On Mon, May 9, 2016 at 3:30 PM, Apollon Oikonomopoulos <[email protected]> wrote: > On 12:17 Sat 07 May , László Böszörményi wrote: >> Sorry for the slow reaction. The FTBFS reason is the package test >> case. It's threaded and the sequence is confused by the file timestamp >> resolution of Hurd and kFreeBSD architectures. It's one second only >> (Linux has millisecond resolution). > > Actually, it's ext4 (and btrfs) that has nanosecond resolution. At least > ext2 and ext3 have 1 second timestamp resolutions as well, so we can't > really rely on the platform being Linux or not. OK, I stand corrected.
>> Question is, how should it be handled? Disable self-tests, ignore the >> results or maybe remove mongodb from the mentioned architectures? > > IMHO, we should just disable the test suite on affected architectures. > Have you checked if the test suite of 2.6 runs (IIRC the test suite > harness has changed significantly after 2.4)? What I did to date is testing 2.4.14 in Sid on kFreeBSD/amd64. Run the test suite ten times - always failed. I've seen three kind of problems, all related to threading. But of course, I'm not a MongoDB developer, I just write what I saw. Good point, didn't try 2.6 yet. Still, I don't think the test suite should be disabled - better to run it and ignore the result on kFreeBSD / Hurd. Cheers, Laszlo/GCS
