[issue22273] abort when passing certain structs by value using ctypes

[Vinay Sajip]

Vinay Sajip added the comment:

> We need more feedback on this suggested flag, especially to stay consistent with CFFI if possible.

Undoubtedly, more feedback would be very helpful. I'm not sure using this flag impacts on consistency with CFFI particularly, since it's an internal implementation detail. Its main purpose would be for ctypes to raise exceptions rather than leading to crashes or undefined behaviour, as we have at the moment.

> Do you know whether CFFI supports passing unions and structs with bitfields in its ABI mode for 64-bit Unix?

I don't believe so, but I'm relatively new to this area. I'm not sure if things have changed recently, but an analogous CFFI issue was closed as WONTFIX in 2015, citing lack of support in libffi:

https://bitbucket.org/cffi/cffi/issues/150/structs-with-bit-fields-as-arguments

Also, the latest CFFI documentation, near the bottom of this section:

https://cffi.readthedocs.io/en/latest/using.html#function-calls

says:

"The limitations are that you cannot pass directly as argument or return type:

* a union (but a pointer to a union is fine);
* a struct which uses bitfields (but a pointer to such a struct is fine);"

The documentation applies these limitations regardless of any specific ABI (presumably to provide consistency).

So, I would guess that, as with ctypes, lack of libffi support is the main obstacle. I suppose one would have to seriously consider contributing there to make much headway here. In this still-open issue from 2013:

https://github.com/libffi/libffi/issues/33

Anthony Green of libffi said he'd welcome a patch, in response to a question by Eli Bendersky. Of course, it may be hard for individual contributors to support this for the range of architectures that libffi covers.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue22273>
_______________________________________