[roottest] Use nbconvert as a library in nbdiff.py#20456
[roottest] Use nbconvert as a library in nbdiff.py#20456guitargeek merged 4 commits intoroot-project:masterfrom
Conversation
Test Results 21 files 21 suites 3d 18h 30m 47s ⏱️ For more details on these failures, see this check. Results for commit 8b6fb40. ♻️ This comment has been updated with latest results. |
b2c687b to
08a8640
Compare
Using nbconvert as a Python library is less fragile, because otherwise there is ambiguity how the command line util should be called (the name changed over the versions as it migrated from ipython to jupyter).
pcanal
left a comment
There was a problem hiding this comment.
LFTM (but see the set of questions)
We only support Python 3 at this point. By updating the reference notebooks to have the Python 3 kernel in the metainfo, we don't need to do extra steps in the `nbdiff.py` script, mapping injecting the `python3` kernel name manually.
|
Thanks for the review! I have added one commit that addresses your comments. In particular, the reference notebooks are not updated to Python 3, so we don't have to manually set the kernel name in the |
|
The failures in Fedora Rawhide are unrelated: they come from an update of the system Maybe it's possible to also "hash" the docker image somehow, and when it changed do a clean build. To be seen later if this should continue to annoy us. |
This reverts commit 2e9c9e6. Things might work, now that we're using nbconvert as a library: root-project/root#20456
This reverts commit 2e9c9e6. Things might work, now that we're using nbconvert as a library: root-project/root#20456
Backport of root-project/root#20456. (cherry picked from commit b2cb261)
Using nbconvert as a Python library is less fragile, because otherwise there is ambiguity how the command line util should be called (the name changed over the versions as it migrated from ipython to jupyter).
This will make it possible to run the notebook tests also on the Fedora Rawhide build that is linked against the Python debug version, and hopefully also fix the test failures on the
macOSplatforms when a Python interpreter other than the system Python is used.