GNOME Bugzilla – Bug 655114
Ensure top_builddir exists for all buildscripts
Last modified: 2011-07-30 19:25:34 UTC
Running 'jhbuild buildone gnome-utils': test -z "/opt/gnome/bin" || /bin/mkdir -p "/opt/gnome/_jhbuild/root-gnome-utils/opt/gnome/bin" /bin/sh ../libtool --mode=install /home/oxyde/.local/bin/install-check gnome-system-log '/opt/gnome/_jhbuild/root-gnome-utils/opt/gnome/bin' libtool: install: /home/oxyde/.local/bin/install-check gnome-system-log /opt/gnome/_jhbuild/root-gnome-utils/opt/gnome/bin/gnome-system-log make[4]: Nothing to be done for `install-data-am'. make[4]: Leaving directory `/home/oxyde/gnome/gnome-utils/logview' make[3]: Leaving directory `/home/oxyde/gnome/gnome-utils/logview' make[2]: Leaving directory `/home/oxyde/gnome/gnome-utils/logview' make[1]: Leaving directory `/home/oxyde/gnome/gnome-utils/logview' make[1]: Entering directory `/home/oxyde/gnome/gnome-utils' make[2]: Entering directory `/home/oxyde/gnome/gnome-utils' make[2]: Nothing to be done for `install-exec-am'. make[2]: Nothing to be done for `install-data-am'. make[2]: Leaving directory `/home/oxyde/gnome/gnome-utils' make[1]: Leaving directory `/home/oxyde/gnome/gnome-utils' I: Deleting toplevel .la file: u'/opt/gnome/_jhbuild/root-gnome-utils/opt/gnome/lib64/libgdict-1.0.la' Traceback (most recent call last):
+ Trace 227843
jhbuild.main.main(sys.argv[1:])
rc = jhbuild.commands.run(command, config, args, help=lambda: print_help(parser))
return cmd.execute(config, args, help)
return self.run(config, options, args, help)
return build.build()
error, altphases = module.run_phase(self, phase)
method(buildscript)
self.process_install(buildscript, self.get_revision())
buildscript.moduleset.packagedb.add(self.name, revision or '', destdir)
self._write_cache()
os.rename(tmp_dbfile_path, self.dbfile)
This is probably fallout from: commit cbbea64ac930b48945c63723b54e98a3b48b1512 Author: Frédéric Péters <fpeters@0d.be> Date: Tue Jul 19 13:26:01 2011 +0200 bot: Move packagedb only if top_builddir exists (it doesn't on build masters) The commit is incomplete; the packagedb will still try to write to the new directory. I don't understand why it doesn't exist; we create it in config.py:setup_env(). Do you have a $JHBUILD_PREFIX/_jhbuild directory?
fwiw that quick fix was made to get "jhbuild bot --start-server" working, I didn't have time to investigate.
(In reply to comment #2) > fwiw that quick fix was made to get "jhbuild bot --start-server" working, I > didn't have time to investigate. Are you on jhbuild master? A probable fix for this was (unintentionally) pushed in commit 6c485ba8. What confuses me is it did come before your commit. Is it likely then that your build server checkout was before that commit, you made an initial diagnosis, worked on the patch, then rebased and pushed? If you could try: git revert cbbea64ac930b48 In your buildbot and let me know if things still work, I'd appreciate it!
Also, Frederic, do you have a $JHBUILD_PREFIX/_jhbuild directory?
(In reply to comment #1) > Do you have a $JHBUILD_PREFIX/_jhbuild directory? Yes. I did repeat 'jhbuild buildone gnome-utils' immediately after the traceback and was unable to reproduce. Not sure if this is important - I did have multiple jhbuilds running at the same time, building separate modules within the same moduleset.
This is not "my" build master but build.gnome.org; I am pretty sure it was pointing to the latest commit as I had to fix the import of ElementTree just before.
(In reply to comment #5) > (In reply to comment #1) > > Do you have a $JHBUILD_PREFIX/_jhbuild directory? > > Yes. > > I did repeat 'jhbuild buildone gnome-utils' immediately after the traceback and > was unable to reproduce. Not sure if this is important - I did have multiple > jhbuilds running at the same time, building separate modules within the same > moduleset. Ohhh...so, one issue with this is that doing so will corrupt the packagedb.xml file. I'm not sure how this would cause your problem, but in the future when we clean up stale files, it *will* explode. This is effectively bug 312910 , I've updated it with a comment.
I pushed a revert of that previous commit. From looking at this bug again, I'm pretty sure it is bug 312910 - if two jhbuild instances try to update packagedb.xml at the same time, it's possible they both write out packagedb.xml.tmp, but the just one wins; then on rename() the other one will fail. *** This bug has been marked as a duplicate of bug 312910 ***