[embedded] Start building embedded support on Linux/ELF, expand archs of the embedded stdlib to cover common embedded targets

[kubamracek]

There is nothing Apple / Mach-O specific in the Embedded Swift setup, let's enable the Embedded Swift support in the Linux toolchains and start building the embedded stdlib for ELF, too.

[kubamracek]

@swift-ci please test

[kubamracek]

@swift-ci please test

[finagolfin]

Nice, any plans to set up a CI that runs the compiler tests on actual hardware? At least, it would be nice to have a 32-bit non-embedded CI set up, that could run all the native executable tests too. I don't think anybody is doing that right now, unless there's still a 32-bit watchOS CI running internally right now somewhere.

I think it wouldn't be hard to set up a linux armv7 CI on one of the existing linux AArch64 images, since 64-bit distros usually support building and running 32-bit binaries also.

[kubamracek]

@swift-ci please test

[kubamracek]

Nice, any plans to set up a CI that runs the compiler tests on actual hardware?

As in, on an actual embedded board? I think we're probably pretty far from being able to run reliable and fast enough tests on those as part of the main Swift CI. But it's certainly an idea worth exploring, yes. Maybe there is some good answer by using an emulator?

At least, it would be nice to have a 32-bit non-embedded CI set up, that could run all the native executable tests too. I don't think anybody is doing that right now, unless there's still a 32-bit watchOS CI running internally right now somewhere.

I think it wouldn't be hard to set up a linux armv7 CI on one of the existing linux AArch64 images, since 64-bit distros usually support building and running 32-bit binaries also.

This seems tractable. I wonder if @shahmishal has any thoughts on that?

[DougGregor]

I think this is causing failures on the incremental bots, which look like:

<unknown>:0: error: fatal error encountered during compilation; please submit a bug report (https://swift.org/contributing/#reporting-bugs)
<unknown>:0: note: No available targets are compatible with triple "riscv32-none-none-eabi"

I put up what I think is a fix at #70057, and am going to kick off testing of a revert PR in case that isn't right.