bpo-45607: Make it possible to enrich exception displays via setting their __note__ field

[iritkatriel]

https://bugs.python.org/issue45607

[iritkatriel]

Should I add PyException_GetNote and PyException_SetNote to the c api?

[erlend-aasland]

Nice feature! LGTM. I found one path not covered by the test suite. I of course also left some pedantic PEP-7 comments 🤓

(BTW, I still find the error handling in pythonrun.c hard to read... and it generates so many indent levels 😕)

[iritkatriel]

Thanks.

(BTW, I still find the error handling in pythonrun.c hard to read... and it generates so many indent levels 😕)

I'm working on it. I made a patch that changed so much it would take a year to review. I'll break it up.

[gvanrossum]

Nice! LGTM, but please fix the issues Erlend brought up.

[iritkatriel]

Nice! LGTM, but please fix the issues Erlend brought up.

I did all of them except the test coverage for the error when passing NULL to the setter. This would only come into play if I add PyException_GetNote and PyException_SetNote to the C api. I think we need those, right?

[gvanrossum]

I did all of them except the test coverage for the error when passing NULL to the setter. This would only come into play if I add PyException_GetNote and PyException_SetNote to the C api. I think we need those, right?

I think we should not add to the C API. If people want to access the note from C, they can use the __note__ attribute (like you do in the exception-printing C code). Even so, what happens when you try del err.__note__?

[iritkatriel]

Ah yes, very good:

>>> del e.__note__
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: __note__ may not be deleted

[gvanrossum]

That behavior is fine. So do you have a unit test for that? If not, it should be easy to add one.

[iritkatriel]

That behavior is fine. So do you have a unit test for that? If not, it should be easy to add one.

I don’t, I’ll add one.

[gvanrossum]

+1