Revert "Extract atomic platform specific implementation into an implementation file"

[CyberShadow]

Reverts #2739

#2739 contains a bugfix that happened to be a breaking change. The breaking change affects vibe.d and many projects using vibe.d, thus breaking the Buildkite autotester.

A follow-up PR, #2745, contains a change which reverts the bugfix, thus unbreaking VIbe.d. However, it looks like that will take a while to fully review.

For now, we can revert #2739 to unbreak master. The submitter seems to agree with this approach.

CC @TurkeyMan

[dlang-bot]

Thanks for your pull request, @CyberShadow!

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#2752"

[dnadlinger]

Yep; first order of business should always be to unbreak the CI, since that brings down all progress globally. This doesn't detract from the original PR, but just addresses a technical/procedural mistake (merging with CI status red).

[TurkeyMan]

Why is buildkite still red?

Alternatively, I could just reintroduce the bug in isolation... that's a 3-line patch.

[TurkeyMan]

Alternative approach: #2753

I don't care which one is chosen.

[CyberShadow]

Why is buildkite still red?

Uh-oh.

It looks like a different error.

master fails with:

../../crypto/vibe/crypto/cryptorand.d(173,6): Error: cannot modify `const` expression `p`

master + this PR fails with:

../../../.dub/packages/eventcore-0.8.44/eventcore/source/eventcore/drivers/posix/processes.d(305,25): Error: undefined identifier `intptr_t`

Alternatively, I could just reintroduce the bug in isolation... that's a 3-line patch.

If it's not too much trouble.

[TurkeyMan]

Link in comment above.

[Geod24]

Why is buildkite still red?

A new release of eventcore was tagged yesterday and that release is broken. It's already fixed: vibe-d/eventcore#120
Now @s-ludwig just needs to tag a new release of eventcore.

[CyberShadow]

A new release of eventcore was tagged yesterday and that release is broken.

How did that happen? Is it broken in all D versions? Or is it broken only with D master (meaning, the new release exposed a regression in our master)?

[TurkeyMan]

I also pushed a fix to vibe.d which corrects against the bug in DMD which I fixed (vibe-d/vibe.d#2349), and that's just waiting for their CI to flush... so we could not do anything, and it will resolve when vibe.d merges the fix... too many choices.

[Geod24]

it will resolve when vibe.d merges the fix

No, it also needs a new release for Buildkite to pick it up.
And as mentioned in my other comment, it transitively breaks a lot of projects. Libraries could get the update for free, but apps that commit dub.selections.json for example will need a manual update.

How did that happen? Is it broken in all D versions?

I don't know, and asked the same question. All D versions AFAICS. Only 2.087.x.

@wilzbach (and @others): Would now be the time to make buildkite required ?
Since it's not the first time it is flat out ignored when relevant.

[wilzbach]

Yeah I will set it as required once it's green again.

[CyberShadow]

Merged #2753.

I understand that the remaining failures should be cleared up once the eventcore bump gets propagated to Vibe.d.