Front page | perl.perl5.porters |
Postings from September 2000
Re: [ID 20000728.005] perl -P broken
Thread Previous
|
Thread Next
From:
David Dyck
Date:
September 5, 2000 17:59
Subject:
Re: [ID 20000728.005] perl -P broken
Message ID:
Pine.LNX.4.05.10009051737190.16729-100000@dd.tc.fluke.com
I've taken another look at this, see my notes at the bottom
On Mon, 31 Jul 2000, Jarkko Hietaniemi wrote:
> On Fri, Jul 28, 2000 at 05:19:16PM -0700, David Dyck wrote:
> >
> > This is a bug report for perl from dcd@tc.fluke.com,
> > generated with the help of perlbug 1.31 running under perl v5.6.0.
> >
> >
> > -----------------------------------------------------------------
> > [Please enter your report here]
> >
> > cd t
> > ./perl harness .//comp/cpp.t
> > .//comp/cpp.........FAILED before any test output arrived
> > FAILED--1 test script could be run, alas--no output ever seen
> >
> >
> > # perl 5.6 rsynced from Jun 23 print ok 3 as expected,
> > ( echo '#define OK "ok 3\n"' ; echo 'print OK;' ) | perl -P -
> > ok 3
> >
> > # but todays (Fri Jul 28 17:17:47 PDT 2000) code doesn't print ok
> > # fetched with
> > # rsync -auvz rsync://ftp.linux.activestate.com/perl-current/ perl-current
> > ( echo '#define OK "ok 3\n"' ; echo 'print OK;' ) | ./perl -P -
>
> That's strange... firstly, I do not off-hand remember any patch lately
> touching the -P option, at least not directly. Secondly, for me the
> test works in Linux, and the configuration differences are:
>
> you me
>
> Linux 2.4.0-test5 2.2.16
> gcc 2.7.2.3 2.95.2 20000313 (Debian GNU/Linux)
> lseeksize *) 4 8
> libc libc.so.5.4.44 libc-2.1.3.so
> libs -lgdbm -ldbm -ldb -ldl -lm -lc
> -lnsl -lndbm -lgdbm -ldb -ldl -lm -lc -lposix -lcrypt -lutil
>
> *) We both have -i686 and -D_LARGEFILE_SOURCE (and -D_FILE_OFFSET_BITS=64)
> but our lseeksizes differ. Go figure.
>
> In other words: I do not know what causes the breakage for you.
> But I suspect our gcc difference the most.
When I reported this bug, all I said that changes between Jun 23
and Jul 28 cause perl -P to break.
I took the effort to apply all the diffs before Jun 23 to
a fresh copy of 5.6.0, and then applied each diff patch
in turn until my test case failed. I discovered
that diffs up to and including
6363.gz and 6366.gz (Jul 11 14:23 and Jul 11 14:24)
(( Change 6363 by gsar@auger on 2000/07/11 18:49:43
Change 6366 by gsar@auger on 2000/07/11 19:11:18 ))
must be applied to reproduce the failure, and
patch -R backs out the failure and my test case works
again. (I was able to test with just miniperl, so it
wasn't too painful).
I hope this will provide more information, otherwise I guess I could
break these diffs into individual patches and apply each one
until the test case fails.
I couldn't find the individual diff files on the
activestate ftp server, but they used to be in
ftp://ftp.linux.activestate.com/pub/staff/gsar/APC/diffs
and I used to get them from
rsync -auvz rsync://ftp.linux.activestate.com/perl-diffs/ diffs
David Dyck
Thread Previous
|
Thread Next