C. Michael Pilato wrote:
> Lieven Govaerts wrote:
>> C. Michael Pilato wrote:
>>> So, I'm noticing something odd as of late in our test suite. Given a
>>> merge test that expects some bit of output and gets a different bit,
>>> an error is raised (as expected). But if *no* output is generated,
>>> no error is raised. That is striking me as really, really bad.
>>
>> Yep, that seems to be the case.
>>>
>>> I don't know the cause, and I'm busy with other semi-related tasks
>>> right now, but wanted to let folks know, because the reliability of
>>> our test suite is no small matter of importance (as I'm sure you'd
>>> all agree).
>>>
>>
>> I find that attached patch solves the issue:
>>
>> ------------------------------------------------------------------------
>> Index: subversion/tests/cmdline/svntest/verify.py
>> ===================================================================
>> --- subversion/tests/cmdline/svntest/verify.py (revision 29217)
>> +++ subversion/tests/cmdline/svntest/verify.py (working copy)
>> @@ -131,6 +131,12 @@ class ExpectedOutput:
>> all_lines_match_re = True
>> else:
>> all_lines_match_re = False
>> +
>> + # If a regex was provided assume that we actually require
>> + # some output. Fail if we don't.
>> + if len(actual) == 0:
>> + return False
>> +
>> for i in range(0, len(actual)):
>> if self.match_all:
>> if not self.is_equivalent_line(expected_re, actual[i]):
>> -------------------------------------------------------------------------
>>
>> Basically here we loop over all actual output lines and try to match
>> them with either all or any (depends on match_all setting) expected
>> output line(s). If there are no actual output lines we don't loop and
>> return True or False depending on what was specified for match_all.
>>
>> I assume that once we pass a regular expression as expected output, we
>> require at least some actual output. I'm running the test suite now,
>> for now it doesn't seem to have any negative consequences.
>>
>> Lieven
>
> +1 to commit, pending success of the test suite.
>
Committed in r29220. I found one failure in merge test 62, but it looks
like this fix actually shows that the test should have been failing
after all.
Lieven
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-02-07 11:28:27 CET