Bug 10783

I can fix it by making specifying output directory mandatory.
However, it can hurt users who run wic without -o option and expect results in /var/tmp/wic

Do you have a better idea how to fix this?

what about using mktemp?

i.e. write to /var/tmp/wic-XXX

via mktemp -d --tmpdir=/var/tmp wic-XXXX

(In reply to comment #2)
> what about using mktemp?
> 
> i.e. write to /var/tmp/wic-XXX

If /var/tmp itself is owned by someone else that'll still break.

I'm not sure how to reconcile this with "existing users expect /var/tmp". We do make backwards-incompatible changes elsewhere from time to time though, if we have to warn people now and then change it in a couple of releases that would be fine. We'd also naturally document it in the migration guide when it does happen.

please, review the fix: http://lists.openembedded.org/pipermail/openembedded-core/2017-January/131679.html

moved to M3. we're late for M2.

The patch has been merged to master.

(In reply to comment #6)
> The patch has been merged to master.

There's some problem with it:

00:40:53.815 [build_intel-corei7-64] | During handling of the above exception, another exception occurred:
00:40:53.815 [build_intel-corei7-64] | 
00:40:53.815 [build_intel-corei7-64] | Traceback (most recent call last):
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/wic", line 330, in <module>
00:40:53.815 [build_intel-corei7-64] |     sys.exit(main(sys.argv[1:]))
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/wic", line 325, in main
00:40:53.815 [build_intel-corei7-64] |     return hlp.invoke_subcommand(args, parser, hlp.wic_help_usage, subcommands)
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/lib/wic/help.py", line 95, in invoke_subcommand
00:40:53.815 [build_intel-corei7-64] |     subcommands.get(args[0], subcommand_error)[0](args[1:], usage)
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/wic", line 258, in wic_create_subcommand
00:40:53.815 [build_intel-corei7-64] |     options.compressor, options.bmap, options.debug)
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/lib/wic/engine.py", line 198, in wic_create
00:40:53.815 [build_intel-corei7-64] |     crobj.main(cmdline)
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/lib/wic/creator.py", line 125, in main
00:40:53.815 [build_intel-corei7-64] |     return self._subcmds[pname](options, *args[1:])
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/lib/wic/plugins/imager/direct_plugin.py", line 101, in do_create
00:40:53.815 [build_intel-corei7-64] |     creator.cleanup()
00:40:53.815 [build_intel-corei7-64] |   File "/srv/jenkins/workspace/builder-slot-0/openembedded-core/scripts/lib/wic/imager/direct.py", line 395, in cleanup
00:40:53.816 [build_intel-corei7-64] |     for fname in os.listdir(self.workdir):
00:40:53.816 [build_intel-corei7-64] | FileNotFoundError: [Errno 2] No such file or directory: '/tmp/wicxbm1fyav'

I guess that
tempfile.mktemp(prefix='wic')

creates a file, where we actually need directory?

I think Mikko copied (2 comments above) a piece which is not most relevant,
the immediate issue reported in build was:

mkdosfs: unable to create /tmp/wicxbm1fyav/rootfs_primary_uefi.1.vfat

and that leads to my guess that tmpdir was not created

correcting my comment, tempfile.mktemp() seems to create nothing.
Even by doc, this function only returns temporary file name,
without creating, and is also deprecated.
mkdtemp should be used to create a directory,
this has fixed issue in local build

Should be fixed by this patchset: http://lists.openembedded.org/pipermail/openembedded-core/2017-February/132189.html

Please, review.