On Tue, Mar 28, 2017 at 12:42 PM, Eric Sandeen <sandeen@xxxxxxxxxxx> wrote: > I am (literally) in the woods this week so can't really comment at length, but generally: > > xfs_io is useful to facilitate testing at times, but usually doesn't have tests built into the tool itself. > David, Eric being the maintainer, so he gets to decide, but he is being somewhat subtle. There is no precedence AFAIK to tests within xfs_io. Instead of comparing to stat with -c flag, you can make sure that output of xfs_io -c 'statx -c' is fully compatible to output of xfs_io -c 'stat -v' and do the test in scripts. > And I think you are right that it is a poor fit for some of the testing you'd like to do. > There is a src/ dir in xfstests for specialized C tests which get called by the test harness; that might also be a reasonable option. > That's actually the shortest path for you and there are quite a few tests in xfstests that took this path. It should be easy for you to copy & paste one of them. That's not instead of adding xfs_io statx - just for tests that are not natural to do with xfs_io. > Thanks, > Eric > >> On Mar 28, 2017, at 9:41 AM, David Howells <dhowells@xxxxxxxxxx> wrote: >> >> Note that the intention is to put the testing of the syscall parameter >> handling into LTP, along with testing of the symlink following and dirfd usage >> since xfstests seems unsuitable for this. >> >> David >
