Failing test for schemes extension

Mads Kiilerich mads at kiilerich.com
Wed Apr 7 06:50:42 CDT 2010


On 04/07/2010 01:39 PM, Michael Glassford wrote:
> On Apr 6, 2010, at 3:43 PM, Mads Kiilerich wrote:
>
>> Michael Glassford wrote, On 04/06/2010 09:31 PM:
>>>> Can someone confirm if this is a 2.6.5 issue?
>>>
>>> I reported below that I saw the same problem with Python 2.5.2, but I
>>> tried it again and don't see it there, so I must have made a mistake
>>> the first time. That makes it seem like it probably is a Python 2.6.5
>>> issue.
>>>
>>> For what it's worth, Python 3.1 behaves the same as Python 2.6.5 when
>>> you adjust for the fact that the urlparse module has been renamed to
>>> urllib.parse (i.e., the code snippet outputs "l:" instead of "l://").
>>
>> Introduced with
>> http://svn.python.org/view/python/branches/release26-maint/Lib/urlparse.py?r1=66717&r2=78235
>> to fix http://bugs.python.org/issue7904
>>
>> /Mads
>
> So that seems to leave two possible fixes:
> 1) Change the expected unit test results to match the new Python
> behavior (which means that the test will start failing for anyone with
> an older version of Python).
> 2) Change the implementation of Mercurial's url.hidepassword function to
> preserve the "//" if it really ought to be there.
>
> Which seems preferable?

That depends ... is it just a test failure where the code works 
correctly anyway? In that case I would rather filter the output to show 
the irrelevant difference.

Where else will this change in behavior show up?

Should we consider it a bug in Python which we have to work around and 
which they should use their time machine to fix?

/Mads


More information about the Mercurial-devel mailing list