utimensat/08.t tests 5&6 fail on ext4 on Linux 4.4.103 + 4.10.0 (pass on Linux 4.4.0) about pjdfstest

[asomers]

Probably has nothing to do with #24. This test deals with the utimensat system call. It's got nothing to do with long pathnames. The problem is apparently that utimensat returned 0 but didn't do anything. It's probably a kernel bug. Or there might be some kind of security setting on your computer that prevents utimensat from backdating the time when run by a non-root user. Or maybe your filesystem only supports 1-second granularity. In the latter case, you should submit a PR to pjdfstest fixing the specification (see chflags/12.t for an example of specifications that differ between filesystems.

[s-m-e]

I assumed #24 and #25 to be related because they are the only two failures I see - but similarly on two totally different Linux distributions. The strange things though is that they both pass on Travis' Ubuntu derivative, which also uses ext4 and almost the same kernel version as on my openSUSE system (4.4.x).

[s-m-e]

@asomers Simple explanation: mke2fs reduces its default (and maximum allowed) inode size to 128 bytes when generating an ext4 filesystem (without journal) into a space of 128 MBytes or smaller. However, a minimum of 256 bytes per inode is required for storing atime and mtime with nano-second granularity. It is irritating though that utimensat does not indicate any sort of problem in this scenario.

[asomers]

Ok, this sounds like a bug in ext4, not pjdfstest.