Apps don't weak-link our overlays today, so if the overlays strong-link
to frameworks apps won't launch on older OSs. This was causing failures
on the buildbot, but is actually a serious issue for our developers.
Swift SVN r31180
This allows us to remove the workaround generic overloads for "print" without sacrificing good diagnostics for its unavailable overloads. (rdar://problem/21499048, rdar://problem/21582758, rdar://problem/22126141)
Swift SVN r31169
This inefficient call took up 8-9% of the compile time of Global ARC Opts when
compiling a release no-assert stdlib with a release compiler with assertions.
NFC.
<rdar://problem/22244924>
Swift SVN r31167
Like decodeTopLevelObjectOfClass(_:forKey:), this API works very nicely
as a generic method in Swift, and this one is actually the one we expect
to be commonly used. One thing to note here is that these methods are
stricter than their ObjC counterparts: they will do a forced checked cast
even when the unarchiver does not use "secure" coding.
This depends on the previous commit; without it, we do not actually
enforce type safety for these methods.
The API notes change is to make the non-generic version of this method
unavailable so that it does not participate in overload resolution.
Without this we prefer the non-generic method unless there's a contextual
type for the result. I've filed rdar://problem/22243198 to track taking
this out once Foundation has updated their headers.
rdar://problem/17060110 (again)
Swift SVN r31154
containsString and localizedCaseInsensitiveContainsString were
introduced in 10.10, release-noted, but never
documented (<rdar://22236574>), so we missed them.
Fixes <rdar://18776075> String.containsString doesn't work in Swift
Swift SVN r31152
dealloc_ref [destructor] is the existing behavior. It expects the
reference count to have reached zero and the isDeallocating bit to
be set.
The new [constructor] variant first drops the initial strong
reference.
This allows DI to properly free uninitialized instances in
constructors. Previously this would fail with an assertion if the
runtime was built with debugging enabled.
Progress on <rdar://problem/21991742>.
Swift SVN r31142
the regressions that r31105 introduced in the validation tests, as well as fixing a number
of other validation tests as well.
Introduce a new UnresolvedType to the type system, and have CSDiags start to use it
as a way to get more type information out of incorrect subexpressions. UnresolvedType
generally just propagates around the type system like a type variable:
- it magically conforms to all protocols
- it CSGens as an unconstrained type variable.
- it ASTPrints as _, just like a type variable.
The major difference is that UnresolvedType can be used outside the context of a
ConstraintSystem, which is useful for CSGen since it sets up several of them to
diagnose subexpressions w.r.t. their types.
For now, our use of this is extremely limited: when a closureexpr has no contextual
type available and its parameters are invalid, we wipe them out with UnresolvedType
(instead of the previous nulltype dance) to get ambiguities later on.
We also introduce a new FreeTypeVariableBinding::UnresolvedType approach for
constraint solving (and use this only in one place in CSDiags so far, to resolve
the callee of a CallExpr) which solves a system and rewrites any leftover type
variables as UnresolvedTypes. This allows us to get more precise information out,
for example, diagnosing:
func r22162441(lines: [String]) {
lines.map { line in line.fooBar() }
}
with: value of type 'String' has no member 'fooBar'
instead of: type of expression is ambiguous without more context
This improves a number of other diagnostics as well, but is just the infrastructural
stepping stone for greater things.
Swift SVN r31130
At -O I see about 10% improvement on Havlak, and 40% improvement on the recently
added array append benchmarks.
COWTree regresses by 15% because we hoist an append that is conditionally
executed. If we were to disallow hoisting of conditionally executed benchmarks
we would see big loses accross our benchmarks (for example, we would not hoist
out of nested loops anymore). So we will have to live with this regression.
rdar://17140639
Swift SVN r31120
This reverts commit cd3f1ba7d1ee2397817e1a165209fdeab8a1c004.
Reverting this b/c it is breaking buildbots with the following:
CMake Error at CMakeLists.txt:522 (message):
CrashReporterClient library is required, but it was not found
Swift SVN r31047
The print abi changed so that there is no longer a way to print one item to a
stream without going through a vararg array. We call the print function in a
String initializer (e.g for string interpolation) incurring unnecessary
overhead. Use an internal specialized version of print instead.
rdar://22133281
Swift SVN r31000
This commit addresses:
<rdar://problem/22121657> stdlib/Slice.swift.gyb test failing on swift_tools-RA_stdlib-RDA_test-ios_device since #1325
Swift SVN r30997
- Add Strict/Defaulted Index types to StdlibUnittest
- Test whether a random access index calls its more efficient
customization by tracking successor calls.
- Fix the RandomAccessIndex.advancedBy(n, limit:) API by de-underscoring
the limit parameter
- Inline some internal transparent default implementations to their only
call site
- Attach _RandomAccessAmbiguity type to RandomAccessIndex
rdar://problem/22085119
Swift SVN r30979
The parse_stdlib tests choke on the constraints that were there previously,
so weaken the API to only work on intervals of the same type for now.
rdar://problem/22126470
Swift SVN r30977
There's still work left to do. In terms of next steps, there's still rdar://problem/22126141, which covers removing the 'workaround' overloads for print (that prevent bogus overload resolution failures), as well as providing a decent diagnostic when users invoke print with 'appendNewline'.
Swift SVN r30976
- Don't repeat type names in methods and properties
- Use 'InPlace' suffix for mutating variants of methods that return
a CG type.
rdar://problem/19788137
Swift SVN r30960
- Remove free Swift functions for advance and distance and replace
them with protocol extension methods:
- advancedBy(n)
- advancedBy(n, limit:)
- distanceTo(end)
- Modernize the Index tests
- Use StdlibUnittest
- Test for custom implementation dispatch
Perf impact: No significant changes reported in the
Swift Performance Measurement Tool.
rdar://problem/22085119
Swift SVN r30958
(or in 'randomShuffle')
This violates our inout aliasing rules, and can cause a crash in certain
situations.
To try to prevent this in the future I've added a debugPrecondition check
to 'swap' that checks addresses. This will only catch inout violations
where neither argument uses writeback, and might some day be optimized
away. Right now, though, it does seem to catch some simple violations
(though not the one in the Radar), so hopefully it will at least help keep
us from introducing new problems.
rdar://problem/21780567
Swift SVN r30956
...replacing it with the new, after passing API review!
* The lazy free function has become a property.
* Before we could extend protocols, we lacked a means for value types to
share implementations, and each new lazy algorithm had to be added to
each of up to four types: LazySequence, LazyForwardCollection,
LazyBidirectionalCollection, and LazyRandomAccessCollection. These
generic adapters hid the usual algorithms by defining their own
versions that returned new lazy generic adapters. Now users can extend
just one of two protocols to do the same thing: LazySequenceType or
LazyCollectionType.
* To avoid making the code duplication worse than it already was, the
generic adapters mentioned above were used to add the lazy generic
algorithms around simpler adapters such as MapSequence that just
provided the basic requirements of SequenceType by applying a
transformation to some base sequence, resulting in deeply nested
generic types as shown here. Now, MapSequence is an instance of
LazySequenceType (and is renamed LazyMapSequence), and thus transmits
laziness to its algorithms automatically.
* Documentation comments have been rewritten.
* The .array property was retired
* various renamings
* A bunch of Gyb files were retired.
Swift SVN r30902
