On Wed, 2010-03-31 at 20:25 -0500, Eric Sandeen wrote:
> Dave Chinner wrote:
> > On Wed, Mar 31, 2010 at 06:38:23PM -0500, Eric Sandeen wrote:
> >> Right now if any of the _check_scratch_fs tests etc fail,
> >> the check script exits but with 0 status.
> >>
> >> This change will cause the status to be non-0 so we can detect
> >> the error.
> >>
> >> Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx>
> >
> > Wouldn't it be better to do:
> >
> > - [ $ok -eq 0 ] && exit 1
> > + if [ $ok -eq 0 ]; then
> > + status=1
> > + exit 1
> > + fi
> >
> > To catch all failures rather than just the fsck failure?
I see this patch is already committed to xfstests-dev.
I don't see the point of setting status=1 before exiting.
If what you meant was to set the status and then *return*
a non-zero value it would make sense.
How do you feel about updating your patch and committing
the result to the -dev tree?
If you change "exit" to "return" in both spots you
can consider it reviewed by me.
-Alex
PS Here is a script that shows what's going on.
#!/bin/bash
echo =============
status=1
echo top-level shell PID is $$
echo initial status is $status
bash <<!
echo -------------
echo sub-shell PID is \$$
echo sub-shell initial status is \$status
foo() { echo foo PID is \$$; status=2; return 3; }
bar() { echo bar PID is \$$; status=4; exit 5; }
echo -------------
status=0
foo # This is run in the current (sub) shell
echo foo result is \$?
echo post-foo status is \$status
echo current shell PID is \$$
echo -------------
status=0
bar # This will exit to the top-level shell
# And the next three will not get executed
echo bar result is \$?
echo post-bar status is \$status
echo current shell PID is \$$
!
R=$?
echo -------------
echo final result is $R
echo final shell PID is $$
echo final status is $status
echo =============
exit 0
> sure, that sounds better. I'll check that version in w/ your SOB
> unless I hear otherwise :)
>
> Thanks,
> -Eric
>
> >> ---
> >>
> >> (aside - we could make it 2 instead of 1, so that a calling script
> >> could fsck and continue ... thoughts?)
> >
> > I think if a test corrupts a filesystem, the test run should stop
> > so the failure can be analysed without needing to reproduce it
> > again...
> >
> > Cheers,
> >
> > Dave.
>
> _______________________________________________
> xfs mailing list
> xfs@xxxxxxxxxxx
> http://oss.sgi.com/mailman/listinfo/xfs
|