Merge swift-DEVELOPMENT-SNAPSHOT-2019-07-22-a into tensorflow.
#26341
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
…lt function argument.
The compiler allows unavailable types to be extended as long as the extension itself is marked unavailable: @available(macOS, unavailable) struct UnavailableStruct { } @available(macOS, unavailable) extension UnavailableStruct { } // no-error However, this doesn't work when the extension is inside a #if. The availability logic that finds the containing ExtensionDecl based on a source range gets confused and finds the IfConfigDecl instead. To fix this, change that logic to ignore IfConfigDecls. Members of an active #if are also represented in the enclosing context, as sibling of the IfConfigDecl, so the compiler will find the extension declaration there. rdar://problem/53079366
This extends the WinSDK module definition further to better modularise the headers. This should improve the header organisation as well as setup additional autolink rules.
`RefCount.h` can be included in a C context - e.g. building the SwiftShims module. Restrict the C++ overloads to the C++ context only. This partially improves the build for Windows ARM64.
AST: attempt to rewrite the expression to be more legible
… on Darwing toolchain
The memcpy in the type layout verifier was not correct for big- endian systems. While we are here change 'long long' to a fixed width unsigned type (uint64_t). It doesn't really make sense to print the value as a signed number since we have zero extended it from its original bit width using the memcpy.
Runtime: print type layout values correctly on big-endian systems
Remove TypeChecker::getWitnessType()
Platform: extend WinSDK further
Revert "Better runtime failure messages (not yet enabled by default)"
…cety-rdar51599533 [CodeCompletion] Add defensive nullptr check
…ion-body [Type checker] Requestify type checking of a function body
Identify constant definitions that look like include guard definitions
… for exhaustive switching over builtin types. I am going to use this to refactor some code in ASTPrinter onto BuiltinType wherein I use an exhaustive switch to print things as appropriate. Exhaustive switching over all builtin types just seems like a useful thing to have as well.
This commit centralizes the code that converts variable length tag values, stored in enum payloads and extra tag bytes, to and from the 4-byte integer values that the runtime uses to represent the enum case. Note that currently big endian machines will store the tag value in the first word of the destination. This reflects the current behaviour of the compiler. I am however expecting to change this so that the value is stored as a true variable-length big-endian integer in the near future, so the tag value will be stored in the last 4 bytes of payloads rather than the first 4 bytes like they are on little-endian systems.
…iteral, indicating the failure reason.
…quest Sema: implement `existentialConformsToSelf` using a request evaluator.
Within invalid code, we might encounter expressions without type information yet. Check for NULL here. Fixes crash from rdar://problem/53120878.
Fixes issue SR-10571 and corresponding rdar://problem/50337188.
…s to the evaluator model These APIs are 'canDeclProvideDefaultImplementationFor' and 'collectAllOverriddenDecls'.
…f6e47d8ce5eb00 [benchmark][readme] Add docs on how to edit the benchmarks in xcode/b…
…ecls IDE+Evaluator: refactor the implementation of two common IDE utilities to the evaluator model
…x wrapped property refereces Fixes SR-11149 / rdar://problem/53204113
I needed this functionality to fix an edge case bug in closure lifetime fixup.
Given certain CFGs, the SSA updater will insert unnecessary phi arguments that
all take the same "initial" value (the .none from the entry block). As an
example, consider the following CFG:
```
+-------+ +------+
| 2 | <-- | 0 |
+-------+ +------+
| |
| |
| v
| +------+ +---+
| | 1 | --> | 3 |
| +------+ +---+
| | |
| | |
| v |
| +------+ |
| | 4 | |
| +------+ |
| | |
| | |
| v |
| +------+ |
+---------> | 5 | |
+------+ |
| |
| |
v |
+-------+ +------+ |
| Throw | <-- | 6 | |
+-------+ +------+ |
| |
| |
v |
+------+ |
| Exit | <-----+
+------+
```
In this case, if our some value is actually in BB3, we will actually insert a
phi node in BB5 that has all .none incoming values. This is causing us from the
perspective of the ownership verifier to be leaking that phi node in
BBThrow. The reason why is that arguments are a location in SIL where the
specific case of the enum is erased, forcing us to use @owned if the enum has
any non-trivial cases. These of course will not be balanced by any
destroy_values reachable from the copy_value (since otherwise, we would have a
.some definition along that path), so it makes sense. Rather than try to insert
destroy_values, we just realize that in this case we know that all the incoming
values are really the .none from the entry block and delete the argument,
eliminating the problem.
Use name mangling to disambiguate field names when types are anonymous.
…alidate-args [Qol] Driver toolchain validate args
[Type checker] Substitute into alternate properties when trying to fix wrapped property references
…4959d83d1bf8bd [closure-lifetime-fixup] Delete dead trivial phi nodes.
…ted types.
The code here was not correct in a situation where an opaque type had constraints that were
refinements of the protocol requirements of an associated type, as in:
```
protocol ParentProtocol {}
protocol SubProtocol: ParentProtocol {}
protocol P {
associatedtype A: ParentProtocol
func foo() -> A
}
struct S: P {
func foo() -> some SubProtocol
}
```
because it assumed that the conformance could be found directly on the opaque type instead of
potentially via an arbitrary MetadataPath. Falling through to the code that already correctly
handles archetype conformances right below the removed code does the right thing. Fixes
rdar://problem/53081207.
…s-protocol-refinement IRGen: Use correct archetype conformance code path for opaque associated types.
…InMiddleOfBlock for getting values in exits. Otherwise we will leak values if the copy block is in the same block as the terminator and we fail to identify a single destroy for the copy block. We are generally pretty good at identifying those single destroy cases, so I am not sure how often this actually happens. Found by the ownership verifier.
…467a74540e4a9c [closure-lifetime-fixup] Use GetValueAtEndOfBlock instead of GetValue…
…05b074b064e388 Revert "Revert "[ownership] Verify functions before we strip ownershi…
#26255) Unresolved imports all have the "same" ModuleDecl: nullptr. But that's not really an inconsistency, and even if there *is* a true inconsistency among unresolved imports, it can be dealt with when the developer fixes their search paths. (Or the imports. Whichever is wrong.) rdar://problem/52943397
…erge Tag build swift-DEVELOPMENT-SNAPSHOT-2019-07-22-a
|
@swift-ci please clean test tensorflow |
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.