ci: Use a regular cargo to build no-std targets.

[waywardmonkeys]

xargo is no longer needed and has been in maintenance mode for over 7 years.

This can run on stable rather than requiring nightly

The build doesn't need rustfmt, so don't require that it be installed.

[waywardmonkeys]

This will conflict with #227, but I'm happy to fix up any merge conflicts after whichever lands first.

[ThierryBerger]

I'm approving, but I have a minor concern over dropping nightly compilation: that means we don't have any ci testing nightly.

It's most definitely correct for this PR, but should we add a periodic CI step on master to test against nightly ? @sebcrozet

[sebcrozet]

Thank you for this PR @waywardmonkeys !

Back in the days, xargo was the only way to ensure that compilation actually fails if any crate (including indirect dependencies) attempts to link against the stdlib. Is that what we gain from cargo now as well? i.e. if we add some crate that needs the stdlib to parry’s Cargo.toml, will it fail to compile when targetting thumbv7em-none-eabihf?

It's most definitely correct for this PR, but should we add a periodic CI step on master to test against nightly ?

@Vrixyz We should keep a nightly test (on PRs) that checkes the simd-nightly feature. (But, yes, this should be in a separate PR.)

[waywardmonkeys]

Back in the days, xargo was the only way to ensure that compilation actually fails if any crate (including indirect dependencies) attempts to link against the stdlib. Is that what we gain from cargo now as well? i.e. if we add some crate that needs the stdlib to parry’s Cargo.toml, will it fail to compile when targetting thumbv7em-none-eabihf?

Yes.

Using cargo with a target that has no stdlib is (now) the typical way to verify that no_std is indeed working correctly.

This is what we do in kurbo for example as well:

https://github.com/linebender/kurbo/blob/16cb9af39eda5010ec30eb086822f741edfdf9f5/.github/workflows/ci.yml#L50

[waywardmonkeys]

As an aside, this would be a great time to bring up the work started in PR #170 and see about finding a victim to do more of that. :)

[ThierryBerger]

Is this related to the failure I'm seeing in #249 ? 👀

[sebcrozet]

Thank you for checking @waywardmonkeys! Looks good.