#general
- e
enough-analyst-54434
02/17/2017, 6:53 PMyes - e
enough-analyst-54434
02/17/2017, 6:53 PM - h
happy-kitchen-89482
02/17/2017, 6:53 PMSome of our
are globbed into targets__init__.py - h
happy-kitchen-89482
02/17/2017, 6:53 PMsome are not - e
enough-analyst-54434
02/17/2017, 6:54 PMIts probably simplest to work backwards from a chroot - e
enough-analyst-54434
02/17/2017, 6:54 PMedit the init.py in it or lack thereof, run pytest in it, rinse repeat - e
enough-analyst-54434
02/17/2017, 6:55 PMthen figure out why we don't assemble that way - if this is truly it - h
happy-kitchen-89482
02/17/2017, 7:28 PMFun fact #1: The test_*.py files don’t even go in the chroot! They are provided as arguments on the cmd line, and pytest does test discovery on them and then presumably attempts to import them directly from the source tree. - h
happy-kitchen-89482
02/17/2017, 7:28 PMMaybe you already knew this, but it was a surprise to me. - e
enough-analyst-54434
02/17/2017, 7:29 PMI knew it at one time, but honestly knew that no longer - h
happy-kitchen-89482
02/17/2017, 7:32 PMIf they did go in the chroot the source trees would get merged into a single one, so presumably this problem wouldn’t occur... - h
happy-kitchen-89482
02/17/2017, 7:35 PMInteresting, so if I run this manually in the same chroot, I can make it go away simply by changing the list of test files presented to pytest. So this isn’t a chroot issue per se. We are confusing pytest with our sources. - h
happy-kitchen-89482
02/17/2017, 7:35 PMTime to dive into the pytest test discovery algorithm - e
enough-analyst-54434
02/17/2017, 7:36 PMlooked wuick and we pass paths relative to buildroot - e
enough-analyst-54434
02/17/2017, 7:36 PMso it may be that that is what messes it up - h
happy-kitchen-89482
02/17/2017, 7:36 PMwell, the algorithm finds python package roots by walking up the tree to the first dir that doesn’t contain an__init__.py - e
enough-analyst-54434
02/17/2017, 7:36 PMit may be must chunk by source root - only for pytest - h
happy-kitchen-89482
02/17/2017, 7:37 PMWhich should work - e
enough-analyst-54434
02/17/2017, 7:37 PMeverything is pluggable in pytest, so I'm sure test discovery could be made to work with multiple test roots - h
happy-kitchen-89482
02/17/2017, 7:41 PMWe’re not really even using the test discovery algorithm though, we’re just giving it an explicit list of modules. But I assume the root discovery part is the same. - h
happy-kitchen-89482
02/17/2017, 7:42 PM - h
happy-kitchen-89482
02/17/2017, 7:42 PMThis is it - h
happy-kitchen-89482
02/17/2017, 8:05 PMafter some tracing, looks like pytest uses import hooks and generally messes around with python imports in a way that doesn’t play nicely with namespace packages - w
witty-crayon-22786
02/17/2017, 8:08 PM - h
happy-kitchen-89482
02/17/2017, 8:12 PMI’ll focus on fixing all this in the new pytest_run - h
happy-kitchen-89482
02/17/2017, 8:12 PMWe’ve gotten this far with the old one... - w
witty-crayon-22786
02/17/2017, 8:26 PMheh - s
straight-waiter-86090
02/17/2017, 9:03 PMhi. I have a problem using the node plugin, which I entered. https://github.com/pantsbuild/pants/issues/4254. Wondering if it could be the same as this: https://github.com/pantsbuild/setup/issues/19. - s
straight-waiter-86090
02/17/2017, 9:04 PMWhen I run with PEX_VERBOSE I see this:IncapableInterpreter: Interpreter /Users/dyoung/.cache/pants/setup/bootstrap-Darwin-x86_64/pants.deby4K/install/bin/python2.7 not capable of running EggInstaller - s
straight-waiter-86090
02/17/2017, 9:05 PMI’m not sure how to work around it based on this description of the
version though.setuptools