[mypyc] fix: reject invalid mypyc_attr args [1/1]#19963
[mypyc] fix: reject invalid mypyc_attr args [1/1]#19963p-sawicki merged 26 commits intopython:masterfrom
mypyc_attr args [1/1]#19963Conversation
for more information, see https://pre-commit.ci
for more information, see https://pre-commit.ci
|
I'm not really sure of the best way to test this, but I have a failing test here that confirms the error and note both work as intended: https://github.com/python/mypy/actions/runs/18171952588/job/51728555657 |
BobTheBuidler
changed the title
mypyc_attr argsmypyc_attr args [1/1]
you could add an irbuild test that succeeds if mypyc reports an error, similar to this one. just change the attribute name to something unsupported, like |
|
I added the test and it looks like its working, but how do I signify to the test that one particular line will emit both an E and a note? |
mypyc/test-data/irbuild-classes.test
Outdated
| @mypyc_attr(invalid_kwarg=True) # E: 'invalid_kwarg' is not a supported `mypyc_attr` | ||
| class InvalidKwarg: | ||
| pass | ||
| @mypyc_attr(str()) # E: All `mypyc_attr` args must be string literals. |
There was a problem hiding this comment.
Do we actually want this? I noticed the current code ignores any values that arent string literals, and I thought an err would be better than a silent ignore, but should we just parse the non-literal string value like anything else in python would?
I dunno, maybe I'm overthinking it and this is fine as-is
There was a problem hiding this comment.
i think it's fine as-is. i doubt this will break any existing code and might catch user errors.
mypyc/test-data/irbuild-classes.test
Outdated
| @mypyc_attr("allow_interpreted_subclasses", "invalid_arg") # E: 'invalid_arg' is not a supported `mypyc_attr` \ | ||
| # N: supported keys: 'allow_interpreted_subclasses', 'free_list_len', 'native_class', 'serializable' |
There was a problem hiding this comment.
style nit: we usually wrap names in double-quotes in errors/notes, so it should be "invalid_arg", "mypyc_attr", "allow_interpreted_subclasses", etc.
for more information, see https://pre-commit.ci
2 participants
This PR emits a builder error when an invalid key is passed into
mypyc_attr. This change could have saved me ~20 minutes or so, when the root of my problem was just a simple typo.