[PATCH] tests: fix builtin module test on pypy

Pierre-Yves David pierre-yves.david at ens-lyon.org
Fri Apr 1 07:09:56 UTC 2016



On 03/30/2016 02:01 PM, timeless wrote:
> On Wed, Mar 30, 2016 at 3:34 PM, Pierre-Yves David
> <pierre-yves.david at ens-lyon.org> wrote:
>>> diff -r ff0d3b6b287f -r 6057e466307e contrib/import-checker.py
>>> --- a/contrib/import-checker.py Tue Mar 29 12:29:00 2016 -0500
>>> +++ b/contrib/import-checker.py Wed Mar 30 21:31:45 2016 +0200
>>> @@ -182,6 +182,8 @@
>>>        yield 'builtins' # python3 only
>>>        for m in 'fcntl', 'grp', 'pwd', 'termios':  # Unix only
>>>            yield m
>>> +    for m in 'cPickle', 'datetime':
>>> +        yield m
>
> i have an alternative to this patch
>
>> The for above have a small comment explaining why these value make sense,
>> Can you add the same thing for the pypy on?
>
> instead of searching the current python and trying to guess what's "standardlib"
> i have hardcoded a list
> but it makes the py3 stuff work much better
> basically what we're doing is a heuristic, and it's really bad
> it doesn't work well w/ virtualenv, and it doesn't work well w/ py3,
> and it sounds like it doesn't work well w/ pypy :)
>
> http://patchwork.serpentine.com/patch/14184/
>
> note that as is, it doesn't actually "fix" pypy, but once people are
> comfortable with it, we could switch away from the heuristic we
> currently have to just using the list there (currently for py3) for
> all py flavors.

With part of your series apply, the pypy test is still necessary. I've 
taken the patch to help moving forward with pyp test fixing (and there 
is already such hack right next to the new code) We can clean that up 
with the rest when the time come.

-- 
Pierre-Yves David


More information about the Mercurial-devel mailing list