On Sun, 14 Apr 2024 18:02:08 +0000 Sam Day <me@samcday.com> said:
Hello!
I recently learned that the AUR seems to have an actively enforced moderation policy preventing packages that either only work on aarch64, or are otherwise heavily focused/only useful on that platform.
RFC0032 hasn't finalized yet, and of course it could still end up being rejected or otherwise put on indefinite hiatus... My outsider perspective is that it likely will​ happen though, so I'm working under that assumption.
In this case, it's only a matter of time before non-x86 communities are accumulating around Arch upstream proper. RFC0032 already notes that:
- Ideally, the AUR web interface / API is extended to allow filtering packages by architecture.
I have a concrete proposal here. Let's say I (or someone else) were to contribute the necessary changes to aurweb, such that it exposed architecture info in the RPC (archlinux/aurweb#485), as well as adding support for filtering by architecture when searching. If this happened, and the search box on the AUR homepage defaulted to x86, would it be possible to relax the current moderation policy that restricts aarch64-only packages from living in the AUR? -Sam
I'd +1 this. I actually have some magic in my PKGBUILDS to deal with low memory systems and that was actually a bi-product of building pkgs on a rpi3 with only 1G RAM (though it could apply to x86 too). The package is portable to any arch that's sane (all the major ones still left)so it's not aarch64 only, but in principle I think this RFC is a good thing. Aarch64 "desktop/laptop" class CPUs other than the m1/2/3 are well on their way (well some say already here - I use one as a daily driver at work, a 32 core Ampere). The time will be coming when alarm (a separate project though IMHO a close cousin) may even start pumping out bootable ISO images (some people have been doing this before on their own). It'd not make sense to make a bigger wedge between close relatives. -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- Carsten Haitzler - raster@rasterman.com