If not targeting an architecture family don't run it's compiler

[ratmice]

This is related to #377, when running

TOCK_TARGETS="cortex-m4" ./build_all.sh

One ends up getting a lot of messages if they don't have the riscv compiler installed,

make: riscv64-unknown-elf-gcc: No such file or directory

This appears to be because Configuration.mk runs the compiler to try and extract the compiler version.
This patch tries to only run the compiler/extract the version if the TOCK_TARGETS contains the string rv32.

I tested two cases:

• With TOCK_TARGETS="cortex-m4", ensuring it doesn't produce the error
• Without TOCK_TARGETS set, ensuring it still does because the default value contains the string.

[ratmice]

May want to hold off on this, I still get the error running:
Edit: Nevermind, I think it's just TOCK_BOARD having been deprecated/removed.

$ make TOCK_BOARD=nrf52840dk

because TOCK_BOARD doesn't influence TOCK_TARGETS, so this error still happens sometimes.
Hmm, I think perhaps TOCK_BOARD doesn't actually do anything at all, at least ripgrep doesn't turn anything up.

Should it rather be removed from here:
https://github.com/tock/tock/blob/master/boards/nordic/nrf52dk/README.md

[ppannuto]

You can check TOCK_ARCH_FAMILIES instead of parsing the targets again yourself, e.g. out of the box

TOCK_ARCH_FAMILIES=cortex-m rv32i

It'd be reasonable to add that to the config print from the verbose build to improve discoverability as part of this PR probably.

---

Yes, TOCK_BOARD should be removed from the documentation. I think the last vestiges of it were removed from here around a year ago as tockloader improved. It's also in imix documentation, which looks like it has more than should be purged, e.g., I don't think any of this paragraph is valid with modern tockloader.

[ratmice]

Thanks, I used TOCK_ARCH_FAMILIES, and added it to the verbose build output like you asked.

Yeah, I'll work on a patch for the nordic stuff that I can figure out, but I was really not sure what to do about imix especially the parts about the application address, so I'll probably just leave that alone and make a note of that in an issue.

[ratmice]

There is apparently one more place to do this check in this configuration dump.

[ratmice]

Fixed in ab7925a

[ratmice]

I figured I should also do the same for the cortex-m family.

[ppannuto]

This all looks good to me, thanks for the patches!