Please add support for loongarch64

[Qinka]

Please add support of loongarch64 for debian sid and unstable.

[tianon]

Since debuerreotype itself does support loongarch64 (see ports-related flags, which is where loongarch64 lives in Debian upstream right now), I'm going to instead assume this request for was for https://hub.docker.com/_/debian to support loongarch64, and I'm going to move it to https://github.com/debuerreotype/docker-debian-artifacts accordingly before answering it. 🙇

[tianon]

I have a few high-level initial issues with supporting this personally in my builds here:

1. it being in Debian Ports really complicates my build matrix (my experience building riscv64 while it was in Ports was pretty negative all around) -- this is fixable and I do have plans to refactor my build matrix eventually so that things like this get easier, but it's non-trivial and not a high priority for me at this time

2. I have no access to loong64 hardware -- I know that Loongson has been very generous with sending folks hardware, but I also have experience receiving hardware from Loongson in the past, and it is still a nightmare (non-mainline kernel, a custom GRUB fork, its very own special implementation of UEFI that nothing else supports correctly, etc), not to mention the social and political issues I've had doing so (having issues with customs/import taxes while trying to receive the box, having a language barrier with the folks providing the box, Loongson apparently now being on the US Federal Government's "Entity List" which makes it hard to determine whether it's even legally OK for me to accept hardware from them, etc), and then at the end of the day, the hardware itself is objectively a terrible experience (slow, unreliable, can't run mainline/stock Debian) and not a very suitable platform for a CI/CD build system

3. not well-supported by other projects/users (likely for similar reasons to the above); see mips64le docker-library/gcc#67 for a parallel example of MIPS support decaying over time

So in short, I unfortunately have no plans to support loong64 at this time.

However, to make this more concrete, here are things that would need to happen for me to reconsider that:

1. availability of hardware

   • from a US-based supplier / US-fed-approved cloud provider
   • running 100% stock upstream mainline Debian, on officially Debian-built kernels and a Debian-built bootloader
   • "reasonable" performance characteristics -- decent number of cores, amount of RAM, overall bus/system latency, etc (to be slightly more concrete, if it takes more than an hour to build one single version/variant of https://github.com/docker-library/gcc, it's probably not good enough)

2. either part of Debian proper (graduating from Ports), or me finding time to refactor the way I perform builds (which would be a lot more attractive to work on if the availability problems were solved reasonably and hardware worth using were available in a form/locale I could actually access)

I know this isn't the answer you're hoping for, but hopefully the transparency into my personal thoughts/feelings/plans on the matter are helpful.

[tianon]

Cross-linking docker-library/official-images#16404 (a related downstream issue).