xfs
[Top] [All Lists]

Re: xfstests: change directory to / before _cleanup_testdir in test 135

To: Boris Ranto <branto@xxxxxxxxxx>
Subject: Re: xfstests: change directory to / before _cleanup_testdir in test 135
From: Alex Elder <aelder@xxxxxxx>
Date: Tue, 08 Mar 2011 14:14:13 -0600
Cc: xfs <xfs@xxxxxxxxxxx>
In-reply-to: <1299590960.12867.102.camel@xxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <1299511558.3183.115.camel@xxxxxxxxxxxxxxxxxxxxxxxxxx> <1299528749.2578.821.camel@doink> <1299590960.12867.102.camel@xxxxxxxxxxxxxxxxxxxxxxxxxx>
Reply-to: aelder@xxxxxxx
On Tue, 2011-03-08 at 14:29 +0100, Boris Ranto wrote:
> On Mon, 2011-03-07 at 14:12 -0600, Alex Elder wrote:
> > On Mon, 2011-03-07 at 16:25 +0100, Boris Ranto wrote:
> > > Nfs tries to umount $testdir in _cleanup_testdir function. The test 135 
> > > calls the function from directory $SCRATCH_MNT that is equal to $testdir 
> > > (at least for nfs). The umount will therefore fail causing the test to 
> > > fail due to the output mismatch.
> > > 
> > > This simple patch fixes the issue for me.
> > > 
> > > Signed-off-by: Boris Ranto <branto@xxxxxxxxxx>
> > 
> > This looks OK to me.  Most other tests do this chdir
> > in their cleanup function.
> > 
> > I did a quick scan and found that test 126 may suffer the
> > same problem.  Can you check this?  We could include the
> > fix for both tests in the same commit.
> > 
> 
> Yes, the test needs cd /, too. Actually the test 126 also does double
> umount thanks to the _cleanup before exit and the trap command. So the
> removal of the call of the _cleanup function before exit is necessary,
> too.

I will commit this patch, fixing the problems in 126 and 135.
I'll massage the commit message a bit to fit what it does,
and the your discussion here.

I'm not going to do anything (at the moment) about the
double mount/umount issues you mention.

Thanks for submitting the fixes.

                                        -Alex

> > 
> > It also looks to me like tests 069, 089 might have a
> > similar issue if they get interrupted.
> 
> Yes, that's also true, if the tests are interrupted and then immediately
> run again the umount will fail due to the processes in background but
> I'm not sure whether it is worth fixing (and if there is a good way to
> fix it).
> I've had bigger problems with double mount/umount in several tests
> (namely: double umount - 124, 128, double mount - 129, 130). The double
> umounts occur due to the use of trapped _cleanup and umount $SCRATCH_MNT
> in the tests. The double mounts occur due to the use of _setup_testdir
> and _scratch_mount (both of them mount $SCRATCH_MNT in the nfs case).
> The problem is that I'm not sure how to fix these without any change in
> the behaviour.
> 
> > 
> >                                     -Alex
> > 
. . .

<Prev in Thread] Current Thread [Next in Thread>