Improve error diagnostic for exception-spec mismatch #5547
Conversation
It seems like the most common Cython 3 snag is the change in function exception behaviour. This PR adds some extra diagnostics to help people navigate this change, and a general mechanism for adding this kind of diagnostics. It needs some tests, but I wasn't to see what tests fail to tell me where to put them.
Cython/Compiler/PyrexTypes.py
Outdated
| copied_src_type.exception_value = self.base_type.exception_value | ||
| if self.base_type.pointer_assignable_from_resolved_type(copied_src_type): | ||
| # the only reason we can't assign is because of exception incompatibility | ||
| msg = "Exception specifications are incompatible." |
There was a problem hiding this comment.
I'm not sure "exception specification" is something we use in the documentation, so it might be difficult to look up. I think we usually talk about "exception values", maybe "exception return values", although that's more constrained than "specification".
There was a problem hiding this comment.
We use it a few times in the docs (and at least of of those times was written by someone who wasn't me)
https://cython.readthedocs.io/en/latest/src/userguide/language_basics.html#language-basics
cdef functions that are not extern are implicitly declared with a suitable exception specification for the return type
https://cython.readthedocs.io/en/latest/src/userguide/migrating_to_cy30.html
There is an easy-to-encounter performance pitfall here with nogil functions with an implicit exception specification of except *.
https://cython.readthedocs.io/en/latest/src/userguide/nogil.html
Be aware that a function with an except * exception specification (typically functions returning void) will be expensive to call because Cython will need to temporarily reacquire the GIL after every call to check the exception state. Most other exception specifications are cheap to handle in a nogil block since the GIL is only acquired if an exception is actually thrown.
In most cases this is efficient since Cython is able to use the function’s exception specification to check for an error
So I'm not feeling too bad about it.
However; I'll replace it with "exception values"
It seems like the most common Cython 3 snag is the change in function exception behaviour. This PR adds some extra diagnostics to help people navigate this change, and a general mechanism for adding this kind of diagnostics.
It needs some tests, but I wasn't to see what tests fail to tell me where to put them.