GNU bug report logs - #10975
'make dist-gz dist-xz' fails

Previous Next

To reply to this bug, email your comments to 10975 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Olaf Lenz <olenz <at> icp.uni-stuttgart.de>
To: bug-automake <at> gnu.org
Subject: 'make dist-gz dist-xz' fails
Date: Thu, 08 Mar 2012 10:22:32 +0100

[Message part 1 (text/plain, inline)]

Hi!

I'm using Automake 1.11.3 and Autoconf 2.68.

Creating two different distribution packages on the command line of make
fails:

> make dist-bzip2 dist-xz
[clip]
tardir=espresso-3.1.0-1-gf2e4716-dirty && ${TAR-tar} chof - "$tardir" |
BZIP2=${BZIP2--9} bzip2 -c >espresso-3.1.0-1-gf2e4716-dirty.tar.bz2
if test -d "espresso-3.1.0-1-gf2e4716-dirty"; then find
"espresso-3.1.0-1-gf2e4716-dirty" -type d ! -perm -200 -exec chmod u+w
{} ';' && rm -rf "espresso-3.1.0-1-gf2e4716-dirty" || { sleep 5 && rm
-rf "espresso-3.1.0-1-gf2e4716-dirty"; }; else :; fi
tardir=espresso-3.1.0-1-gf2e4716-dirty && ${TAR-tar} chof - "$tardir" |
XZ_OPT=${XZ_OPT--e} xz -c >espresso-3.1.0-1-gf2e4716-dirty.tar.xz
tar: espresso-3.1.0-1-gf2e4716-dirty: Cannot stat: No such file or directory
tar: Exiting with failure status due to previous errors
if test -d "espresso-3.1.0-1-gf2e4716-dirty"; then find
"espresso-3.1.0-1-gf2e4716-dirty" -type d ! -perm -200 -exec chmod u+w
{} ';' && rm -rf "espresso-3.1.0-1-gf2e4716-dirty" || { sleep 5 && rm
-rf "espresso-3.1.0-1-gf2e4716-dirty"; }; else :; fi

The problem seems to be a race condition.
"make dist-bzip2" populates the distribution dir, tars it and deletes it
after 5 seconds. During that time "make dist-xz" has started and noticed
that the distdir already exists. Unfortunately, it is deleted while
dist-xz is still packing.

Olaf

-- 
Dr. rer. nat. Olaf Lenz
Institut für Computerphysik, Pfaffenwaldring 27, D-70569 Stuttgart
Phone: +49-711-685-63607

[olenz.vcf (text/x-vcard, attachment)]

[signature.asc (application/pgp-signature, attachment)]

Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Tim Landscheidt <tim <at> tim-landscheidt.de>
To: bug-automake <at> gnu.org
Subject: Re: bug#10975: 'make dist-gz dist-xz' fails
Date: Fri, 09 Mar 2012 19:53:10 +0000

Olaf Lenz <olenz <at> icp.uni-stuttgart.de> wrote:

> [...]
> The problem seems to be a race condition.
> "make dist-bzip2" populates the distribution dir, tars it and deletes it
> after 5 seconds. During that time "make dist-xz" has started and noticed
> that the distdir already exists. Unfortunately, it is deleted while
> dist-xz is still packing.

For a possible workaround, see
<URI:http://permalink.gmane.org/gmane.comp.sysutils.automake.bugs/4940>.
I agree that it would be nice to fix this properly.

Tim

Message #11 received at 10975 <at> debbugs.gnu.org (full text, mbox):

From: Stefano Lattarini <stefano.lattarini <at> gmail.com>
To: 10975 <at> debbugs.gnu.org
Subject: Re: bug#10975: 'make dist-gz dist-xz' fails
Date: Fri, 09 Mar 2012 21:36:42 +0100

[Message part 1 (text/plain, inline)]

Hi Olaf, Tim.

On 03/09/2012 08:53 PM, Tim Landscheidt wrote:
> Olaf Lenz <olenz <at> icp.uni-stuttgart.de> wrote:
> 
>> [...]
>> The problem seems to be a race condition.
>> "make dist-bzip2" populates the distribution dir, tars it and deletes it
>> after 5 seconds.
>
Nope -- it deletes it *immediately*, and then re-tries to delete it after 5
seconds if the first deletion fails.  This is done in an attempt to accommodate
systems like MinGW where an "active" directory cannot be deleted:

 <http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10470>

 During that time "make dist-xz" has started and noticed
>> that the distdir already exists. Unfortunately, it is deleted while
>> dist-xz is still packing.
> 
> For a possible workaround, see
> <URI:http://permalink.gmane.org/gmane.comp.sysutils.automake.bugs/4940>.
> I agree that it would be nice to fix this properly.
> 
> Tim
> 
I've reproduced the bug with the attached test case (committed to maint).

I've also verified that the bug was also present in automake 1.11.2 and
automake 1.11.1, so this is not a regression introduced by the "sleep 5"
change referenced above.

For the moment, and until the bug is properly fixed, my advice is: don't
run multiple "dist-*" targets in the same make invocations.

Regards,
  Stefano

[0001-coverage-expose-automake-bug-10975.patch (text/x-diff, attachment)]

Previous Next

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