Parse list value for terminado_settings

[krassowski]

Fixes jupyterlab/jupyterlab#12540. Allows configuring terminado_settings from command line with:

--ServerApp.terminado_settings="shell_command=['/bin/zsh']"

or if users are willing to accept the default tokenization by shlex.split with:

--ServerApp.terminado_settings="shell_command='/bin/zsh'"

Are there tests for configuration settings?

For context it is used in: jupyter_server_terminals

[blink1073]

This seems like a reasonable change for now. Ideally we'd have a custom parser for each of the allowed fields. Here's an example of a config setting test.

[codecov-commenter]

Codecov Report

Merging #949 (33fcc01) into main (4a34e6a) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #949   +/-   ##
=======================================
  Coverage   72.30%   72.30%           
=======================================
  Files          64       64           
  Lines        8136     8136           
  Branches     1358     1358           
=======================================
  Hits         5883     5883           
  Misses       1843     1843           
  Partials      410      410           

Impacted Files	Coverage Δ
jupyter_server/serverapp.py	65.76% <ø> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[blink1073]

https://github.com/ipython/traitlets/releases/tag/5.4.0

[blink1073]

Thanks! Failures are unrelated

[krassowski]

Thanks! Test Minimum Versions failure is relevant (other tests also fail on #971). Should I:

• mark the test case for skipping on older traitlets, or
• bump the minimum version of traitlets to 5.4 (currently 5.1)?

[blink1073]

Oh, the minimum version check is related. We'll need a version guard on the new test for the traitlets version.

[blink1073]

mark the test case for skipping on older traitlets, or

This one 😄

[blink1073]

Thanks again!

[vidartf]

@krassowski Note that the test that was added was failing locally on Windows for me. This change to the test make it pass, but I'm not sure if this means the bug is fixed on Windows or not? #983 (comment)

[westurner]

Is there already a way to use busybox ash (or bash) compiled to WASM with this setting?

emscripten-forge could build busybox, bash, and zsh, I think