pytorch
267b554b - fx: Fix type_matches for Optional[List[int]] arguments (#56790)

Commit
4 years ago
fx: Fix type_matches for Optional[List[int]] arguments (#56790) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/56790 If the argument doesn't match `List[int]`, this code falls through to `issubclass(argument_type, List[int])` which is invalid and raises a `TypeError`. If this happens during the processing of a `Union` (e.g. `Optional`), the other union types aren't given the chance to match against the signature. This also stop normalize_function from indescriminately swallowing exceptions, which let this bug go unnoticed. Test Plan: Imported from OSS Reviewed By: ngimel Differential Revision: D27987746 Pulled By: mruberry fbshipit-source-id: c5aa5f61a215f0f39925e7053f33bff4b5d5acc2
Author
Parents
Loading