this article summarizes the situation and explains a few things that were not clear to me from previous articles.
Neywiny•7mo ago
The windows support is interesting because Python dropped 32-bit x86 for Linux, yet kept it for Windows (and maybe ARM). That means at the end of the day unless they're doing stupid things, the code can't be assuming 64-bit, which meant once I modded the auto tooling to remove the lock-out it worked just fine. So what my thought is, is that I guess I understand both sides. It's not hard to just compile it yourself if you have the expected configuration, but it's also usually not much work to compile a package twice for a release flow once you have the pipeline setup.
em-bee•7mo ago
the python example continues to confuse me. in a comment in the proposal discussion cpython was used as an example for a case that would cause problems if it were to drop 32-bit support, and now you say that python already did drop 32-bit for linux. and that seemingly without consequences for fedora. that dropping 32-bit support would not have consequences is what i would have expected anyways, because fedora stopped offering a kernel and installer with 32-bit support some 5 years ago.
Neywiny•7mo ago
https://github.com/python/cpython/blob/847d1c2cb4014f122df64.... i686 is Windows only. I see there's a warning about this later if you're unsupported and therefore a 0 case, but either this becomes an error later on or they softened the impact since I did this. Or I'm looking at the wrong check.
em-bee•7mo ago