GNU bug report logs - #7198
ls-misc failure with Oct 10 snapshot

Previous Next

Full log

Message #23 received at 7198 <at> debbugs.gnu.org (full text, mbox):

From: Jim Meyering <jim <at> meyering.net>
To: Paul Eggert <eggert <at> cs.ucla.edu>
Cc: 7198 <at> debbugs.gnu.org
Subject: Re: bug#7198: ls-misc failure with Oct 10 snapshot
Date: Thu, 14 Oct 2010 09:26:01 +0200

Paul Eggert wrote:
> On 10/13/2010 12:37 AM, Jim Meyering wrote:
>
>> What is the server running?
>
> NetApp 6.5.6.  NFS is configured to be NFSv3, running over TCP.
>
>> First step for that one should be to avoid the warnings from perl,
>> e.g., via the patch I suggested.
>
> Yes, thanks, somehow I missed the patch to tests/misc/ls-misc
> that you suggested in
> <http://lists.gnu.org/archive/html/bug-coreutils/2010-10/msg00038.html>.
> I just tried it now, and it does fix the perl warnings.
>
> The problem goes away if I do the build on a local disk, so it does
> seem to be specific to NFS.
>
> I reproduced the problem with the standard RHEL tools, i.e., without
> using the newer versions of GCC etc. that I normally use.
>
> When I strace the make, the bug goes away.  This indicates that it's
> timing-related, and may be hard to debug.
>
> Ooo! Ooo!  The "timing-related" in the previous paragraph made the
> light bulb go on in my head.  It's clock skew.  The NFS server's
> clock is a tiny bit ahead of the NFS client's clock, and so the output
> of "ls -l" contains the date, not the time.

Good catch!

> I don't see how clock-skew could also explain the test-rename bug noted in
> <http://lists.gnu.org/archive/html/bug-coreutils/2010-10/msg00050.html>,
> though.  That part of NFS is not supposed to care about clock skew.
>
> Anyway, the following combined patch fixes the ls-misc failure:
...
> Subject: [PATCH] tests: work around portability and clock-skew problems
>
> * tests/misc/ls-misc (push_ls_colors): Don't assume LS_COLORS
> is set.  This part of the fix is by Jim Meyering.
> (sl-dangle2, sl-dangle3, sl-dangle4, sl-dangle5): Don't assume
> that newly-created files will have time stamps in the past.  They
> might not, due to clock skew, if the file systems are remote.

Thanks!
I've pushed that.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.