If When the basic CI passes, I'll send it to the bots.

There is no CI for cross building yet.

There is no CI for cross building yet.

Let's run it on the Emscripten container then; that's a cross-build, AFAIU.

Accidentally, I've got some ARM toolchains installed. I'll throw it at them and see how it fares.

Just out of curiosity: what distributions still don't have Autoconf so we need pregeneratedconfigure under version control? Won't it be better to let a user or a package manager to generate it on their system instead?

Package managers are not our concern. We keep configure and other scripts in git to lower the barrier for contributors.

Ourconfigure.ac more than just autotools. It also requires automake, a recent version of autoconf-archive and pkg-config's M4 macros. Not every distro has a recent enough autoconf archive package. Some distros have the pkg-config M4 files in a separate package. Until a couple of weeks ago ourconfigure.ac did not work with autoconf 2.71.

Well, if this file cannot be removed because of specific requirements for its generation, maybe it can be marked as generated akin to outputs of Argument Clinic, Freeze, and Parser/asdl_c.py? It will allow GitHub to collapse it. By the way, I created#30031 to address this.

BTW, this is a case where theconfigure diff matters at little bit more than normal: we have no CI for changes like this, so being able to see that the refactored code actually produces the same result is crucial. Luckily, this diff is pretty short 🙂

Accidentally, I've got some ARM toolchains installed. I'll throw it at them and see how it fares.

Works for me.

This PR is stale because it has been open for 30 days with no activity.

I have no time to follow up this clean-up. Closing.