Don't look for jupyter at startup with raw kernels #8291
Conversation
| } | ||
| } | ||
|
|
||
| @debounceAsync(500) | ||
| @swallowExceptions('Failed to pre-warm daemon pool') | ||
| private async PreWarmDaemonPool() { | ||
| if (!this.extensionChecker.isPythonExtensionActive) { | ||
| if (!this.extensionChecker.isPythonExtensionActive || this.rawSupported.isSupported) { |
There was a problem hiding this comment.
I'm confused by this? Why not prewarm the daemon pool when using raw? We still use daemons for raw.
There was a problem hiding this comment.
Here we're attempting to pre-warm the daemon pool against the python env that starts jupyter.
When using raw, we pre-warm against the python env that's associated with the kernel.
Its possible that the kernel selected is not the same as the python env used to start jupyter,
If using non-raw, then it makes sense
There was a problem hiding this comment.
I guess the problem then is nobody knows (at that point in the code) why raw isn't allowed. Can you put a comment explaining why?
There was a problem hiding this comment.
Or some other way to make it obvious this daemon pool is just for jupyter?
Part of #7849
For #8350
We pre-warm daemons (with jupyter interpreter) and look for Jupyter when we're dealing with raw kernels.
This is totally unnecessary, the only time jupyter interpreter is used is when exporting & the like (its ok for that to be fractionally slower than slowing all of extension & kernel startup).