Fixing a bug in which filtered dependencies based on complex filterin…#2983
Conversation
…g schema (specific versions) were not being filtered.
|
I wonder if you can write a test for this bug as well? |
|
I'm not quite sure how to trigger the bug actually. It showed up for some dependencies on CMake when we changed from I found the code path only by deduction. I know that this change gets rid of the bug that we were triggering. |
|
This definitely needs a test before it gets merged (or we will be bitten by this again sooner or later). |
|
@mboisson How did the bug manifest itself exactly? Was it via an error like |
|
Reopening to see if mysterious py 3.5 failure is reproduced. |
|
@mboisson I merged your branch with develop to see because it was a little out of sync with the Py3 changes. On build-node the tests run ok: Let's hope travis is happy with them too. |
|
Trying once more... |
|
@bartoldeman If a test frequently fails for no good reason with Python 3.5, we should try and get to the bottom of it, not re-trigger the tests until it passes. ;) Is this the same test that was failing for your PR #2994, i.e. |
I moved much of the filter_deps code to "dep_is_filtered" function, and reuse that function in _finalize_dependencies, which was only checking for the name of the dependency, instead of the full version check.