Fixes #653 use deprecated_call as context_manager

[chiller]

No description provided.

[RonnyPfannschmidt]

the simplefilter happening outside of a context manager affects global state

[chiller]

is moving warnings.simplefilter('always') inside WarningsChecker.__enter__ a solution?

[RonnyPfannschmidt]

it already does that, see the line before return

[chiller]

Ah, I've found that, currently am trying to figure out why it only works if append=True is unset:
WarningsRecorder.__enter__:warnings.simplefilter('always') - works fine
WarningsRecorder.__enter__:warnings.simplefilter('always', append=True) - the existing code, breaks my PR tests

[RonnyPfannschmidt]

append decides whether the filter is inserted at the end or the beginning in the warnings filter list

[chiller]

Yes, but I have not yet figured out why that would change the outcome of the deprecated_call

[chiller]

I believe this is because in a list of filters - the first one that matches will apply, and upon inspecting the filters (before the always is appended):

[('ignore', None, <type 'exceptions.DeprecationWarning'>, None, 0), 
('ignore', None, <type 'exceptions.PendingDeprecationWarning'>, None, 0),
 ('ignore', None, <type 'exceptions.ImportWarning'>, None, 0),
('ignore', None, <type 'exceptions.BytesWarning'>, None, 0)]

I see now why appending always will never work. The question is : is there a point in appending this filter in the first place(instead of prepending) since it is possible to have one that will match before it(being shadowed).

[RonnyPfannschmidt]

@chiller i think we need a option to clear the errors list or to insert at the beginning for the recorder
if you are a pyconuk, perhaps talk with @flub

[chiller]

I am, will do

[chiller]

Discussed with @flub that it doesn't make sense to append the filter

[RonnyPfannschmidt]

good investigative work

instead of merging master, please use a rebase and a force push so we can keep the main history less eratic

[chiller]

ok, done

[hpk42]

please retarget to "features" branch -- "master" is since today meant for bugfixes only. sorry for the inconvenience.

[RonnyPfannschmidt]

we can manually merge into feature

[chiller]

@RonnyPfannschmidt I didn't understand your comment, should I branch my code from feature instead and create a new PR? I am happy to do that. Is there maybe an easier way?

[RonnyPfannschmidt]

@chiller i'll merge myself in a bit (github does not have an ui to make the same pull request against different branch without a mess)

[RonnyPfannschmidt]

it was merged into the features branch

[chiller]

thanks

[coveralls]

Coverage increased (+0.002%) to 92.596% when pulling 9f77a85 on chiller:653-deprecated-context-manager into c30eafa on pytest-dev:master.