Subprocess kill is not async #535
Conversation
There was a problem hiding this comment.
Thanks David. This is also consistent with what we did in JKM.
|
Do we need a patch release to get the change available for those other PRs? |
|
I think this would be good as the jupyter_server PR produces intermittent failures during CI. I don't know about the nbclient behavior. |
|
This could explain some hanging tests we see in python 3.5 for nbclient, I'll kick off a release now |
|
6.1.1 is released |
|
Thank you. I started drilling into the odd CI failure we've been seeing in the Notebook repo since about a month ago, and believe it is associated with JC 6.0. The tests are failing in kernel.js for its auto-restart test. This theory is based on taking two of the failing PRs - of completely unrelated areas - and pinning their JC versions to 5.3.4 and both succeed. Because 6.0 consists of numerous fixes from the master branch - some 2 years old - this might be difficult, but I'm going to focus on some of the PRs I added since they could be related to restarts in some manner (as most of my PRs are in the areas of kernel management). At any rate, we might need a 6.1.2 sometime in the near future. |
|
Sounds great thanks for digging into it. I was going to start candidate testing 6.0 here but the virus has prompted delaying new changes for a couple weeks on our stacks. |
|
PR #536 has been submitted to address the NB CI failures. |
The issue was revealed in jupyter/nbclient#37 and jupyter-server/jupyter_server#191.