You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I did this for the icicle-python project (which uses rust), but it greatly simplifies things to only have to build wheels once per OS. The main issue here is building for linux, but there are nice docker images for manylinux: https://github.com/icicle-emu/icicle-python/blob/main/.github/workflows/CI.yml#L102. You also do not need to support new python versions, because the binaries work for whatever ABI you pick
Hi! I think this is a good approach that can simplify a lot as you say.
I took a look at this idea some time ago. IIRC the main limitation was in the bindings/python/utils.cpp file, with the conversions from PyLong to the different Triton types. I think some of them can be fixed by using functions available in the API. However, for functions such as PyLong_AsUint256 and PyLong_AsUint512 the fix does not seem straightforward (if you want to keep compatibility with all the current Python version).
Anyway, I will take another look and see what I can do.
Btw, did you run into a similar issues in icicle-python? Any idea on how to deal with this in a clean way?
I did this for the icicle-python project (which uses rust), but it greatly simplifies things to only have to build wheels once per OS. The main issue here is building for linux, but there are nice docker images for manylinux: https://github.com/icicle-emu/icicle-python/blob/main/.github/workflows/CI.yml#L102. You also do not need to support new python versions, because the binaries work for whatever ABI you pick
Related issues:
The text was updated successfully, but these errors were encountered: