[Distributed] Enforce strict order of synthesized id and actorSystem properties; avoid wrong offset crashes #58745
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ktoso
changed the title
ktoso
commented
May 8, 2022
| // important. The `hint` below makes sure the system is inserted right after. | ||
| auto id = derived.Nominal->getDistributedActorIDProperty(); | ||
| derived.addMemberToConformanceContext(pbDecl, /*hint=*/id); | ||
| derived.addMemberToConformanceContext(propDecl, /*hint=*/id); |
There was a problem hiding this comment.
This ensures the system is always AFTER the ID
ktoso
commented
May 8, 2022
| llvm_unreachable("DistributedActor.id MUST be synthesized earlier, " | ||
| "because it is forced by the Identifiable conformance. " | ||
| "If we attempted to do synthesis here, the earlier phase " | ||
| "failed and something is wrong: please report a bug."); |
There was a problem hiding this comment.
We had to move away from its synthesis here a while back so this was dead code;
Actually I should just delete the hangling here I guess.
ktoso
commented
May 8, 2022
| @@ -27,33 +26,25 @@ import WinSDK | |||
| /// for learning about `distributed actor` isolation, as well as early | |||
| /// prototyping stages of development where a real system is not necessary yet. | |||
| @available(SwiftStdlib 5.7, *) | |||
| public struct LocalTestingDistributedActorSystem: DistributedActorSystem, @unchecked Sendable { | |||
| public final class LocalTestingDistributedActorSystem: DistributedActorSystem, @unchecked Sendable { | |||
There was a problem hiding this comment.
This UNDOES the workaround (which actually did not work after all).
This is NOT a change in the API of the system
ktoso
commented
May 8, 2022
| @@ -286,7 +274,7 @@ func test() async { | |||
|
|
|||
| let localDA = LocalTestingDA_Int() | |||
| print("localDA = \(localDA.id)") | |||
There was a problem hiding this comment.
This would have crashed recently -- this patch fixes it.
ktoso
requested review from
DougGregor,
xedin and
drexin
and removed request for
DougGregor
|
@swift-ci please test |
This was referenced May 8, 2022
|
Note that even the 32 bit watch passed here... and does not on 5.7, so we're missing something there 🤔 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Resolves rdar://92712849 - the root cause of all those issues
Resolves rdar://92910719 - which disabled the still failing test since the workaround was not good enough
This is the actual solution for all our mystical offset issues -- the order of the synthesized AST fields MUST match the order IRGen enforces (and the DA needs in any case); where the id and system are the FIRST fields, and must be specifically: id, system because that's what IRGen emits and many places seem to expect the order matches between the original AST and whatever gets derived form it and IR.
This probably also resolves a few other crashers we had with similar crash reasons of weird mismatching offsets, I'll be verifying those.