This is why I'm glad we have the validate step. Glad that @rchiodo caught this.

My fix for 2.7 initially did this:

1. Try to launch the daemon
2. If the daemon comes back as a IPythonExecutionService then we failed to launch the daemon
3. Return undefined from the daemon launcher
4. In the kernel process if the daemon launcher returns undefined, then use process execution service to launch the kernel spec

This would work, and was working for my test case. But it could very easily pick up the wrong interpreter. It would certainly pick the wrong one for anything that was just an interpreter and not an venv / conda env.

New flow is:

1. Try to launch the daemon
2. If the daemon comes back as an IPythonExecutionService then use that to launch the process in the daemon launcher
3. Only use the general process execution service for non-python.

I might check this in first, since I want this fix in release. But maybe we could use a functional test here? I'll look into if it's reasonable to write one.

Functional tests (that actually use jupyter) use conda at the moment. I believe you said conda doesn't have this problem.

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities (and 0 Security Hotspots to review)
0 Code Smells

No Coverage information
0.0% Duplication