Re: [PATCH v2] fstests: faster group file creation

[Dave Chinner <david@fromorbit.com>]

On Fri, May 06, 2022 at 03:13:11PM +0800, Zorro Lang wrote:
> On Fri, May 06, 2022 at 03:10:17PM +1000, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> > 
> > We don't need to execute every test just to check it's groups are
> > valid. Just grab all the groups with grep, pull out the unique ones,
> > then check them.
> > 
> > This also avoids the problem of editor swap files being present in
> > the test directory and breaking the build because they are not
> > executable.
> > 
> > Building on a clean, already built tree so it only builds the
> > group lists:
> > 
> > $ time make
> > ....
> > Building udf
> >  [GROUP] /home/dave/src/xfstests-dev/tests/udf/group.list
> > Building xfs
> >  [GROUP] /home/dave/src/xfstests-dev/tests/xfs/group.list
> > 
> > real    0m36.917s
> > user    0m15.032s
> > sys     0m26.219s
> > $
> > 
> > Patched:
> > 
> > $ time make
> > ....
> > Building udf
> >  [GROUP] /home/dave/src/xfstests-dev/tests/udf/group.list
> > Building xfs
> >  [GROUP] /home/dave/src/xfstests-dev/tests/xfs/group.list
> > groups "frobnozzle" not mentioned in documentation.
> > gmake[3]: *** [../../include/buildgrouplist:8: group.list] Error 1
> > gmake[2]: *** [../include/buildrules:31: xfs] Error 2
> > gmake[1]: *** [include/buildrules:31: tests] Error 2
> > make: *** [Makefile:51: default] Error 2
> > 
> > real    0m1.751s
> > user    0m0.863s
> > sys     0m1.067s
> > 
> > $
> > 
> > Just a little bit faster, and as you can see that it still detects
> > groups that are not documented. There was also an open .001.swp file
> > in the XFS directory and that doesn't throw a failure anymore,
> > either.
> > 
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> 
> Looks good to me. It's much faster than before, especially when rebuild for
> small changes.

Yup, my run scripts always run make before starting a run, so when
I'm iterating single tests for failure analysis, this knocks 20-30s
out of the cycle time. check starting up is now the thing that is
really slow - it's still taking around 10s to get to the first
test these days....

> Reviewed-by: Zorro Lang <zlang@kernel.org>

Thanks!

-Dave.
-- 
Dave Chinner
david@fromorbit.com