Hi,
Now up and working with the latest pytest (think only distributed
testing was an issue). It is still backward compatible with older
versions of pytest (before the split).
Funcarg is at module level so visible with py.test --funcargs.
pth file more robust since it doesn't complain if cov-core got deleted
and pth file left.
More doc for coverage config file since couple of people have asked questions.
I moved the doc to the docstring in the module. I see that the
website just points to the external plugins at pypi, so that means the
doc can't get out of sync right?
:)
On 20 March 2011 22:21, holger krekel <hol...@merlinux.eu> wrote:
> Hi Meme,
>
> On Sun, Mar 20, 2011 at 22:00 +1100, meme dough wrote:
>> Hi,
>>
>> Currently looking at fixing pytest-cov to work with latest pytest. I
>> see two issues so far.
>>
>> 1. How to determine if non distributed, or distributed master, or
>> distributed slave.
>>
>> This used to look at the session class Session = Central (non
>> distributed), DSession = distributed master, and SlaveSession =
>> distributed slave.
>>
>> So now I'm thinking look at
>> session.config.pluginmanager.hasplugin('dsession') which means
>> distributed master. For distributed slave check if config.slaveinput
>> is there which means distributed slave. If none of those then must be
>> non distributed.
>
> I guess that's ok. There is no "official" way at the moment.
>
>> Is there a better way to determine at pytest_sessionstart if the
>> process is central (non distributed), distributed master or
>> distributed slave?
>>
>> 2. Need to know topdir.
>>
>> This used to use config.topdir to know what the root dir is. This is
>> needed since distributed testing may be co-located or not. If not
>> then have to rewrite all source paths back to what the master is set
>> to so that all coverage files combine correctly even if have many
>> separate tx envs on diff machines on diff paths.
>>
>> Has this moved somewhere? I see config.fspath, is that it? I look
>> more I just got to this bit.
>
> No, there now is a "looponfailroots" (see --help) and distributed
> testing uses "rsyncroots". The problem is that "topdir" was a bit
> of an artifical concept.
>
> Maybe it helps, that item.nodeid contains a "::" separated test id where
> the first part is a path relative to the current directory and
> normalized to use "/" as separator.
>
> best,
> holger
>
_______________________________________________
py-dev mailing list
py-dev@codespeak.net
http://codespeak.net/mailman/listinfo/py-dev
