This feature is required for the implementation of pre-specialization, because one needs to check if a specialized SIL function with a given name exists in the standard library.
Swift SVN r30307
The old code misused clang::CompilerInstance::createOutputFile (which should
really be sunk down to LLVM). That API specifically handles the case where you
can write to the final output file, but not to anything else in that directory.
I'm fixing this (and cleaning up the code) because of a copy/paste error Dmitri
noticed, where the temp file for the .swiftdoc output would be created as if it
were for the .swiftmodule file. This is also now fixed, and the copy/paste code
has been refactored.
This change should only really affect the case described above, so it's not urgent.
Swift SVN r30206
This flag is required for performing the propagation of global and static "let" values into their uses.
Let variables have now a [let] attribute in the SIL textual form.
Swift SVN r30153
Add a new convention to describe what happens with
nonzero_result on a type that isn't imported as Bool.
This isn't really a safe convention to implement, but
calls are fine.
Implements <rdar://21715350>.
Swift SVN r29953
Compiler output at least up to serialization should be deterministic at this point,
at least when not taking SIL into account. This /should/ mean that changing a
function body should not affect the final built swiftmodule, which means downstream
targets don't need to be rebuilt. Leaving the previous swiftmodule output in place
signals that.
A while back I put in a push to get all the non-determinism out of type checking,
importing, and serialization itself; it looks like we've finally made it. Let's keep
it that way!
rdar://problem/20539158 and others
Swift SVN r29923
We need a SIL level unsafe cast that supports arbitrary usage of
UnsafePointer, generalizes Builtin.reinterpretCast, and has the same
semantics on generic vs. nongeneric code. In other words, we need to
be able to promote the cast of an address type to the cast of an
object type without changing semantics, and that cast needs to support
types that are not layout identical.
This patch introduces an unchecked_bitwise_cast instruction for that
purpose. It is different from unsafe_addr_cast, which has been our
fall-back "unknown" cast in the past. With unchecked_bitwise_cast we
cannot assume layout or RC identity. The cast implies a store and
reload of the value to obtain the low order bytes. I know that
bit_cast is just an abbreviation for bitwise_cast, but we use
"bitcast" throught to imply copying a same sized value. No one could
come up with a better name for copying an objects low bytes via:
@addr = alloca $wideTy
store @addr, $wideTy
load @addr, $narrowTy
Followup patches will optimize unchecked_bitwise_cast into more
semantically useful unchecked casts when enough type information is
present. This way, the optimizer will rarely need to be taught about
the bitwise case.
Swift SVN r29510
Still no implementation yet; we'll need to renovate how boxes work a bit to make them projectable (and renovate SILGen to generate typed boxes for the insn to be useful).
Swift SVN r29490
Represents a heap allocation containing a value of type T, which we'll be able to use to represent the payloads of indirect enum cases, and also improve codegen of current boxes, which generates non-uniqued box metadata on every allocation, which is dumb. No codegen changes or IRGen support yet; that will come later.
This time, fix a paste-o that caused SILBlockStorageTypes to get replaced with SILBoxTypes during type substitution. Oops.
Swift SVN r29489
This reverts commit r29475 because it conflicts with reverting r29474,
and it looks like that commit is breaking the build of the SpriteKit
overlay.
Swift SVN r29481
Still no implementation yet; we'll need to renovate how boxes work a bit to make them projectable (and renovate SILGen to generate typed boxes for the insn to be useful).
Swift SVN r29475
Represents a heap allocation containing a value of type T, which we'll be able to use to represent the payloads of indirect enum cases, and also improve codegen of current boxes, which generates non-uniqued box metadata on every allocation, which is dumb. No codegen changes or IRGen support yet; that will come later.
Swift SVN r29474
That's how everything behaved anyway. Might as well make it explicit and
stop special-casing it.
I've left in compatibility for modules built with older compilers so that
people using the OS toolchains aren't immediately unable to debug their apps.
As soon as we change the module format in a more significant way, I can take
this out.
Groundwork for rdar://problem/21254367; see next commit.
Swift SVN r29437
initializer has been type-checked, rather than a bit for the entire
PatternBindingDecl.
<rdar://problem/21057425> Crash while compiling attached test-app.
Swift SVN r29049
Based on Dave’s hack, this allows one to define a “default implementation” as, e.g.,
protocol P {
func foo()
}
extension P {
final func foo() { … }
}
Swift SVN r28949
This is more complex than it could be if ExtensionDecl and NominalTypeDecl
had a common ancestor in the Decl hierarchy, however this is not possible
right now because TypeDecl inherits from ValueDecl.
Fixes <rdar://problem/20981254>.
Swift SVN r28941
This shouldn't affect anything in practice but it's best to be deterministic.
(Although I'm not sure why the previous mode was nondeterministic.)
Swift SVN r28580
Unfortunately, we still have non-determinism coming from elsewhere, so we
can't start trying to test this yet.
Motivated by rdar://problem/20539158
Swift SVN r28576
This came out of today's language review meeting.
The intent is to match #available with the attribute
that describes availability.
This is a divergence from Objective-C.
Swift SVN r28484
Now that we don't have generic parameter lists at arbitrary positions
within the extended type of an extension declaration, simplify the
representation of the extended type down to a TypeLoc along with a
(compiler-synthesized) generic parameter list.
On the parsing side, just parse a type for the extended type, rather
than having a special grammar. We still reject anything that is not a
nominal type (of course), but it's simpler just to call it a type.
As a drive-by, fix the crasher when extending a type with module
qualification, rdar://problem/20900870.
Swift SVN r28469
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
When reading the generic parameters of a constrained protocol
extension, cross-refencing an associated type would perform name
lookup into the protocol extension itself, causing fatal recursion
during deserialization. Fixed by avoiding additional deserialization
when looking for an associated type. Fixes rdar://problem/20812303.
Swift SVN r28228
Rather than swizzle the superclass of these bridging classes at +load time, have the compiler set their ObjC runtime base classes, using a "@_swift_native_objc_runtime_base" attribute that tells the compiler to use a different implicit base class from SwiftObject. This lets the runtime shed its last lingering +loads, and should overall be more robust, since it doesn't rely on static initialization order or deprecated ObjC runtime calls.
Swift SVN r28219
Otherwise, when a file contains a local enum, the decl for its synthesized
== ends up disappearing, and we get a dangling XREF.
rdar://problem/20429123
Swift SVN r28131
preserve the original method name.
This heuristic is based on the Objective-C selector and therefore
doesn't really handle factory methods that would conflict with
initializers, but we can hope that those simply don't come up in
the wild.
It's not clear that this is the best thing to do --- it tends to
promote the non-throwing API over what's probably a newer, throwing
API --- but it's significantly easier, and it unblocks code without
creating deployment problems.
Swift SVN r28066
@warn_unused_result can be attached to function declarations to
produce a warning if the function is called but its result is not
used. It has two optional parameters that can be placed in
parentheses:
message="some message": a message to include with the warning.
mutable_variant="somedecl": the name of the mutable variant of the
method that should be suggested when the subject method is called on
a mutable value.
The specific use we're implementing this for now is for the mutating
and in-place operations. For example:
@warn_unused_result(mutable_variant="sortInPlace") func sort() -> [Generator.Element] { ... }
mutating func sortInPlace() { ... }
Translate Clang's __attribute__((warn_unused_result)) into
@warn_unused_result.
Implements rdar://problem/18165189.
Swift SVN r28019
When deserializing a protocol, the conformance lookup table would not
contain entries for the inherited protocols of that protocol. They
were stashed in the "Protocols" array in TypeDecl (which will
eventually go away), but since there are no conformances for a
protocol, the conformance lookup table never got updated.
Nothing important seems to query this now; that will change soon.
Swift SVN r27967
Preparation to fix <rdar://problem/18151694> Add Builtin.checkUnique
to avoid lost Array copies.
This adds the following new builtins:
isUnique : <T> (inout T[?]) -> Int1
isUniqueOrPinned : <T> (inout T[?]) -> Int1
These builtins take an inout object reference and return a
boolean. Passing the reference inout forces the optimizer to preserve
a retain distinct from what’s required to maintain lifetime for any of
the reference's source-level copies, because the called function is
allowed to replace the reference, thereby releasing the referent.
Before this change, the API entry points for uniqueness checking
already took an inout reference. However, after full inlining, it was
possible for two source-level variables that reference the same object
to appear to be the same variable from the optimizer's perspective
because an address to the variable was longer taken at the point of
checking uniqueness. Consequently the optimizer could remove
"redundant" copies which were actually needed to implement
copy-on-write semantics. With a builtin, the variable whose reference
is being checked for uniqueness appears mutable at the level of an
individual SIL instruction.
The kind of reference count checking that Builtin.isUnique performs
depends on the argument type:
- Native object types are directly checked by reading the
strong reference count:
(Builtin.NativeObject, known native class reference)
- Objective-C object types require an additional check that the
dynamic object type uses native swift reference counting:
(Builtin.UnknownObject, unknown class reference, class existential)
- Bridged object types allow the dymanic object type check to be
bypassed based on the pointer encoding:
(Builtin.BridgeObject)
Any of the above types may also be wrapped in an optional. If the
static argument type is optional, then a null check is also performed.
Thus, isUnique only returns true for non-null, native swift object
references with a strong reference count of one.
isUniqueOrPinned has the same semantics as isUnique except that it
also returns true if the object is marked pinned regardless of the
reference count. This allows for simultaneous non-structural
modification of multiple subobjects.
In some cases, the standard library can dynamically determine that it
has a native reference even though the static type is a bridge or
unknown object. Unsafe variants of the builtin are available to allow
the additional pointer bit mask and dynamic class lookup to be
bypassed in these cases:
isUnique_native : <T> (inout T[?]) -> Int1
isUniqueOrPinned_native : <T> (inout T[?]) -> Int1
These builtins perform an implicit cast to NativeObject before
checking uniqueness. There’s no way at SIL level to cast the address
of a reference, so we need to encapsulate this operation as part of
the builtin.
Swift SVN r27887
Printing a module as Objective-C turns out to be a fantastic way to
verify the (de-)serialization of foreign error conventions, so
collapse the parsing-driving Objective-C printing test of throwing
methods into the general test for methods.
Swift SVN r27880
Printing a module as Objective-C turns out to be a fantastic way to
verify the (de-)serialization of foreign error conventions, so
collapse the parsing-driving Objective-C printing test of throwing
methods into the general test for methods.
Swift SVN r27870
Extensions cannot be uniquely cross-referenced, so cross-references to
extensions are serialized with the extended nominal type name and the
module in which the extension resides. This is not sufficient when
cross-referencing the generic type parameters of a constrained
protocol extension, because we don't know whether to get the
archetypes of the nominal type or some extension thereof. Serialize
the canonical generic signature so that we can pick an extension with
the same generic signature; it doesn't matter which we pick, so long
as we're consistent.
Fixes rdar://problem/20680169. Triggering this involves some
interesting interactions between the optimizer and standard library;
the standard library updates in the radar will test this.
Swift SVN r27825
the printed interface.
Previously we printed the typechecked and uniqued requirements and the result was non-sensical.
Long-term the requirements will be preserved in a better form but for now print the requirements
and serialize them.
rdar://19963093
Swift SVN r27680
reference to something of class type. This is required to model
RebindSelfInConstructorExpr correctly to DI, since in the class case,
self.init and super.init *take* a value out of class box so that it
can pass the +1 value without performing an extra retain. Nothing
else in the compiler uninitializes a DI-controlled memory object
like this, so nothing else needs this. DI really doesn't like something
going from initialized to uninitialized.
Yes, I feel super-gross about this and am really unhappy about it. I
may end up reverting this if I can find an alternate solution to this
problem.
Swift SVN r27525
Allow an unversioned 'deprecated' attribute to specify unconditional
deprecation of an API, e.g.,
@availability(*, deprecated, message="sorry")
func foo() { }
Also support platform-specific deprecation, e.g.,
@availability(iOS, deprecated, message="don't use this on iOS")
func bar() { }
Addresses rdar://problem/20562871.
Swift SVN r27355
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
