This matches how dispatch_once works in C, dramatically cutting the cost of a global accessor by avoiding the runtime call in the hot path and giving the global a unique branch for the CPU to predict away. For now, only do this for Darwin; non-ObjC platforms don't necessarily expose their "done" value as ABI like ours do.
While we're here, change "once" to take a thin function pointer. We don't ever emit global initializers with context dependencies, and this simplifies the runtime glue between swift_once and dispatch_once/std::call_once a bit.
Swift SVN r28166
Like the FIXME comment states, this method should probably
be changed to be:
public func getMirror() -> MirrorType {
return _OptionalMirror<T>(self)
}
but I don't know how to test that change.
Swift SVN r28157
We had failed to recognize that legacy ObjC class mirrors have a
distinct type. Fixes <rdar://20811737>
However, there's a bug in legacy mirrors that prevents all base classes
from being properly reflected: <rdar://20813711>
Swift SVN r28156
Two pieces to this:
- Peephole OptionalEvaluationExpr(InjectOptionalExpr(BindOptionalExpr(X))) to bitcast x to the result type.
- Enhance OptionalEvaluationExpr to delete the failure block if not needed.
This is the same as r28150, but it includes a fix for the case when a non-address-only type
is initializing a contextally-provided-and-addressible buffer, tested by the new
testContextualInitOfNonAddrOnlyType testcase.
Swift SVN r28153
Two pieces to this:
- Peephole OptionalEvaluationExpr(InjectOptionalExpr(BindOptionalExpr(X))) to
bitcast x to the result type.
- Enhance OptionalEvaluationExpr to delete the failure block if not needed.
This is the same as r28111, except that we finalize the initialization in the
address-only case. A reduced testcase for the specific issue is added to
optional-casts.swift.
Swift SVN r28150
<rdar://20494686>
String itsef should only expose Unicode-correct algorithms, like proper
substring/prefix/suffix search, enumerating words/lines/paragraphs, case
folding etc. Promoting sequence-centric algorithms to methods on String
is not acceptable since it invites users to write wrong code. Thus,
String has to lose its SequenceType conformance.
Nevertheless, we recognize that sometimes it is useful to manipulate the
String contents on lower levels (UTF-8, UTF-16, Unicode scalars,
extended grapheme clusters), for example, when implementing high-level
Unicode operations, so we can't remove low-level operations
altogether. For this reason, String provides nested "views" for the
first three low-level representations, but grapheme clusters were in a
privileged position -- String itself is a collection of grapheme
clusters. We propose to add a characters view that will represent the
String as a collection of Character values.
Swift SVN r28065
Currently they do nothing but allow stdlib code to use regular (Bool)
types. However, soon the wrappers for the _native variants will
provide point-of-use sanity checking.
These need to be fully generic to support class protocols and
single-payload enums (not just for optional). It also avoids a massive
amount of overloading for all the reference type variations
(AnyObject, Native, Unknown, Bridge) x 2 for optional versions of
each.
Because the wrapper is generic, type checking had to be deferred until
IRGen. Generating code for the wrapper itself will result in an
IRGen-time type error. They need to be transparent anyway for proper
diagnostics, but also must be internal.
Note that the similar external API type checks ok because it
forces conformance to AnyObject.
The sanity checks are disabled because our current facilities for
unsafe type casting are incomplete and unsound. SILCombine can
remove UnsafeMutablePointer and RawPointer casts by assuming layout
compatibility. IRGen will later discover layout incompatibility and
generate a trap.
I'll send out a proposal for improving the casting situation so we can
get the sanity checks back.
Swift SVN r28057
and link it properly. This is needed to embed LLVM bitcode sections
in the standard library and overlays. The linker doesn't support
embedded bitcode and reexport flags (among others).
rdar://problem/20750099
Swift SVN r28052
Change _stdlib_isOSVersionAtLeast on tvOS to actually check the running
OS version rather than always returning false. It still always returns false on
watchOS; rdar://problem/20234735 tracks that.
rdar://problem/20663303
Swift SVN r28024
