Log details of why test discovery failed

[DonJayamanne]

Assume we have a test file with syntax errors, now attempt to discover tests.
The discovery process will fail with no indication of the cause.
Using the old approach we'd see the full error message.

We probably need to implement:

• Logging of test discovery
• And returning of discovered results

See here https://github.com/Microsoft/vscode-python/wiki/Testing-Architecture-Review#adapter-for-discovery-running--debugging

[ericsnowcurrently]

This may mean printing out stdout and stderr in the same way we do in the functional tests.

Also, we might also consider leveraging pytest's exit codes.

[DonJayamanne]

Also, we might also consider leveraging pytest's exit codes.

I disagree. why not make it simpler:
The adapter code returns the following info, all in one JSON:

• Include stdout
• Include stderr
• Include the discovered tests

stderr in the same way we do in the functional tests.

Doing this breaks the JSON returned. The output from the discovery needs to be JSON. I don't think we should be looking for begin and end JSON markers.

Also, we shouldn't have to code for pytest exit codes, the whole point of using the adapter is so we don't have to worry about such things. What about unit test & nose or another in the future?

This way, if there are errors, we notify the user.
Either way, what ever is in stdout and stderr in the JSON returned we can dump it into the Python TestLog, as was done in the older version of the extension.

[karrtikr]

validated