On 4/20/19 3:53 AM, Eryu Guan wrote: > On Thu, Apr 18, 2019 at 12:01:31PM -0500, Eric Sandeen wrote: >> With the xfs_repair fixes for how the link count on root inodes >> are handled, xfs/033 started failing because we didn't have to >> re-fix the link count at the end of the repair run. >> >> Capturing and comparing all of the xfs_repair output seems somewhat >> fragile in general. It should suffice to just run repair, then >> check that it succeeded with _scratch_xfs_repair, rather than >> comparing against all the detailed output. So, this patch does that. >> Full repair output is still saved to the $seqres.full file. >> >> Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx> >> --- >> >> diff --git a/.gitignore b/.gitignore >> index c13fb713..2482dd61 100644 >> --- a/.gitignore >> +++ b/.gitignore >> @@ -251,7 +251,6 @@ >> /dmapi/src/suite2/src/test_rights >> >> # Symlinked files >> -/tests/xfs/033.out >> /tests/xfs/071.out >> /tests/xfs/096.out >> >> diff --git a/common/repair b/common/repair >> index c94939c9..5e53f5f2 100644 >> --- a/common/repair >> +++ b/common/repair >> @@ -114,7 +114,7 @@ _check_repair() >> _scratch_unmount >> >> _zero_position $value "$structure" >> - _scratch_xfs_repair 2>&1 | _filter_repair >> + _scratch_xfs_repair 2>&1 | _filter_repair >> $seqres.full > > Looks like this will break ohter tests that depend on the output of > _check_repair, e.g. xfs/030 and xfs/148 (xfs/031 and xfs/149 also call > _check_repair, but they have their local definition of _check_repair) Whoops sorry, I should have caught that. Thanks, -Eric
