On 2020-10-12 14:23, Graham Inggs wrote:
On Mon, 12 Oct 2020 at 05:18, Drew Parsons <dpars...@debian.org> wrote:
Must have been a temporary package inconsistency?
All tests are passing now.
The autopkgtest looks flaky to me [1]. I see successes and failures
with the same triggers.
Looks like they might be generating random file names using random
numbers without guaranteeing uniqueness. So with random.randint(1, 100)
in play, there's a "1%" probability of name clash.
The exact problem is hard to pin down, i.e. the error message doesn't
clearly match a filename clash.
For instance a recent amd64 error is
ERROR:
test_write_readback.TestPythonMatlabFormat.test_dtype_structured_with_offsets_titles
----------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/lib/python3/dist-packages/nose/case.py", line 197, in
runTest
self.test(*self.arg)
File
"/tmp/autopkgtest-lxc.58_whuw2/downtmp/autopkgtest_tmp/tests/test_write_readback.py",
line 862, in test_dtype_structured_with_offsets_titles
np.dtype(desc).itemsize + random.randint(1, 100)
ValueError: title already used as a name or title.
But the indicated code line "np.dtype(desc).itemsize + random.randint(1,
100)" is just a random value for the field
desc_with_itemsize['itemsize']. What does it have to do with names? Not
clear why the indicated ValueError is triggered by this.
Drew
