Merge tag 'swift-DEVELOPMENT-SNAPSHOT-2020-01-20-a' into tensorflow #29347
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
…y expr in unused
This reverts commit 3e932c0. The compiler does not support @_alwaysEmitIntoClient properties specially wrt property descriptors. The revert commit would introduce an ABI incompatability when a keypath to Array.first is formed: let greetings = ["hello", "hola"] let count = greetings[keyPath: \[String].first?.count] Runmning on an older runtime would lead to linker errors against $sSa5firstxSgvpMV the property descriptor for Array.first. rdar://58484319
…hod. Done using Xcode's refactoring engine! Should be NFC. I am going to turn this method into a visitor on the op code. NOTE: Xcode's refactoring engine messd with the whitespace, so I also ran this through git-clang-format.
…9165) The check for whether or not we can use a function builder on a property checked whether the attached getter had a body, which is not always true for module interfaces. Just disable that part of the check when compiling a module interface. rdar://58535753
[build-script] Install SwiftSyntax to usr/lib instead of usr/lib/swift
* Expose separate function sections option to compiler * Add function sections test
…eRequirementBetweenConcrete-during-completion [AST] Handle visiting UnresolvedType in TypeMatcher
…on-Equatable raw type The error recovery logic around derived conformances is a little bit tricky. Make sure we don't crash if a type explicitly provides a RawValue type witness that is not equatable, but omits the witnesses for init(rawValue:) and the rawValue property. Fixes <rdar://problem/58127114>.
[Diagnostics] Port diagnostics for array/dictionary literals
[CodeCompletion] Always look into decls to find the parsed expression
[CodeCompletion] Reduce accumulating memory per fast completion
…stall names Using the new linker directives $ld$previous requires the compiler to know the previous install names for the symbols marked as removed. This patch teaches the compiler to take a path to a Json file specifying the map between module names and previous install names. Also, these install names can be platform-specific. Progress towards: rdar://58281536
This validation may call many `stat(2)`. Since we don't expect system files are edited. Disable it for code completion, even if they are edited, they are validated when the user manually build the project. rdar://problem/58550697
`@differentiable` attributes may contain `where` clauses referencing generic parameters from some generic context, just like `@_specialize` attributes. Without special ASTScope support for `@differentiable` attributes, ASTScopeLookup.cpp logic tries to resolve the generic parameter `DeclName`s in `where` clauses based on source location alone (`ASTScopeImpl::findChildContaining`) and fails. The fix is to add a special `DifferentiableAttributeScope`, mimicking `SpecializeAttributeScope`. Every `@differentiable` attribute has its own scope, derived from the declaration on which it is declared. Unlike `@_specialize`, `@differentiable` may also be declared on `AbstractStorageDecl` declarations (subscripts and variables). Upstreams #27451. Progress towards TF-828: upstream `@differentiable` attribute type-checking.
This adjusts the test diffing to ignore the whitespace changes across different platforms. `raw_fd_ostream` will write out `OF_Text` files with the platform's line endings, which may differ from the source file's.
…le-fix Sema: Fix crash when synthesizing RawRepresentable conformance with non-Equatable raw type
…ntend jobs Unless you do this, the flag has no effect when used with the driver; it only worked in conjunction with -Xfrontend. Noticed while working on <rdar://problem/58490723>.
This is a workaround to fix weak linking of frameworks that define types with availability, but then define extensions of those types without availability. This can come up if the framework itself is built with a newer deployment target than the client that uses the framework. Since the type checker only enforces that an extension has availability if the extension is less available than the deployment target, we were failing to weak link the members of the extension in this case. This is not a perfect fix; ideally such frameworks should be built with -require-explicit-availability, and all reported warnings fixed by adding explicit availability. However, it allows clients to weak link when using existing swiftinterface files that have already shipped in the mean time, and it should not cause any problems once the frameworks are properly annotated in the future. Fixes <rdar://problem/58490723>.
…ity-fix Fix driver to pass -require-explicit-availability flag to frontend jobs
TBDGen: teach the compiler to take a json file indicating previous install names
test: adjust Migrator tests (NFCI)
AST: Weak-link declarations in extensions of weak-linked types
…metadata for now Recently we started using the runtime demangler to realize imported Objective-C class metadata, to save on code size since no metadata accessor function needs to be emitted in this case. This introduced a regression where OBJC_CLASSREF symbols were no longer being emitted in some cases, which breaks autolinking. The fix for this was tracked by rdar://56136123, but unfortunately had to be reverted because it caused other problems. Until the original fix can be re-applied, let's put in a temporary workaround where we avoid going through the runtime demangler for imported Objective-C classes. This regresses code size but unblocks everyone involved, until the fix for the demangling code path (8247525) can be re-applied. Fixes <rdar://56621277>.
Use the presence or absence of the PGO reader instead. SILGen was wiping out this option then checking for it being wiped out anyways.
This was being done at an odd point in the frontend presumably because by that point the private discriminator had been fully computed. Instead, push the conditions for generating the prefix data down to debug info generation and stop mutating IRGenOptions::DebugFlag in the frontend.
[Build System: Python] Test restructure
[NFC][Build System: Python] Cleaned-up the python_lint.py script.
[NFC][Build System: build-script] Remove the old which module in swift_build_support.
…hich replaces the existing module from swift_build_support. This new module uses the build_swift.shell.ExecutableWrapper API to create a wrapper class around 'xcrun'. The wrapper class is instantiated and exposed under the name build_swift.wrappers.xcrun.
[Build System: build-script] Adds a new xcrun module to build_swift.
burmako
force-pushed
the
tensorflow-merge
branch
from
2f45701 to
457df8f
Compare
rxwei
reviewed
Jan 23, 2020
| @@ -208,7 +208,6 @@ bool GenericParamScope::doesContextMatchStartingContext( | |||
| return false; | |||
| } | |||
|
|
|||
| // SWIFT_ENABLE_TENSORFLOW | |||
There was a problem hiding this comment.
It's awesome that these all have been upstreamed. Thanks for removing the comments :)
- Fix `resolveDifferentiableAttrDerivativeGenericSignature`.
- Generic signature should be returned directly instead of retrieved from
generic environment.
- Delete `DifferentiableAttributeParameterIndicesRequest`.
- That request was upstreamed and renamed to
`DifferentiableAttributeTypeCheckRequest`.
Stdlib compilation (`TensorFlow` module) fixed.
…h a new (unexpected?) failure
Use new `-Xfrontend` flag instead of `-Xllvm` flag.
|
@swift-ci Please clean test tensorflow |
|
Current status:
Proposed next steps:
|
dan-zheng
approved these changes
Jan 27, 2020
This was referenced Jan 28, 2020
burmako
pushed a commit
that referenced
this pull request
Jan 28, 2020
It would seem that despite getting a green checkmark from CI, #29347 broke `tensorflow` HEAD. There are now a few SourceKit tests which fail only on macOS. In this PR, I'm disabling the failing tests pending a full-fledged investigation.
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.
No description provided.