This fixes a non-determinism during deserialization:
The constructor of SerializedSILLoader iterates over ASTContext::LoadedModules.
If a function is contained in multiple loaded modules, it was non deterministic which version was deserialized.
Fixes rdar://problem/20992687.
Swift SVN r28704
When a derived class specializes its base class, e.g. 'class Derived: Base<Int>', the natural abstraction levels of its methods may differ from the original base class's more abstract methods. Handle this by using the reabstraction machinery to thunk values when necessary. Merge the existing optionality thunking support into the reabstraction code, where witness thunking and similar convention adjustments may also be able to use it, if we desire. Fixes rdar://problem/19760292.
Swift SVN r28505
Reverts r28087. We're going back to the C++ interface for SIMD, and the changes in this patch are needless complication for that design.
Swift SVN r28384
Modules occupy a weird space in the AST now: they can be treated like
types (Swift.Int), which is captured by ModuleType. They can be
treated like values for disambiguation (Swift.print), which is
captured by ModuleExpr. And we jump through hoops in various places to
store "either a module or a decl".
Start cleaning this up by transforming Module into ModuleDecl, a
TypeDecl that's implicitly created to describe a module. Subsequent
changes will start folding away the special cases (ModuleExpr ->
DeclRefExpr, name lookup results stop having a separate Module case,
etc.).
Note that the Module -> ModuleDecl typedef is there to limit the
changes needed. Much of this patch is actually dealing with the fact
that Module used to have Ctx and Name public members that now need to
be accessed via getASTContext() and getName(), respectively.
Swift SVN r28284
Currently GenericSignature::getCanonicalSignature isn't able to canonicalize the set of requirements due to fragile dependencies on generic signatures matching AllArchetypes order of their originating GenericParamLists. However, we shouldn't let that stop us from getting the mangling right, so implement a "getCanonicalManglingSignature" that builds the true canonical signature by feeding it into an ArchetypeBuilder and shedding unnecessary constraints. For now, just handle conformance and base class constraints; still to do are same-type constraints.
Swift SVN r28191
The design we landed on for SIMD is to define the vector types as nested types of their element, e.g. Float.Vector4, Int32.Vector2, etc. Update the Clang importer and other mapping facilities to match.
Swift SVN r28087
Replace ReST-flavored documentation comments with Markdown.
rdar://problem/20180412
In addition to full Markdown support, the following extensions are
supported. These appear as lists at the top level of the comment's
"document". All of these extensions are matched without regard to
case.
Parameter Outlines
------------------
- Parameters:
- x: ...
- y: ...
Separate Parameters
-------------------
- parameter x: ...
- parameter y: ...
- Note:
Parameter documentation may be broken up across the entire comment,
with a mix of parameter documentation kinds - they'll be consolidated
in the end.
Returns
-------
- returns: ...
The following extensions are also list items at the top level, which
will also appear in Xcode QuickHelp as first-class citizens:
- Attention: ...
- Author: ...
- Authors: ...
- Bug: ...
- Complexity: ...
- Copyright: ...
- Date: ...
- Experiment: ...
- Important: ...
- Invariant: ...
- Note: ...
- Postcondition: ...
- Precondition: ...
- Remark: ...
- Remarks: ...
- See: ...
- Since: ...
- Todo: ...
- Version: ...
- Warning: ...
These match most of the extra fields in Doxygen, plus a few more per request.
Other changes
-------------
- Remove use of rawHTML for all markup AST nodes except for those
not representable by the Xcode QuickHelp XSLT - <h>, <hr/>, and of
course inline/block HTML itself.
- Update the doc comment RNG schema to more accurately reflect Xcode
QuickHelp.
- Clean up cmark CMake configuration.
- Rename "FullComment" to "DocComment"
- Update the Swift Standard Documentation (in a follow-up commit)
- Update SourceKit for minor changes and link against cmark
(in a follow-up commit).
Swift SVN r27727
When we're checking for a particular witness in a protocol conformance
because the result is needed elsewhere, capture the diagnostic we
would emit and then store it in the ASTContext. We will emit these
diagnostics when doing the full check of the conformance. Addresses
the rest of rdar://problem/20564378.
Swift SVN r27449
Move the map that keeps track of conforming decl -> requirement from ASTContext
to a nominal type's ConformanceLookupTable, and populate it lazily.
This allows getSatisfiedProtocolRequirements() to work with declarations from module files.
Test on the SourceKit side.
Part of rdar://20526240.
Swift SVN r27353
Now that we can pick up search paths from frameworks (necessary to debug
them properly), we can end up with exponential explosions leading to the
same search path coming up thousands of times, which destroys compilation
time /and/ debugger responsiveness. This is already hitting people with
frameworks compiled for app extensions (due to a mistaken approximation
of whether or not something is a framework), but we're turning this on for
all frameworks in the immediate future.
rdar://problem/20291720
Swift SVN r27087
Have TypeChecker's constructor register itself as the lazy resolver,
and its destructor unregister itself. This introduces the
completely-sensible restriction that there can only be one type
checker active for an ASTContext at a time, which is the case already.
Use ASTContext's lazy resolver in a single place that's been causing
trouble (rdar://problem/20363958, rdar://problem/19773096), where
deserializing a protocol conformance can cause us to pass a null lazy
resolver into the protocol conformance table, which doesn't handle it
well. This commit fixes those issues, which I'm unable to reduce down
to a sane-enough test case to commit.
This commit implies a ton of cleanup work to eliminate LazyResolver
parameters from *everywhere*, deriving them from the ASTContext in the
few places they're needed as well. It's a good direction, but that
cleanup can be evolutionary.
Swift SVN r26824
This tool takes the input of two versions of the same sdk and outputs
their diff to facilicate the auto-migration of sdk clients.
In this initial commit, we take the path of one sdk and generate
a tree describing its API content. Next, we will diff trees generated
from different versions of the sdk.
In addition, this commit also refactored out part of swift-sdk-analyzer
to the common utils shared with swift-sdk-digester.
Swift SVN r26656
Stop storing a conformances array on ExtensionDecls. Instead, always use the conformance lookup table to retrieve conformances (which is lazy and supports multi-file, among other benefits).
As part of this, space-optimize ExtensionDecl's handling of conformance loaders. When one registers a conformance loader, it goes into a DenseMap on ASTContext and gets erased once we've loaded that data, so we get two words worth of space back in each ExtensionDecl.
Swift SVN r26353
Previously, we would require the type checker to be able to build a
conformance, which meant we would actually have to lie in the AST
about having a conformance (or crash; we did the form). Now, we can
form the conformance in the AST and it will be checked in the type
checker when needed. The intent here is to push conformance creation
into the conformance lookup table.
To get here, we had to stop relying on the broken, awful,
ASTContext-wide conformance "cache". A proper cache can come back once
the model is sorted out.
Swift SVN r26250
Previously, a multi-pattern var/let decl like:
var x = 4, y = 17
would produce two pattern binding decls (one for x=4 one for y=17). This is convenient
in some ways, but is bad for source reproducibility from the ASTs (see, e.g. the improvements
in test/IDE/structure.swift and test/decl/inherit/initializer.swift).
The hardest part of this change was to get parseDeclVar to set up the AST in a way
compatible with our existing assumptions. I ended up with an approach that forms PBDs in
more erroneous cases than before. One downside of this is that we now produce a spurious
"type annotation missing in pattern"
diagnostic in some cases. I'll take care of that in a follow-on patch.
Swift SVN r26224
The problem here was that the _preconditionImplicitlyUnwrappedOptionalHasValue
compiler intrinsic was taking the optional/IUO argument as inout as a performance
optimization, but DI would reject it (in narrow cases, in inits) because the inout
argument looks like a mutation.
We could rework this to take it as an @in argument or something, but it is better
to just define this problem away: the precondition doesn't actually care about the
optional, it is just testing its presence, which SILGen does all the time. Have
SILGen open code the switch_enum and just have the stdlib provide a simpler
_diagnoseUnexpectedNilOptional() to produce the error message.
This avoids the problem completely and produces slightly better -O0 codegen.
Swift SVN r25254
An optional @objc requirement within a protocol can be left
unsatisfied in a well-formed program. However, there may still be a
conflict within the Objective-C runtime if the conforming class
defines a method with the corresponding Objective-C selector(s) for
that requirement, which means that the Swift and Objective-C semantics
will differ. Diagnose such issues.
More steps along the road to fixing rdar://problem/18383574.
Diagnose conflicts between unsatisfied, optional @objc requirements and
Swift SVN r24830
They don't work properly, and if we want eager static initialization,
we'll add a Swift feature for it. Fixes rdar://problem/18423731.
Swift SVN r24814
storage for arbitrary values.
A buffer doesn't provide any way to identify the type of
value it stores, and so it cannot be copied, moved, or
destroyed independently; thus it's not available as a
first-class type in Swift, which is why I've labelled
it Unsafe. But it does allow an efficient means of
opaquely preserving information between two cooperating
functions. This will be useful for the adjustments I
need to make to materializeForSet to support safe
addressors.
I considered making this a SIL type category instead,
like $@value_buffer T. This is an attractive idea because
it's generally better-typed. The disadvantages are that:
- it would need its own address_to_pointer equivalents and
- alloc_stack doesn't know what type will be stored in
any particular buffer, so there still needs to be
something opaque.
This representation is a bit gross, but it'll do.
Swift SVN r23903
Include a mapping from Objective-C selectors to the @objc methods that
produce Objective-c methods with those selectors. Use this to lazily
populate the Objective-C method lookup tables in each class. This makes
@objc override checking work across Swift modules, which is part of
rdar://problem/18391046.
Note that we use a single, unified selector table, both because it is
simpler and because it makes global queries ("is there any method with
the given selector?") easier.
Swift SVN r23214
Diagnose cases where the use of @objc will produce Objective-C methods
that end up overriding an Objective-C method in a superclass, when
that override is not properly represented as an override in the Swift
type system. This can happen when the Objective-C methods are produced
by different kinds of entities. For example:
class Super {
@objc var property: Int
}
class Sub : Super {
@objc func setProperty(property: Int) { }
}
In Swift, Sub.setProperty and Super.property are completely
unrelated. However, both produce an Objective-C instance method with
the selector "setProperty:", so we end up with unexpected overriding
behavior. Diagnose this whenever it occurs, regardless of the kind of
@objc entity that produced the Objective-C methods: initializers,
deinitializers, methods, properties, or subscripts.
Implements the rest of the intended functionality of
rdar://problem/18391046, with the caveat that there are two remaining
classes of bugs:
1) Superclasses defined in a module (or imported from a Clang
module) aren't handled properly yet; we might not see those methods.
2) We won't properly detect all of these failures when the methods
are scattered across different source files in the same module.
Swift SVN r23170
@objc methods, initializers, deinitializers, properties, and
subscripts all produce Objective-C methods. Diagnose cases where two
such entities (which may be of different kinds) produce the same
Objective-C method in the same class.
As a special exception, one can have an Objective-C method in an
extension that conflicts with an Objective-C method in the original
class definition, so long as the original class definition is from a
different model. This reflects the reality in Objective-C that the
category definition wins over the original definition, and is used in
at least one overlay (SpriteKit).
This is the first part of rdar://problem/18391046; the second part
involves checking that overrides are sane.
Swift SVN r23147
Before it was done with a big switch statement, which remained a switch until the final code.
Now it's done by getting an integer index for both enums and just doing an integer compare.
This results in a simple compare in the final code.
Note that the == function is only generated for enums without payload. Getting the integer
index of such enums is a cheap operation.
Swift SVN r23129
This is a type that has ownership of a reference while allowing access to the
spare bits inside the pointer, but which can also safely hold an ObjC tagged pointer
reference (with no spare bits of course). It additionally blesses one
Foundation-coordinated bit with the meaning of "has swift refcounting" in order
to get a faster short-circuit to native refcounting. It supports the following
builtin operations:
- Builtin.castToBridgeObject<T>(ref: T, bits: Builtin.Word) ->
Builtin.BridgeObject
Creates a BridgeObject that contains the bitwise-OR of the bit patterns of
"ref" and "bits". It is the user's responsibility to ensure "bits" doesn't
interfere with the reference identity of the resulting value. In other words,
it is undefined behavior unless:
castReferenceFromBridgeObject(castToBridgeObject(ref, bits)) === ref
This means "bits" must be zero if "ref" is a tagged pointer. If "ref" is a real
object pointer, "bits" must not have any non-spare bits set (unless they're
already set in the pointer value). The native discriminator bit may only be set
if the object is Swift-refcounted.
- Builtin.castReferenceFromBridgeObject<T>(bo: Builtin.BridgeObject) -> T
Extracts the reference from a BridgeObject.
- Builtin.castBitPatternFromBridgeObject(bo: Builtin.BridgeObject) -> Builtin.Word
Presents the bit pattern of a BridgeObject as a Word.
BridgeObject's bits are set up as follows on the various platforms:
i386, armv7:
No ObjC tagged pointers
Swift native refcounting flag bit: 0x0000_0001
Other available spare bits: 0x0000_0002
x86_64:
Reserved for ObjC tagged pointers: 0x8000_0000_0000_0001
Swift native refcounting flag bit: 0x0000_0000_0000_0002
Other available spare bits: 0x7F00_0000_0000_0004
arm64:
Reserved for ObjC tagged pointers: 0x8000_0000_0000_0000
Swift native refcounting flag bit: 0x4000_0000_0000_0000
Other available spare bits: 0x3F00_0000_0000_0007
TODO: BridgeObject doesn't present any extra inhabitants. It ought to at least provide null as an extra inhabitant for Optional.
Swift SVN r22880
This reverts commit r22829, because reverting r22828 depends on it.
Reverting r22828 because it
was apparently causing an assertion on the bot:
Swift SVN r22830
llvm::Optional lives in "llvm/ADT/Optional.h". Like Clang, we can get
Optional in the 'swift' namespace by including "swift/Basic/LLVM.h".
We're now fully switched over to llvm::Optional!
Swift SVN r22477
This patch adds SILGen for the API availability (#os(...)) construct. To do so,
it (1) adds version range information to the AvailabilityQueryExpr AST in Sema,
during type refinement context construction; and (2) uses that version range,
during SILGen, to emit a call to the standard library's
_stdlib_isOSVersionAtLeast function.
Swift SVN r22348
This avoids a pointless copy every time an array literal is written, and will let us retire the horrible "alloc_array" instruction and globs of broken IRGen code. Implements rdar://problem/16386862, and probably fixes a bunch of bugs related to alloc_array brokenness.
Swift SVN r22289
Like the Clang decls, this happens at the end of the type-checking, just as
a simple walk through the loaded decls of the loaded modules. This caught
all of the issues in this commit series and will hopefully keep us honest in
the future.
(By the way, we don't verify right when we return a deserialized decl for the
same reason we don't verify right when we return an imported decl: parts of
the decl may be delayed, and (a) we don't want to force things to be imported
or deserialized sooner than necessary, yet (b) we want to verify as much as
possible.)
rdar://problem/16968891
Swift SVN r22200
While we work out the remaining performance improvements in the type checker, we can improve the user experience for some "runaway solver" bugs by setting a limit on the amount of temporary memory allocated for type variables when solving over a single expression.
Exponential behavior usually manifests itself while recursively attempting bindings over opened type variables in an expression. Each one of these bindings may result in one or more fresh type variables being created. On average, memory consumption by type variables is fairly light, but in some exponential cases it can quickly grow to many hundreds of megabytes or even gigabytes. (This memory is managed by a distinct arena in the AST context, so it's easy to track.) This problem is the source of many of the "freezing" compiler and SourceKit bugs we've been seeing.
These changes set a limit on the amount of memory that can be allocated for type variables while solving for a single expression. If the memory threshold is exceeded, we can surface a type error and suggest that the user decompose the expression into distinct, less-complex sub-expressions.
I've set the current threshold to 15MB which, experimentally, avoids false positives but doesn't let things carry on so long that the user feels compelled to kill the process before they can see an error message. (As a point of comparison, the largest allocation of type variable data while solving for a single expression in the standard library is 592,472 bytes.) I've also added a new hidden front-end flag, "solver-memory-threshold", that will allow users to set their own limit, in bytes.
Swift SVN r20986
