# Re: Tests

Date: 2003-09-14 21:17:34 CEST

On Sun, 2003-09-07 at 15:04, D.J. Heap wrote:
> It's being applied to expected_output. Here's the relevant test output
> on Windows:
>
> EXPECTED STDOUT:
> Sending working_copies/trans_tests-6/foo
> Transmitting file data .
> Committed revision 3.
> ACTUAL STDOUT:
> Sending working_copies\trans_tests-6\foo
> Transmitting file data .
> Committed revision 3.
> EXCEPTION: SVNLineUnequal
> FAIL: trans_tests.py 6: committing eol-style change forces text send
>
> DJ
>
>

I've looked deeper into the three test failures I've been getting, and
the trans and update tests are just because of the path separator.

The trans_test can be fixed by fixing the _tweak_paths method which in
that file is only used by that test -- it appears that Subversion has
changed behavior slightly and is now sending back the local separator
when it did not used to.

The update test can be fixed by adding a _tweak_paths type method and
using it in the test that fails.

The question is, I guess, is this the way to handle it? Would a patch
doing that be accepted, or does someone have a 'vision' for how this
should be handled now that Subversion seems to be returning local
separators more often (which is a good thing, IMO)?

The commit test failure is slightly different -- the expected output
seems to be the absolute path to the changed file, but all that is
outputted is the filename (which is identical behavior as on Linux, as
far as I can see); so, why is the expected output a full path on
Windows and apparently not on Linux?

I was hoping to find more evidence of the Windows filesystem problem to
narrow in on it -- but these all appear to be simple test buglets from
what I can see. At this point, I'm out of ideas to try on narrowing in