan 27, 2004 at 06:41:45PM -0000, Flash Sheridan wrote:
> > -----------------------------------------------------------------
> > The following code incorrectly prints -1 under Perl v5.8.2 build 808, but
> > correctly prints 0 for 5.6.1 build 631, with CygWin true.
> > CygWin's true is working correctly, as verified by echo %ERRORLEVEL%;
> > I originally noticed the problem with
> > an internal tool. More complicated code relying on the expected behavior
> > has been working for over a year.
> > Because of ActiveState's warning about dual installations, I haven't
> > yet tried both versions of Perl on the same machine;
> > the bug is on a new Windows XP box.
> >
> >
> > use strict; use warnings; use diagnostics; use English; use 5.6.0;
> > my $theSysText = qx("true 2>&1");
> > print "\nCHILD_ERROR: $CHILD_ERROR\n";
>
> First, a disclaimer - I don't use Windows, so some of the following is
> speculative. However, the '2>&1' syntax is specific to UNIX (Bourne-like)
> shells, and is unlikely to work under Windows. A return code of -1
> usually means that Perl was unable to execute the command. For example on
> my Linux system:
>
> $ ./perl -e '$a=`/xxx`; print "\$?=$? \$a=[$a] \$!=[$!]\n"'
> $?=-1 $a=[] $!=[No such file or directory]
>
> Try printing out $! to see what the error is, then see if the problem goes
> away by removing the 2>&1.
Did perl get smarter about not invoking a shell when doing "foo 2>&1"
between those two versions?
If so, it may be that true is not in the path, and the error was hidden
(because $? was the return code of the shell, not of true) under 5.6.1.
I'm curious about this "activestate warning" about dual installations,
do you have a link to it?
Thread Previous
|
Thread Next