On Tue, 01 Oct 2019 at 23:01:14 +0100, Simon McVittie wrote: > The test in question asserts that we can: > > * read the modification time of a file, with microsecond precision > * set a modification time 50 microseconds later > * read the modification time back > * get the same modification time we set > > and the failure mode is that the file's new mtime is 1 microsecond older > than what we asked for.
I think this is because it's using a GDateTime API that involves floating point. I have a patch (replacing g_date_time_add_seconds() with an equivalent g_date_time_add() that adds microseconds) which I'll test and hopefully upload tomorrow. smcv