Thanks for your report. Le 03/20/17 à 13:32, Santiago Vila a écrit : > ./test > .missing b'b/c' on destination side > missing b'b' on destination side > ... create new inode for b'a' > b'a/c': missing > F > ====================================================================== > FAIL: test2 (__main__.Test) > ---------------------------------------------------------------------- > Traceback (most recent call last): > File "./test", line 371, in test2 > self.assertListEqual([b'a', b'a/c'], list(self.fssync.check(b''))) > AssertionError: Lists differ: [b'a', b'a/c'] != [b'a/c'] > > First differing element 0: > b'a' > b'a/c' > > First list contains 1 additional elements. > First extra element 1: > b'a/c' > > - [b'a', b'a/c'] > + [b'a/c'] > > ---------------------------------------------------------------------- > Ran 2 tests in 0.382s
This error was recently reported by reproducible-builds (2017-02-04): https://tests.reproducible-builds.org/debian/history/fssync.html At that time, I already tried to reproduce it on my machine, without success. I suspected reprotest and tried it. Your report really confirms it's something else. > The bug should be reproducible with sbuild on a single CPU virtual machine, > provided you try enough times (as the failure happens randomly). OK, so I've just set up a single-CPU VM. Still unable to reproduce the failure. I don't think the number of cores matters here. The issue seems to be at the line 368 of 'test': os.rename(x + b'c', x + b'd') The test assumes that this changes the modification time of the folder pointed to by x. On my side, it's increased by a few ms. Maybe your /tmp folder use a FS with a very low time precision. I tried tmpfs and ext4. Julien
