core/atomic.d: Add GDC atomic support code

[ibuclaw]

Following on from #2596, rewrite to work with the new atomic API.

This has been peer reviewed by @TurkeyMan

Best viewed with whitespace ignored, as the removals are only highlighted by indenting the DMD specific code into a version(DigitalMars) block.

[dlang-bot]

Thanks for your pull request, @ibuclaw!

Bugzilla references

Your PR doesn't reference any Bugzilla issue.

If your PR contains non-trivial changes, please reference a Bugzilla issue or create a manual changelog.

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub fetch digger
dub run digger -- build "master + druntime#2831"

[TurkeyMan]

I already reviewed this, but here's my 👍

[TurkeyMan]

What's sad is that no unit tests will cover this GDC code... is that fine?

[ibuclaw]

What's sad is that no unit tests will cover this GDC code... is that fine?

Erm, all the unittests in core.atomic.

[TurkeyMan]

Yeah but the DMD CI won't prove this code path, it could regress against GDC?

[ibuclaw]

Yeah but the DMD CI won't prove this code path, it could regress against GDC?

If it isn't syntactically valid, an error would be thrown. Any API changes need to be synchronized, however I would expect the person making the change would do the required due dilligence, otherwise there's no basis for accepting them.

[kinke]

Iain, what's your ultimate goal wrt. these upstreamings? Do you aim for not having to maintain druntime/Phobos forks or do you plan to restrict the upstreamings to manageably-sized diffs?

Edit: Never mind, I've just seen that discussion in the previous PR. Certainly less work for us downstream compilers, but the code doesn't get more readable/tight. Not sure if LDC will follow suit; the recent refactorings are certainly a hassle to adapt to, but on the other hand, upstreaming everything will make refactorings a bigger hassle here upstream, and in the worst case, less likely due to the additional work for potentially interested contributors.