New targets for esp32s2 and esp32s3

[ivmarkov]

Given that for these MCUs Espressif provides separate toolchains, perhaps we need separate targets too. At the very least, the linker field in TargetOptions as well as the cpu should be different.

Also: shall we create a target for esp32c3? I get it that it is not really 'xtensa', but it needs the same changes. Possibly based on some of the existing "riscv-" targets?

[ivmarkov]

... and while we are at it, how about putting some extra info in the targets, like unix, vendor = espressif and env = newlib. Mentioned here.

For those who have no interest in using this with the ESP-IDF SDK, these would do no harm, as they won't be trying to build the STD crates anyway. However, for me this extra info is very beneficial, as most of my patches against STD are checking for generic stuff, like cfg(unix) and target_env = newlib. A few also check for target_os = none and target_vendor = espressif.

[MabezDev]

Hi @ivmarkov,

I've just updated the fork to 1.51, with a target def for the esp32s2, it seems esp32s3 support has not yet landed in llvm.

Also: shall we create a target for esp32c3?

I'm not yet convinced this is necessary, and I think we should try and lean on upstream support where we can until we really need custom definitions.

As for the custom defs for libstd porting, I'm not apposed to them haven't really looked into the libstd porting effort enough yet.

[ivmarkov]

Closing as the esp32s2 target is there, and the ESP-IDF-specific targets for Xtensa & RiscV are part of the libstd pull request.