Restores the coveralls coverage by changing the coveralls rc file
Limits coverage to the single mainstream run - Python 2.7 with NumPy 1.7 and SciPy 0.12
Test coveralls issue removing if, then
Fixed coveragerc to allow anythin in */statsmodels/*
Changes for single coverage run
Merge remote-tracking branch 'remotes/upstream/master' into travis-mi…
@josef-pkt @jseabold This should fix the two issues related to coveralls identified in the previous PR.
we have to wait a bit to hear back from travisci and coveralls, there are 3 testruns in the queue
as reminder: if there are no code changes, e.g. only docs, then we can also skip a test run [skip ci]
Coverage is up 3% - presumably due to matplotlib testing.
A hair away form 80 now.
Change test condition for coverage
Fix for missing font warning
Looks like it is working as expected now.
The repo root information in coveralls needs to be changed to reflect the changed system path.
I updated the path on coverall.io
(I haven't checked yet if we lost older files, but that doesn't matter much)
Yes, switching the path messes up the old coveralls reports
I don't think anyone is looking at old coverage files, should be just a short transition problem.
do you know why we don't get coveralls.io emails and comments anymore? Did you change a setting? Or did coveralls.io turn it off because we complained about incorrect messages?
looks good to merge
I have been getting them occasionally in the past couple of days from builds using the old travis scripts. I do not get them for all builds that submit coverage though.
I didn't change anything. I don't know if they fixed the incorrect stuff or what. I don't pay attention to the coveralls stuff because it's very low signal to noise ratio given all the problems.
Simplification using globals
Coverage remained the same when pulling b7e91ca on bashtage:travis-miniconda into d991e25 on statsmodels:master.
Ok, we do get the coverage messages (again)
Can I hit the merge button?
Fine by me (I didn't look). Any chance on marking knownfail or fixing the GMM test failure noise?
One more build to simplify things. I also mark known fails.
BTW, I have emails on. These were off before, any preferences here?
Marked known fails
Suppressed noise from matplotlib about .matplotlibrc location
Github seems to be having problems, but it is passing...and so is complete. Ignores 3 SciPy 13 failures now.
I have restarted by githib-induced errors:
which are passing now.
Final modifications to simplify coverage choices
This really is finished now.