Adrian Prantl [Thu, 18 Jul 2013 00:28:02 +0000 (00:28 +0000)]
Replace llvm::DIBuilder::DisableDebugLocations() with two RAII interfaces
inspired by CodegenFunction::LexicalScope.
- NoLocation temporarily turns off debug locations altogether.
This is useful for emitting instructions that should be
counted towards the function prologue.
- BuiltinLocation temporarily switches to an artificial debug location
that has a valid scope, but no line information. This is useful when
emitting compiler-generated helper functions that have no source
location associated with them.
Richard Smith [Wed, 17 Jul 2013 23:53:16 +0000 (23:53 +0000)]
Reinstate r186040, with additional fixes and more test coverage (reverted in
r186331).
Original commit log:
If we friend a declaration twice, that should not make it visible to
name lookup in the surrounding context. Slightly rework how we handle
friend declarations to inherit the visibility of the prior
declaration, rather than setting a friend declaration to be visible
whenever there was a prior declaration.
Jordan Rose [Wed, 17 Jul 2013 17:16:42 +0000 (17:16 +0000)]
[analyzer] Handle C++11 member initializer expressions.
Previously, we would simply abort the path when we saw a default member
initialization; now, we actually attempt to evaluate it. Like default
arguments, the contents of these expressions are not actually part of the
current function, so we fall back to constant evaluation.
Jordan Rose [Wed, 17 Jul 2013 17:16:38 +0000 (17:16 +0000)]
[analyzer] Handle C string default values for const char * arguments.
Previously, SValBuilder knew how to evaluate StringLiterals, but couldn't
handle an array-to-pointer decay for constant values. Additionally,
RegionStore was being too strict about loading from an array, refusing to
return a 'char' value from a 'const char' array. Both of these have been
fixed.
Jordan Rose [Wed, 17 Jul 2013 17:16:33 +0000 (17:16 +0000)]
[analyzer] Treat std::initializer_list as opaque rather than aborting.
Previously, the use of a std::initializer_list (actually, a
CXXStdInitializerListExpr) would cause the analyzer to give up on the rest
of the path. Now, it just uses an opaque symbolic value for the
initializer_list and continues on.
At some point in the future we can add proper support for initializer_list,
with access to the elements in the InitListExpr.
Pavel Labath [Wed, 17 Jul 2013 08:33:58 +0000 (08:33 +0000)]
[analyzer] Add very limited support for temporary destructors
Summary:
This patch enables ExprEndgine to reason about temporary object destructors.
However, these destructor calls are never inlined, since this feature is still
broken. Still, this is sufficient to properly handle noreturn temporary
destructors and close bug #15599. I have also enabled the cfg-temporary-dtors
analyzer option by default.
_Complex load/store didn't have their alignment set properly, which was visible when GCC's torture tests use volatile _Complex.
Update some existing tests to check for alignment, and add a new test which also has over-aligned volatile _Complex (since the imaginary part shouldn't be overaligned, only the real part).
Eli Friedman [Tue, 16 Jul 2013 22:40:53 +0000 (22:40 +0000)]
Make Expr::isConstantInitializer match IRGen.
Sema needs to be able to accurately determine what will be
emitted as a constant initializer and what will not, so
we get accurate errors in C and accurate -Wglobal-constructors
warnings in C++. This makes Expr::isConstantInitializer match
CGExprConstant as closely as possible.
Summary:
These can appear when comments contain command lines with quoted line
breaks. As the text (including escaped newlines and '//' from consecutive lines)
is a single line comment, we used to break it even when it didn't exceed column
limit. This is a temporary solution, in the future we may want to support this
case completely - at least adjust leading whitespace when changing indentation
of the first line.
Daniel Jasper [Tue, 16 Jul 2013 18:22:10 +0000 (18:22 +0000)]
Revamp the formatting of C++11 braced init lists.
The fundamental concept is:
Format as if the braced init list was a function call (with parentheses
replaced by braces). If there is no name/type before the opening brace
(e.g. if the braced list is nested), assume a zero-length identifier
just before the opening brace.
This behavior is gated on a new style flag, which for now replaces the
SpacesInBracedLists style flag. Activate this style flag for Google
style to reflect recent style guide changes.
Samuel Benzaquen [Tue, 16 Jul 2013 15:47:24 +0000 (15:47 +0000)]
Add more types to ASTNodeKind. Refactor common instantiation code.
Summary:
Add support for CXXCtorInitializer and TemplateArgument types to ASTNodeKind.
This change is to support more matchers from clang/ASTMatchers/ASTMatchers.h in the dynamic layer (clang/ASTMatchers/Dynamic).
This patch removes unused parameter allProperties and converts remaining
parameters in ArrayRef'ize Sema::ActOnAtEnd to ArrayRef.
Patch by Robert Wilhelm.
Manuel Klimek [Tue, 16 Jul 2013 13:20:30 +0000 (13:20 +0000)]
Fixes another hard to test problem with iterator invalidation.
As every match call can recursively call back into the memoized match
via a nested traversal matcher (for example:
stmt(hasAncestor(stmt(hasDescendant(stmt(hasDescendant(stmt()))))))),
and every memoization step might clear the cache, we must not store
iterators into the result cache when calling match on a submatcher.
Daniel Jasper [Tue, 16 Jul 2013 11:37:21 +0000 (11:37 +0000)]
clang-format: Improve detection of function types.
This fixes an incorrect detection that led to a formatting error.
Before:
some_var = function (*some_pointer_var)[0];
After:
some_var = function(*some_pointer_var)[0];
Tim Northover [Tue, 16 Jul 2013 10:22:19 +0000 (10:22 +0000)]
Fix test on release builds.
Unfortunately I don't think there's a good way to validate branch targets on
release builds. Fortunately it's a minor part of this test (and based on
generic code) so I don't mind dropping it.
Tim Northover [Tue, 16 Jul 2013 09:47:53 +0000 (09:47 +0000)]
ARM: implement low-level intrinsics for the atomic exclusive operations.
This adds three overloaded intrinsics to Clang:
T __builtin_arm_ldrex(const volatile T *addr)
int __builtin_arm_strex(T val, volatile T *addr)
void __builtin_arm_clrex()
The intent is that these do what users would expect when given most sensible
types. Currently, "sensible" translates to ints, floats and pointers.
Eli Friedman [Tue, 16 Jul 2013 00:01:31 +0000 (00:01 +0000)]
Fix member refs with using decl + anonymous union.
Make sure we call BuildFieldReferenceExpr with the appropriate decl
when a member of an anonymous union is made public with a using decl.
Also, fix a crash on invalid field access into an anonymous union.
Move the "->" to "." fixit from r186128 into a separate note since
recovery is not attempted with the fixit. Also move the associated test
case from FixIt/fixit.cpp to SemaCXX/member-expr.cpp since the fixit is
no longer automatically applied.
Samuel Benzaquen [Mon, 15 Jul 2013 19:25:06 +0000 (19:25 +0000)]
Add support for type traversal matchers.
Summary:
Fixup the type traversal macros/matchers to specify the supported types.
Make the marshallers a little more generic to support any variadic function.
Update the doc script.
Re-revert r86040, which was un-reverted in r186199.
This breaks the build of basic patterns with repeated friend
declarations. See the added test case in SemaCXX/friend.cpp or the test
case reported to the original commit log.
Original commit log:
If we friend a declaration twice, that should not make it visible to
name lookup in the surrounding context. Slightly rework how we handle
friend declarations to inherit the visibility of the prior
declaration, rather than setting a friend declaration to be visible
whenever there was a prior declaration.
Fix to PR12262 - assertion when substituting explicit template arguments
does not substitute a sizeof-pack expression.
The solution is proposed by Richard Smith.
Differential Revision: http://llvm-reviews.chandlerc.com/D869
Richard Smith [Sun, 14 Jul 2013 05:41:45 +0000 (05:41 +0000)]
Add missing include guards into headers in lib/Headers. While it may appear
that these headers should not be included more than once, they are in fact
included twice when building our builtins module (in order for it to generate
submodules for them), and without this, any modular build enabling AVX and
including any builtin header fails.
Testing this is tricky because including any of these headers in a modular
build is liable to fail, due to unrelated builtin headers in the same module
including headers which might not be available on the system running the tests.
Suggestion on that front are welcome (but we're getting close to being able to
run a buildbot that has modules enabled for all tests, which would nicely solve
the testing problem).
Richard Smith [Sun, 14 Jul 2013 02:01:48 +0000 (02:01 +0000)]
If an unimported submodule of an imported module contains a declaration of a
global allocation or deallocation function, that should not cause that global
allocation or deallocation function to become unavailable.
David Blaikie [Sat, 13 Jul 2013 21:08:14 +0000 (21:08 +0000)]
PR16214, PR14467: DebugInfo: use "RequireCompleteType" to decide when to emit the full definition of a type in -flimit-debug-info
This simplifies the core benefit of -flimit-debug-info by taking a more
systematic approach to avoid emitting debug info definitions for types
that only require declarations. The previous ad-hoc approach (3 cases
removed in this patch) had many holes.
The general approach (adding a bit to TagDecl and callback through
ASTConsumer) has been discussed with Richard Smith - though always open
to revision.
David Blaikie [Sat, 13 Jul 2013 21:08:08 +0000 (21:08 +0000)]
Correctly classify pack expansions as NON_CANONICAL_UNLESS_DEPENDENT
Test coverage for non-dependent pack expansions doesn't demonstrate a
failure prior to this patch (a follow-up commit improving debug info
will cover this commit specifically) but covers a related hole in our
test coverage.
Fix some uninstantiable code in ASTVector::insert. I've added a
cheap-and-dirty compile test for this, because I don't have the time to
figure out a nice way to get a real ASTContext to implement executable
tests - but we probably should have them for this ADT.
Richard Smith [Sat, 13 Jul 2013 02:00:19 +0000 (02:00 +0000)]
C++ modules: Don't call DeclContext::lookup when half-way through deserializing
decls. That can reenter deserialization and explode horribly by trying to merge
a declaration that we've not got very far through deserializing yet.
[ms-cxxabi] Don't consider function templates for name backrefs
They don't seem to be used for back references, presumably because a
function template is unlikely to reoccur, while a class template name
may reoccur as a type.
This fixes a mangling issue for llvm::hash_combine() in Hashing.h.
Richard Smith [Fri, 12 Jul 2013 20:38:49 +0000 (20:38 +0000)]
Unrevert r186040, reverted in r186185, with fix for PR16597.
Original commit log:
If we friend a declaration twice, that should not make it visible to
name lookup in the surrounding context. Slightly rework how we handle
friend declarations to inherit the visibility of the prior
declaration, rather than setting a friend declaration to be visible
whenever there was a prior declaration.
Revert r186040 to fix PR16597 while Richard investigates what the best
fix is.
Original commit log:
If we friend a declaration twice, that should not make it visible to
name lookup in the surrounding context. Slightly rework how we handle
friend declarations to inherit the visibility of the prior
declaration, rather than setting a friend declaration to be visible
whenever there was a prior declaration.