It is possible for the SIL optimizers, IRGen, etc. to request information
from the AST that only the type checker can provide, but the type checker
is typically torn down after the “type checking” phase. This can lead to
various crashes late in the compilation cycle.
Keep the type checker instance around as long as the ASTContext is alive
or until someone asks for it to be destroyed.
Fixes SR-285 / rdar://problem/23677338.
On Darwin platforms, libdispatch and libBlocksRuntime are re-exported from
libSystem (via LC_REEXPORT_DYLIB). Other platforms do not have libdispatch and
libBlocksRuntime in their C runtime, so we need to explicitly link against them.
Now that we are building BlocksRuntime with hidden visibility, we do not
accidentally get the symbols from libdispatch.
Module references get indexed as a 'module' symbol; they get USRs similar to how clang would assign a USR for a module reference.
JIRA: https://bugs.swift.org/browse/SR-8677
Addressing review feedback: this avoids calling shared_from_this() from
outside the implementation. Note: it is not possible to use private
inheritance of enable_shared_from_this to prevent this issue in general,
because enabled_shared_from_this relies on the shared_ptr constructor
being able to detect that the type has this inheritance, which means it
must be public.
While I'm not aware of this causing an issue in practice it was just
begging to trigger a use-after-free if we ever used this reference in a
new method since the editor document is accessed asynchronously.
While the SwiftLangSupport doesn't currently outlive the global context
it clarifies the ownership to just store the weak reference we care
about instead of referencing the owning context. In the future we may
want to ref-count the lang support, in which case we'd need to do this
to avoid a cycle anyway.
When the server shuts down we may still have outstanding async work to
build an AST, so use a shared_ptr + weak_ptr instead of unique_ptr +
unowned references.
When the server shuts down we may still have outstanding async work that
can attempt to trigger a notification, so use a shared_ptr + weak_ptr
instead of unique_ptr + unowned references.
When triggering replaceText we were locking the EditorDocument then the
ASTProducer, but during AST building we do the reverse. This would cause
a deadlock if these two bits of code raced. Break the cycle by dropping
the editor lock when calling the ASTManager.
Thank you TSan!
When a document is opened twice without closing it we re-initialize its
internal data. Make sure we don't race with code that accesses that data.
Found by TSan!
Package up the logic that generates a full Clang module name, so that
(a) we don't have to deal with clang::Module in quite as many places
in the /Swift/ compiler, and (b) we can avoid the cost of a temporary
string in a few places.
The main places where this is /not/ adopted is where we don't just
want to know the parent module name, but actually the module itself.
This is mostly indexing-related queries, which use the very similar
ModuleEntity class also defined in Module.h. I didn't quite see an
obvious way to unify these, but that might be where we want to go.
No functionality change.
If an imported Objective-C method had no owning clang module, passCursorInfoForDecl() would call clang::Module::getFullModuleName() on a null instance. The current implementation of this method happens to terminate its loop immediately and return an empty string, but it’s best not to depend on this. Fixes a UBSan error in SourceKit/Mixed/cursor_mixed_header.swift.
The right way is findBufferContainingLoc. getBufferIdentifierForLoc is
both slower and wrong in the presence of #sourceLocation.
I couldn't come up with a test for the change in IDE/Utils.cpp because
refactoring still seems to be broken around #sourceLocation. I'll file
bugs for that.
But not even that much uglier; at all three call sites this will save
an allocation, and for the most important one (SourceKit) we can now
avoid creating a temporary CompilerInvocation just to copy into a
longer-lived one.
With this change, Driver no longer depends on Frontend, which means...
well, slightly faster builds of the compiler itself, but not much
else.
Most of this patch is just removing special cases for materializeForSet
or other fairly mechanical replacements. Unfortunately, the rest is
still a fairly big change, and not one that can be easily split apart
because of the quite reasonable reliance on metaprogramming throughout
the compiler. And, of course, there are a bunch of test updates that
have to be sync'ed with the actual change to code-generation.
This is SR-7134.
- getAsDeclOrDeclExtensionContext -> getAsDecl
This is basically the same as a dyn_cast, so it should use a 'getAs'
name like TypeBase does.
- getAsNominalTypeOrNominalTypeExtensionContext -> getSelfNominalTypeDecl
- getAsClassOrClassExtensionContext -> getSelfClassDecl
- getAsEnumOrEnumExtensionContext -> getSelfEnumDecl
- getAsStructOrStructExtensionContext -> getSelfStructDecl
- getAsProtocolOrProtocolExtensionContext -> getSelfProtocolDecl
- getAsTypeOrTypeExtensionContext -> getSelfTypeDecl (private)
These do /not/ return some form of 'this'; instead, they get the
extended types when 'this' is an extension. They started off life with
'is' names, which makes sense, but changed to this at some point. The
names I went with match up with getSelfInterfaceType and
getSelfTypeInContext, even though strictly speaking they're closer to
what getDeclaredInterfaceType does. But it didn't seem right to claim
that an extension "declares" the ClassDecl here.
- getAsProtocolExtensionContext -> getExtendedProtocolDecl
Like the above, this didn't return the ExtensionDecl; it returned its
extended type.
This entire commit is a mechanical change: find-and-replace, followed
by manual reformatted but no code changes.
That is, don't look through InOutType anymore, and update callers to
call getInOutObjectType() as well (or not, where it was obvious to me
that InOutType could not appear).
This surfaces more remaining uses of getInOutObjectType() directly.
Having this be a single buffer hardcoded in the SourceManager and set
by all clients is silly. SourceFiles with the 'Main' kind are allowed
to have hashbang lines (`#!`), other files are not. And anyone
manually setting up a Lexer can decide for themselves.
No intended behavioral change.
The information about whether a variable/property is initialized is lost in the
public interface, but is, unfortunately, required because it results in a symbol
for the initializer (if a class/struct `init` is inlined, it will call
initializers for properties that it doesn't initialize itself). This is
important to preserve for TBD file generation.
Using an attribute rather than just a bit on the VarDecl means this fits into
the scheme for module interfaces: textual/valid Swift.
Introduce ExtensionDecl::getExtendedNominal() to provide the nominal
type declaration that the extension declaration extends. Move most
of the existing callers of the callers to getExtendedType() over to
getExtendedNominal(), because they don’t need the full type information.
ExtensionDecl::getExtendedNominal() is itself not very interesting yet,
because it depends on getExtendedType().
We were always returning true from those functions in SKEditorConsumer
and false in the test consumers. On the client side we would then ignore
the return value. So it's clearer to have the functions not return
anything.
Otherwise, hits assertion, or crashes in no-assertion build.
Added 'EditableTextBuffer::getSize()' for getting size after previous updates
without actually applying them.
rdar://problem/34206143
