Don't Serialize Declarations with package Access
#70100
Merged
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
This accidentally started happening when I adjusted getEffectiveAccess to return `Public` for `Package` declarations in swiftlang#69709. As a result, the optimizer thought it had more opportunities to go after declarations that used to be opaque. Unfortunately, this resulted in a miscompile as the devirtualizer was able to look through now-serialized package (static) functions. In one specific instance, the optimizer created direct calls to hidden accessors instead of going through the dispatch thunk.
|
@swift-ci smoke test |
slavapestov
reviewed
Nov 30, 2023
| @@ -1080,7 +1080,7 @@ void SILGenModule::emitNonCopyableTypeDeinitTable(NominalTypeDecl *nom) { | |||
| auto serialized = IsSerialized_t::IsNotSerialized; | |||
| bool nomIsPublic = nom->getEffectiveAccess() >= AccessLevel::Public; | |||
| // We only serialize the deinit if the type is public and not resilient. | |||
| if (nomIsPublic && !nom->isResilient()) | |||
| if (nomIsPublic && !nom->hasPackageAccess() && !nom->isResilient()) | |||
There was a problem hiding this comment.
is it possible that getEffectiveAccess() == Public && nom->hasPackageAccess()
There was a problem hiding this comment.
Yes. This PR #69709 changed the effective access level of package declarations to public since those declarations are visible across module boundaries.
There was a problem hiding this comment.
Would it be too disruptive to have getEffectiveAccess() return a new enum that does not include 'Package' then? Or maybe move the computation to the SIL linkage level, so that AccessLevel::Package becomes SILLinkage::Public? getEffectiveAccess() is a terrible hack for -enable-testing and we should not hang new behavior off of it. There's also a FormalLinkage concept for type metadata in IRGen, IIRC. It's all too convoluted.
elsh
approved these changes
Dec 1, 2023
tshortli
added a commit
to tshortli/swift
that referenced
this pull request
Dec 18, 2023
swiftlang#70100 prohibited `package` declarations from ever being serialized in order to solve a problem in which the declarations were being serialized inappropriately. That's too heavy handed, though, because an `@_alwaysEmitIntoClient` function with `package` access *must* be serialized because it has `public_nonabi` linkage. Resolves rdar://104711625
tshortli
added a commit
to tshortli/swift
that referenced
this pull request
Dec 18, 2023
swiftlang#70100 prohibited `package` declarations from ever being serialized in order to solve a problem in which the declarations were being serialized inappropriately. That's too heavy handed, though, because an `@_alwaysEmitIntoClient` function with `package` access *must* be serialized because it has public non-abi linkage. Resolves rdar://104711625
tshortli
added a commit
to tshortli/swift
that referenced
this pull request
Dec 19, 2023
…ol tests. Missing from swiftlang#70100.
tshortli
added a commit
to tshortli/swift
that referenced
this pull request
Dec 19, 2023
meg-gupta
pushed a commit
to meg-gupta/swift
that referenced
this pull request
Dec 21, 2023
swiftlang#70100 prohibited `package` declarations from ever being serialized in order to solve a problem in which the declarations were being serialized inappropriately. That's too heavy handed, though, because an `@_alwaysEmitIntoClient` function with `package` access *must* be serialized because it has public non-abi linkage. Resolves rdar://104711625
meg-gupta
pushed a commit
to meg-gupta/swift
that referenced
this pull request
Dec 21, 2023
…ol tests. Missing from swiftlang#70100.
meg-gupta
pushed a commit
to meg-gupta/swift
that referenced
this pull request
Dec 21, 2023
Catfish-Man
pushed a commit
to Catfish-Man/swift
that referenced
this pull request
Jan 19, 2024
swiftlang#70100 prohibited `package` declarations from ever being serialized in order to solve a problem in which the declarations were being serialized inappropriately. That's too heavy handed, though, because an `@_alwaysEmitIntoClient` function with `package` access *must* be serialized because it has public non-abi linkage. Resolves rdar://104711625
Catfish-Man
pushed a commit
to Catfish-Man/swift
that referenced
this pull request
Jan 19, 2024
…ol tests. Missing from swiftlang#70100.
Catfish-Man
pushed a commit
to Catfish-Man/swift
that referenced
this pull request
Jan 19, 2024
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.
This accidentally started happening when I adjusted getEffectiveAccess to return
PublicforPackagedeclarations in #69709. As a result, the optimizer thought it had more opportunities to go after declarations that used to be opaque. Unfortunately, this resulted in a miscompile as the devirtualizer was able to look through now-serialized package (static) functions. In one specific instance, the optimizer created direct calls to hidden accessors instead of going through the dispatch thunk.Ban serializing witness tables, vtables, and functions for package types once more.