On Fri, 15 Apr 2011, Thomas Koch wrote:
I'd expect anyone running on Windows to see these test failures.
Andi..
So what do you think about this issue - can we ignore this or claim it's a
windows bug or hope that 'just' the test code is wrong?
Here is what I think: I think that there is a bug. I don't know where but it
happens only on Windows. I tried to fix it in the past but I don't run
Windows much so I didn't try that hard. For a more definitive answer or
explanation, someone with a stronger itch on Windows is going to have to
step in. Until then, and it's been years, that bug shall remain a bit of a
mystery. Given that no one has reported a more general bug in the same area
of PyLucene indexing on Windows, it could mean that either there are no
users (unlikely) or that this bug is just with the unit test (more likely).
I'd suggest to at least apply the mentioned fix (i.e. uncomment the close
call in test_PyLucene) to make this test run on windows. Of course someone
should confirm this doesn't break the tests on Linux (or other OS)... just
my thoughts.
I uncommented line 260 in test_PyLucene.py as you suggested and tests seem
to still pass on Mac OS X. I didn't try Linux or Solaris. In fact, I think
that this commented line very much looks like an earlier attempt of mine at
elucidating this bug that I didn't clean up properly.
Checked into rev 1092774 of branch_3x.
Andi..
