traitlets.py Type validate method ate what would have been a valuable error message to see

[stevehaertel]

traitlets.py has a Type class with a validate method.

If this validate method fails, it just raises the error "must be a type".

I was trying to start Jupyter with a custom login class, and I would keep getting this error when trying to start Jupyter.

It turns out that the specific conda environment had problems, and I had to manually changed the traitlets.py code to except error as e, then raise an error of str(e). On then did I finally see what the root cause of the problem was in my logs.

Would it be possible to append a string output of the specific error so we can save everyone some troubleshooting/debugging time?

[rmorshea]

This seems like it's really only a problem in Python 2 since by default Python 3 will now show all exceptions that occurred in the stack. My suggestion would be to upgrade to Python 3 (if possible).

However, if you're working in Python 3 and it's still hiding the underlying error, then it's probably a problem with how the error is getting logged in the Jupyter application itself since this sort of failure should be at the ERROR logging level.

[stevehaertel]

It's something inside Jupyter that is deciding what level of error it is, I guess.

[2019-11-29 09:51:51] Starting Jupyter Enterprisegateway.
[2019-11-29 09:51:52] Jupyter Enterprise Gateway started at https://hostname:####.
[C 09:51:53.141 NotebookApp] Bad config encountered during initialization:
[C 09:51:53.141 NotebookApp] The 'login_handler_class' trait of <notebook.notebookapp.NotebookApp object at 0x1000036c25c0> instance must be a type, but 'notebook.auth.loginauth.LoginHandler' could not be imported

Note: I haven't started running any application inside Jupyter yet - this is just starting a Jupyter instance.

EDIT: Yes, it was Python 3.

@rmorshea , based on the above information, would you say that the problem is likely because Jupyter code is not logging at an ERROR level? (i.e., I should be going to Jupyter to submit an issue?)

[rmorshea]

@stevehaertel since it's already logging the message I don't think changing the level will do anything. Jupyter is just suppressing the traceback even though it should have access to it. You could try using jupyter notebook --debug but as I said I don't expect that to work.

[rmorshea]

Post a message here and see if anyone can help: https://gitter.im/jupyter/jupyter