-
-
Notifications
You must be signed in to change notification settings - Fork 42
Conversation
…w files from the report (not making a huge impact on the overall coverage score, but the rest are things that should be tested...
Weird....travis-ci tests mysteriously failed on Python 2.6. This should have nothing to do with this PR--probably something in the build changed. I can't reproduce locally... |
The appveyor test on Python 2.6 worked fine too... |
Okay, I was able to reproduce this. It actually really does fail on Python 2.6 when adding the |
Interesting...
|
Okay, interesting, and subtle. This has to do with recent updates to pytest-cov that enables it, in a clever but hackish way, to collect coverage data on code run in subprocesses. However, this hack breaks in a bad way if the path you give to the coveragerc file is a relative path, and your test changes directories before running the subprocess. The breakage actually occurs very early on in the Python interpreter startup (for the subprocess) in site.py--Python 2.7 and up are able to ignore the exception and resume startup, but Python 2.6's site.py is less flexible and it will crash the whole interpeter startup if an unhandled exception occurs while processing a .pth file. I'll report this to pytest-cov. As a workaround, for now, I think I can pass a full path in for the config file. |
…because*. pwd should work, instead...
Add a coveragerc to tweak the coverage report somewhat. Excludes a few files from the report (not making a huge impact on the overall coverage score, but the rest are things that should be tested...