scoder

In general, C compile time features rather than Cython translation time features. Generating different C files is rarely a good idea. A common thing to do is to write configuration...

I looked through CuPy a bit. There are large chunks of multiply defined structs like here: https://github.com/cupy/cupy/blob/76f3f85f14842685b4b8048f22536f39d2079c92/cupy_backends/cuda/api/_runtime_typedef.pxi I certainly don't understand enough of it to give "final" answers (and this...

I think https://github.com/cython/cython/issues/2596 should be the most general way forward for this.

I'm changing the milestone to 3.0 because this ticket is about the deprecation. Removing the feature in 3.1 would be way too soon, but there will be a deprecation warning...

I agree that this behaviour should be changed. Returning the corresponding Enum object is the expected result. In the case of an invalid enum integer value, I'd also expect an...

>I propose to fix the missing coverage by killing off the feature.> +1, I also knew nothing about it.

For packaging Python applications (whether its binary modules are built by Cython or not), there are much better tools like pyinstaller or pyrun. See #2011 and #2849. For wrapping multiple...

I can see this being a nice optimisation. If it was fast, people would probably make use of this pattern (more). I've seen it being used a few times, and...

I looked through the usages of `CO_COROUTINE` in CPython (main) and they look safe, except for this one, which I don't quite understand yet: https://github.com/python/cpython/blob/50b9a7762f06335277d9962edc8d39498601a4e4/Objects/frameobject.c#L208-L211 All other usages only occur...

> That code's for checking if it's legal to set the frame)s `f_lineno`. While tracing you can do that to jump execution around, but you can't do it into/out of...