MRG: Omit 2.6, fix image_tester deprecation #1353
Conversation
|
@larsoner Somehow the bundled |
larsoner
changed the title
|
Lots of CI fixes in this one. Ready for review/merge from my end. No need to look at |
|
(This drops 2.6 support and adds 3.6 support BTW.) |
| with TestingCanvas() as c: | ||
| if os.getenv('TRAVIS', 'false') == 'true' and \ | ||
| c.app.backend_name == 'pyqt4': | ||
| # TODO: Fix this (issue #1042 |
There was a problem hiding this comment.
#1042 mentioned python26. Is this still an issue?
|
@larsoner First of all, great work! I tried to follow all your of changes. I didn't catch any obvious while going through the changed files. What I found so far in the travis runs:
I didn't look into appveyor yet. |
|
@larsoner Do you know what is needed to get backends (eg. qt5) tested in python 3.6? Or what holds us from doing so? |
|
Qt5 wouldn't even create a context on Linux, so it will probably be a pain -- see #960 That OSMesa I compiled a long time ago so it might need some updating, not sure. The deprecation stuff we should deal with separately, it will require refactoring our testing code. I'll try to fix the Flake stuff, though, that should run. |
|
Qt5 from the defaults channel doesn't yet support OpenGL: ContinuumIO/anaconda-issues#1267 But the version from conda-forge does, so could use that channel? |
|
Perhaps, but let's do it in another PR :) |
|
Flake8 should be fixed. I don't think this fixes everything but it at least improves the situation and gets more testing done. @astrofrog @kmuehlbauer okay to merge once the CIs are happy? |
|
@larsoner Yes, sure go ahead. But we should not forget the other issues (eg. osmesa). BTW @astrofrog @davidh-ssec do you know, if we can utilise osmesa from conda-forge (mesalib)? |
I saw this log and thought it could be cleaned up a bit. Hopefully this does it.
Closes #1020.