Douglas Gregor [Wed, 16 Feb 2011 19:08:06 +0000 (19:08 +0000)]
Teach the CXCodeCompleteResults results structure, which stores
code-completion results accessed via libclang, to extend the lifetime
of the allocator used for cached global code-completion results at
least until these completion results are destroyed. Fixes
<rdar://problem/8997369>.
Douglas Gregor [Wed, 16 Feb 2011 18:16:54 +0000 (18:16 +0000)]
Improve the invalidation logic for the cache of global code
completions. We now compute a hash of the names of all top-level
declarations and macro definitions, and invalidate the cache when the
hash value changes.
Zhanyong Wan [Wed, 16 Feb 2011 05:19:17 +0000 (05:19 +0000)]
Improves CMakeLists.txt for Clang's unit tests: make "linked
components" and "used libs" arguments of add_clang_unittest().
Reviewed by jyasskin and chapuni.
Douglas Gregor [Wed, 16 Feb 2011 01:39:26 +0000 (01:39 +0000)]
When searching for visible declarations (e.g., for code completion),
be sure to look at all of the Objective-C class declarations within a
@class. Fixes <rdar://problem/8876207>.
Douglas Gregor [Wed, 16 Feb 2011 00:51:18 +0000 (00:51 +0000)]
When trying to provide a code completion item for a call to "super" in
Objective-C, also look in the categories and class extensions of our
superclasses. Fixes <rdar://problem/8853540>.
John McCall [Wed, 16 Feb 2011 00:49:34 +0000 (00:49 +0000)]
Don't call objc_read_weak as part of emitting a block literal.
Nobody ever gave me a clear reason for why we were doing this, and
now it's apparently causing serious problems, so if *not* having this
causes problems, we get to solve them the right way this time.
Devang Patel [Tue, 15 Feb 2011 23:36:28 +0000 (23:36 +0000)]
Only c++ class arguments with non trivial constructor or destructor needs a reference.
C struct arguments do not need this adjustment.
This fixes 7 failures in callfuncs.exp from gdb testsuite.
Rafael Espindola [Tue, 15 Feb 2011 22:23:51 +0000 (22:23 +0000)]
Add a hack to avoid adding '\01' to asm names when possible. It would be
better for clang to always compute the right name, but for now this hack
fixes PR9177 and lets us build firefox with LTO :-)
Douglas Gregor [Tue, 15 Feb 2011 20:33:25 +0000 (20:33 +0000)]
When code-completing within a list of declaration specifiers,
separately handle the case of a local declaration-specifier list,
including all types in the set of options. Fixes
<rdar://problem/8790735> and <rdar://problem/8662831>.
Douglas Gregor [Tue, 15 Feb 2011 19:17:31 +0000 (19:17 +0000)]
When we encounter an Objective-C class name in an expression, followed
by the code completion token, treat this as a class message send where
the opening square bracket is missing. Fixes <rdar://problem/6970911>.
Allow resolving headers from a PCH even after headers+PCH were moved to another path.
Store in PCH the directory that the PCH was originally created in.
If a header file is not found at the path that we expect it to be and the PCH file
was moved from its original location, try to resolve the file by assuming that
header+PCH were moved together and the header is in the same place relative to the PCH.
John McCall [Tue, 15 Feb 2011 09:22:45 +0000 (09:22 +0000)]
Assorted cleanup:
- Have CGM precompute a number of commonly-used types
- Have CGF copy that during initialization instead of recomputing them
- Use TBAA info when initializing a parameter variable
- Refactor the scalar ++/-- code
[analyzer] Reflect changes for tablegen'ing the checkers.
-Update tablegen files for checkers, use the tablegen class name for the checker class name.
-Update ClangSACheckersProvider to not look into hidden checker packages.
John McCall [Tue, 15 Feb 2011 07:12:36 +0000 (07:12 +0000)]
Return a declaration to the parser when creating a field in C++ so that
the parser will complete the declarator with a valid decl and thus trigger
delayed diagnostics for it. It certainly looks like we were intentionally
returning null here, but I couldn't find any good reason for it, and there
wasn't a comment, so farewell to all that.
John McCall [Tue, 15 Feb 2011 06:40:56 +0000 (06:40 +0000)]
Perform zero-initialization of virtual base classes when emitting
a zero constant for a complete class. rdar://problem/8424975
To make this happen, track the field indexes for virtual bases
in the complete object. I'm curious whether we might be better
off making CGRecordLayoutBuilder *much* more reliant on
ASTRecordLayout; we're currently duplicating an awful lot of the ABI
layout logic.
Warn if method for a deprecated method is implemented.
Warn if class for a deprecated class is implemented.
Warn if category for a deprecated class is implemented.
All under control of -Wdeprecated-implementations.
// rdar://8973810.
[analyzer] Overhauling of the checker registration mechanism.
-Checkers will be defined in the tablegen file 'Checkers.td'.
-Apart from checkers, we can define checker "packages" that will contain a collection of checkers.
-Checkers can be enabled with -analyzer-checker=<name> and disabled with -analyzer-disable-checker=<name> e.g:
Enable checkers from 'cocoa' and 'corefoundation' packages except the self-initialization checker:
-analyzer-checker=cocoa -analyzer-checker=corefoundation -analyzer-disable-checker=cocoa.SelfInit
-Introduces CheckerManager and CheckerProvider. CheckerProviders get the set of checker names to enable/disable and
register them with the CheckerManager which will be the entry point for all checker-related functionality.
Currently only the self-initialization checker takes advantage of the new mechanism.
[analyzer] Remove ManagerRegistry which is not used. In the future we may load analyzer plugins dynamically but
registration through static constructors should be avoided.
[analyzer] Move include/clang/StaticAnalyzer/AnalysisConsumer.h -> lib/StaticAnalyzer/Frontend/AnalysisConsumer.h since
FrontendActions.cpp is the only user.
John McCall [Mon, 14 Feb 2011 07:13:47 +0000 (07:13 +0000)]
When parsing an out-of-line member function declaration, we must delay
access-control diagnostics which arise from the portion of the declarator
following the scope specifier, just in case access is granted by
friending the individual method. This can also happen with in-line
member function declarations of class templates due to templated-scope
friend declarations.
We were really playing fast-and-loose before with this sort of thing,
and it turned out to work because *most* friend functions are in file
scope. Making us delay regardless of context exposed several bugs with
how we were manipulating delay. I ended up needing a concept of a
context that's independent of the declarations in which it appears,
and then I actually had to make some things save contexts correctly,
but delay should be much cleaner now.
I also encapsulated all the delayed-diagnostics machinery in a single
subobject of Sema; this is a pattern we might want to consider rolling
out to other components of Sema.
John McCall [Sun, 13 Feb 2011 04:07:26 +0000 (04:07 +0000)]
Give some convenient idiomatic accessors to Stmt::child_range and
Stmt::const_child_range, then make a bunch of places use them instead
of the individual iterator accessors.
Implement objective-c++'s block pointer type matching involving
types which are contravariance in argument types and covariance
in return types. // rdar://8979379.
When reading the AST, delay loading of the redeclaration chain to avoid deeply nested calls.
Temporarily set the first (canonical) declaration as the previous one, which is the one that
matters, and mark the real previous DeclID to be loaded & attached later on.
projects / clang / log