On 1/15/20 10:21 AM, Darrick J. Wong wrote: > On Wed, Jan 15, 2020 at 07:08:02AM -0800, Christoph Hellwig wrote: >> Hi Jan and Baihua, >> >> the xfs/191 test case has been failing for me basically since it >> was added. Does it succeed for anyone with an upstream kernel >> and xfsprogs? > > It never succeeds here. <Jan has left Red Hat> yeah I think this has always been a mess, not sure why it took so long to highlight it. Seems like a combination of WTF? and general rot. > --D > >> Here is my diff between the golden and the actual >> output: > >> --- /root/xfstests/tests/xfs/191-input-validation.out 2016-09-21 20:34:14.961574921 +0000 >> +++ /root/xfstests/results//xfs/191-input-validation.out.bad 2020-01-15 15:05:25.580935340 +0000 >> @@ -1,2 +1,13 @@ >> QA output created by 191-input-validation >> silence is golden supposedly these should fail, but they pass: >> +pass -n size=2b /dev/vdc >> +pass -d agsize=8192b /dev/vdc >> +pass -d agsize=65536s /dev/vdc I don't know why it wouldn't be valid to use block & sector units for these values. But this whole mkfs reworking discussion was so long ago :( >> +pass -d su=0,sw=64 /dev/vdc not sure why this passes >> +pass -d su=4096s,sw=64 /dev/vdc >> +pass -d su=4096b,sw=64 /dev/vdc s & b suffixes seem to be ignored here, which seems like a bug. >> +pass -l su=10b /dev/vdc again this seems to DTRT. according to the test, these should pass, but they fail: >> +fail -n log=15 /dev/vdc uh, "-n log=" isn't even a mkfs option so of course it fails >> +fail -r rtdev=/mnt/test/191-input-validation.img /dev/vdc this fails because reflink is default now >> +fail -r size=65536,rtdev=/mnt/test/191-input-validation.img /dev/vdc ditto >> +fail -i log=10 /dev/vdc -i log is not a valid option either so of course it fails. My first thought was to nuke it all, but I think the test could be salvaged by anyone motivated to do so. -Eric
