Consider dropping bespoke pep517 backend

[rossburton]

Is your feature request related to a problem?

No response

Describe the solution you'd like

As per https://cython.readthedocs.io/en/latest/src/userguide/source_files_and_compilation.html#basic-setup-py, it's now possible to build cython extensions in setuptools. Would you consider dropping the bespoke backend?

Describe alternatives you've considered

.

Additional context

No response

Code of Conduct

• I agree to follow the aio-libs Code of Conduct

[webknjaz]

Partially, which is why I encouraged upstream to implement that originally.
It doesn't cover every aspect of the wrapper, however. I'd want to see which parts can be migrated to be declarative first.

Keep in mind that a copy of this backend exists in multiple repositories and any changes would have to happen in all the synchronized repos. They must not diverge too much (short of project-specific bits like dist names et al.)
It should remain possible to easily sync changes with something like nvim -d, comparing diff per-file.

Note that since your motivation here is unclear, I can't guide you more at this point.

[rossburton]

Where is the canonical location for the backend? I've a number of issues to file.

My primary motivation is reproducible builds of yarl/frozenlist/proplist which are all not reproducible (see cython/cython#5949).

My final .so has host paths embedded, despite the compiler being told to remap build paths, because temporary directories outside the expected build tree are used.