David Majnemer [Fri, 13 Mar 2015 22:36:55 +0000 (22:36 +0000)]
MS ABI: Generate default constructor closures
The MS ABI utilizes a compiler generated function called the "vector
constructor iterator" to construct arrays of objects with
non-trivial constructors/destructors. For this to work, the constructor
must follow a specific calling convention. A thunk must be created if
the default constructor has default arguments, is variadic or is
otherwise incompatible. This thunk is called the default constructor
closure.
N.B. Default constructor closures are only generated if the default
constructor is exported because clang itself does not utilize vector
constructor iterators. Failing to export the default constructor
closure will result in link/load failure if a translation unit compiled
with MSVC is on the import side.
Benjamin Kramer [Fri, 13 Mar 2015 16:10:42 +0000 (16:10 +0000)]
Sema: Replace the SetVector/DenseMap/std::sort combination with a simple std::map
This guarantees the order and doesn't increase malloc counts a lot as there are
typically very few elements int the map. Provide a little iterator adapter to
keep the same interface as we had with the flat sorted list.
Chandler Carruth [Fri, 13 Mar 2015 08:29:54 +0000 (08:29 +0000)]
[Modules] Teach Clang to survive ambiguous macros which come from system
headers even if they arrived when merging non-system modules.
The idea of this code is that we don't want to warn the user about
macros defined multiple times by their system headers with slightly
different definitions. We should have this behavior if either the
macro comes from a system module, or the definition within the module
comes from a system header. Previously, we would warn on ambiguous
macros being merged when they came from a users modules even though they
only showed up via system headers.
By surviving this we can handle common system header macro differences
like differing 'const' qualification of pointers due to some headers
predating 'const' being valid in C code, even when those systems headers
are pre-built into a system module.
Benjamin Kramer [Thu, 12 Mar 2015 23:41:40 +0000 (23:41 +0000)]
CodeGen: Base the conditional cleanup machinery on variadic templates
This is complicated by the fact that we can't simply use side-effecting
calls in an argument list without losing all guarantees about the order
they're emitted. To keep things deterministic we use tuples and brace
initialization, which thankfully guarantees evaluation order.
David Majnemer [Thu, 12 Mar 2015 17:44:49 +0000 (17:44 +0000)]
MS ABI: Allow a nullptr_t exception to be caught by void * catch handler
A nullptr exception object can be caught by any pointer type catch
handler. However, it is not possible to express this in the exception
info for the MS ABI. As a middle ground, allow such exception objects
to be caught with pointer-to-void catch handlers.
Logan Chien [Thu, 12 Mar 2015 17:27:53 +0000 (17:27 +0000)]
[docs] Update the doxygen configuration file.
Update the doxygen configuration file and the Makefile build rules
to provide better output (simply use the default stylesheet and template
from the Doxygen distribution.)
This CL has upgrade doxygen.cfg.in to Doxygen 1.8.6.
Benjamin Kramer [Thu, 12 Mar 2015 16:25:19 +0000 (16:25 +0000)]
ASTMatchers: Add an explicit dependency on libclangBasic.
In a static build the dependency is picked up implictly, but not in a shared
library build. This is needed for the new ObjC matchers that reference Selector.
Manuel Klimek [Thu, 12 Mar 2015 15:48:15 +0000 (15:48 +0000)]
Add support for a few Objective-C matchers.
Add some matchers for Objective-C selectors and messages to
ASTMatchers.h. Minor mods to ASTMatchersTest.h to allow test files with
".m" extension in addition to ".cpp". New tests added to
ASTMatchersTest.c.
Benjamin Kramer [Thu, 12 Mar 2015 14:28:47 +0000 (14:28 +0000)]
Sema: Don't emit a missing prototype warning for deleted functions.
This is a bit more involved than I anticipated, so here's a breakdown
of the changes:
1. Call ActOnFinishFunctionBody _after_ we parsed =default and
=delete specifiers. Saying that we finished the body before parsing
=default is just wrong. Changing this allows us to use isDefaulted
and isDeleted on a decl in ActOnFinishFunctionBody.
2. Check for -Wmissing-prototypes after we parsed the function body.
3. Disable -Wmissing-prototypes when the Decl isDeleted.
Benjamin Kramer [Thu, 12 Mar 2015 14:28:38 +0000 (14:28 +0000)]
Use Sema's PrintingPolicy when diagnosing DeclSpecs.
Sema overrides ASTContext's policy on the first emitted diagnostic
(doesn't matter if it's ignored or not). This means changing the order
of diagnostic emission in Sema suddenly changes the text of diagnostic
emitted from the parser.
In the test case -Wmissing-prototypes (ignored) was the culprit, use
'int main' to suppress that warning so we see when this regresses.
Also move it into Sema/ as it's not testing any C++.
Aaron Ballman [Thu, 12 Mar 2015 14:14:48 +0000 (14:14 +0000)]
Reverting r232034, as it broke one of the bots with link errors. Details at: http://bb.pgr.jp/builders/ninja-clang-x64-mingw64-RA/builds/6352/steps/build/logs/stdio
Aaron Ballman [Thu, 12 Mar 2015 13:49:45 +0000 (13:49 +0000)]
Instead of dereferencing std::vector::end() (which is UB and causes failed assertions in debug builds with Visual Studio), use data() + size() to calculate the end iterator. Amends r231952.
Aaron Ballman [Thu, 12 Mar 2015 13:21:19 +0000 (13:21 +0000)]
Added some matchers for objective c selectors and messages to ASTMatchers.h. Minor mods to ASTMatchersTest.h to allow test files with ".m" extension in addition to ".cpp". New tests added to ASTMatchersTest.c.
Patch by Dean Sutherland, reviewed by Manuel Klimek. From http://reviews.llvm.org/D7710
Alexey Bataev [Thu, 12 Mar 2015 08:53:29 +0000 (08:53 +0000)]
[OPENMP] Initial codegen for 'omp sections' and 'omp section' directives.
If only one section is found in the sections region, it is emitted just like single region.
Otherwise it is emitted as a static non-chunked loop.
#pragma omp sections
{
#pragma omp section
{1}
...
#pragma omp section
{n}
}
is translated to something like
Justin Bogner [Thu, 12 Mar 2015 00:52:56 +0000 (00:52 +0000)]
Driver: Keep -isysroot flags in crash scripts if we're dumping a VFS
For crashes with a VFS (ie, with modules), the -isysroot flag is often
necessary to reproduce the crash. This is especially true if some
modules need to be rebuilt, since without the sysroot they'll try to
read headers that are outside of the VFS.
I find it likely that we should keep some of the other -i flags in
this case as well, but I haven't seen that come up in practice yet so
it seems better to be conservative.
Justin Bogner [Thu, 12 Mar 2015 00:14:35 +0000 (00:14 +0000)]
Driver: Print the clang version and original command in crash scripts
When a crash report script doesn't work for a reproduction on your
machine for one reason or another, it can be really tricky to figure
out why not. The compiler version that crashed and the original
command line before stripping flags are very helpful when this comes
up.
Hal Finkel [Wed, 11 Mar 2015 19:14:15 +0000 (19:14 +0000)]
[PowerPC] ABI support for the QPX vector instruction set
Support for the QPX vector instruction set, used on the IBM BG/Q supercomputer,
has recently been added to the LLVM PowerPC backend. This vector instruction
set requires some ABI modifications because the ABI on the BG/Q expects
<4 x double> vectors to be provided with 32-byte stack alignment, and to be
handled as native vector types (similar to how Altivec vectors are handled on
mainline PPC systems). I've named this ABI variant elfv1-qpx, have made this
the default ABI when QPX is supported, and have updated the ABI handling code
to provide QPX vectors with the correct stack alignment and associated
register-assignment logic.
David Majnemer [Wed, 11 Mar 2015 18:36:39 +0000 (18:36 +0000)]
MS ABI: Implement copy-ctor closures, finish implementing throw
This adds support for copy-constructor closures. These are generated
when the C++ runtime has to call a copy-constructor with a particular
calling convention or with default arguments substituted in to the call.
Because the runtime has no mechanism to call the function with a
different calling convention or know-how to evaluate the default
arguments at run-time, we create a thunk which will do all the
appropriate work and package it in a way the runtime can use.
Richard Smith [Wed, 11 Mar 2015 18:21:02 +0000 (18:21 +0000)]
[modules] When merging the pattern of a class template definition into a prior
definition, be sure to update the definition data on all declarations, not just
the canonical one, since the pattern might not be in the list of pending
definitions (if it used to be canonical itself).
One-line fix by me; reduced testcase by Daniel Jasper!
Ed Schouten [Wed, 11 Mar 2015 08:46:01 +0000 (08:46 +0000)]
Fix up default header paths for CloudABI.
CloudABI is a pure cross compilation target. This means that we should
not add /usr/include and /usr/local/include. Instead, headers are stored
in $sysroot/$triple/include.
The method of going back to the sysroot (by using "../../..") is also
used in this function for some of the other environments (e.g., MinGW).
Ed Schouten [Wed, 11 Mar 2015 08:42:46 +0000 (08:42 +0000)]
Add target information for CloudABI on x86-64.
CloudABI can be identified by the __CloudABI__ preprocessor definition. The
system uses ELF executables.
CloudABI uses Unicode 7.0.0 for the encoding of wchar_t. As Unicode 7.0.0 is
synchronized with ISO/IEC 10646:2012 (released on 2012-06-01),
__STDC_ISO_10646__ is defined as 201206L.
David Majnemer [Wed, 11 Mar 2015 06:45:39 +0000 (06:45 +0000)]
Sema: Properly track mangling number/name for linkage for using decls
Using declarations which are aliases to struct types have their name
used as the struct type's name for linkage purposes. Otherwise, make
sure to give an anonymous struct defined inside a using declaration a
mangling number to disambiguate it from other anonymous structs in the
same context.
Alexey Bataev [Wed, 11 Mar 2015 04:48:56 +0000 (04:48 +0000)]
[OPENMP] Fix for ExprWithCleanups in 'omp atomic' constructs.
This patch allows using of ExprWithCleanups expressions and other complex expressions in 'omp atomic' construct
Differential Revision: http://reviews.llvm.org/D8200
Richard Smith [Wed, 11 Mar 2015 01:44:51 +0000 (01:44 +0000)]
[modules] Avoid accidentally completing the redeclaration chain when updating
all the existing declarations of a record-like entity with a pointer to its
definition.
Aaron Ballman [Tue, 10 Mar 2015 17:19:18 +0000 (17:19 +0000)]
The semantic spelling enumeration should retain values to the spelling list indexes used by the attribute. The only attribute affected by this in practice is the OpenCLImageAccessAttr, which has duplicate semantic spellings that are automatically stripped.
We do not implicitly create an OpenCLImageAccessAttr, so this change only affects out of tree users. There is no way to test this behavior specifically that I can see, since this only affects implicit creation of attributes.
Sanjay Patel [Tue, 10 Mar 2015 15:19:26 +0000 (15:19 +0000)]
[X86, AVX] Replace vinsertf128 intrinsics with generic shuffles.
We want to replace as much custom x86 shuffling via intrinsics
as possible because pushing the code down the generic shuffle
optimization path allows for better codegen and less complexity
in LLVM.
This is the sibling patch for the LLVM half of this change:
http://reviews.llvm.org/D8086
Yaron Keren [Tue, 10 Mar 2015 07:33:23 +0000 (07:33 +0000)]
Teach raw_ostream to accept SmallString.
Saves adding .str() call to any raw_ostream << SmallString usage
and a small step towards making .str() consistent in the ADTs by
removing one of the SmallString::str() use cases, discussion at
Alexey Bataev [Tue, 10 Mar 2015 07:28:44 +0000 (07:28 +0000)]
[OPENMP] Initial codegen for 'omp task' directive.
The task region is emmitted in several steps:
Emit a call to kmp_task_t *__kmpc_omp_task_alloc(ident_t *, kmp_int32 gtid, kmp_int32 flags, size_t sizeof_kmp_task_t, size_t sizeof_shareds, kmp_routine_entry_t *task_entry).
Here task_entry is a pointer to the function:
kmp_int32 .omp_task_entry.(kmp_int32 gtid, kmp_task_t *tt) {
TaskFunction(gtid, tt->part_id, tt->shareds);
return 0;
}
Copy a list of shared variables to field shareds of the resulting structure kmp_task_t returned by the previous call (if any).
Copy a pointer to destructions function to field destructions of the resulting structure kmp_task_t.
Emit a call to kmp_int32 __kmpc_omp_task(ident_t *, kmp_int32 gtid, kmp_task_t *new_task), where new_task is a resulting structure from previous items.
Differential Revision: http://reviews.llvm.org/D7560
Alexey Bataev [Tue, 10 Mar 2015 05:15:26 +0000 (05:15 +0000)]
[OPENMP] Improved code for generating debug info + generation of all OpenMP regions in termination scope
Patch adds proper generation of debug info for all OpenMP regions. Also, all OpenMP regions are generated in a termination scope, because standard does not allow to throw exceptions out of structured blocks, associated with the OpenMP regions
Differential Revision: http://reviews.llvm.org/D7935
Rafael Espindola [Tue, 10 Mar 2015 04:40:21 +0000 (04:40 +0000)]
Revert "[OPENMP] Improved code for generating debug info + generation of all OpenMP regions in termination scope Patch adds proper generation of debug info for all OpenMP regions. Also, all OpenMP regions are generated in a termination scope, because standard does not allow to throw exceptions out of structured blocks, associated with the OpenMP regions Differential Revision: http://reviews.llvm.org/D7935"
This reverts commit r231752.
It was failing to link with cmake:
lib64/libclangCodeGen.a(CGOpenMPRuntime.cpp.o):/home/espindola/llvm/llvm/tools/clang/lib/CodeGen/CGOpenMPRuntime.cpp:function clang::CodeGen::InlinedOpenMPRegionRAII::~InlinedOpenMPRegionRAII(): error: undefined reference to 'clang::CodeGen::EHScopeStack::popTerminate()'
clang-3.7: error: linker command failed with exit code 1 (use -v to see invocation)
Alexey Bataev [Tue, 10 Mar 2015 04:22:11 +0000 (04:22 +0000)]
[OPENMP] Improved code for generating debug info + generation of all OpenMP regions in termination scope
Patch adds proper generation of debug info for all OpenMP regions. Also, all OpenMP regions are generated in a termination scope, because standard does not allow to throw exceptions out of structured blocks, associated with the OpenMP regions
Differential Revision: http://reviews.llvm.org/D7935
Richard Smith [Tue, 10 Mar 2015 02:00:53 +0000 (02:00 +0000)]
PR21687: when adding a redeclaration of a function with an implicit exception
specification, update all prior declarations if the new one has an explicit
exception specification and the prior ones don't.
Patch by Vassil Vassilev! Some minor tweaking and test case by me.
Richard Smith [Tue, 10 Mar 2015 01:41:22 +0000 (01:41 +0000)]
[modules] Don't clobber a destructor's operator delete when adding another one;
move the operator delete updating into a separate update record so we can cope
with updating another module's destructor's operator delete.
Richard Smith [Tue, 10 Mar 2015 00:19:04 +0000 (00:19 +0000)]
[modules] This check is run before we resolve the header, not after, so just
check that private headers are in a list matching the role. (We can't perform
the opposite checks for non-private headers because we infer those.)
Alexey Samsonov [Mon, 9 Mar 2015 21:50:19 +0000 (21:50 +0000)]
[UBSan] Split -fsanitize=shift into -fsanitize=shift-base and -fsanitize=shift-exponent.
This is a recommit of r231150, reverted in r231409. Turns out
that -fsanitize=shift-base check implementation only works if the
shift exponent is valid, otherwise it contains undefined behavior
itself.
Make sure we check that exponent is valid before we proceed to
check the base. Make sure that we actually report invalid values
of base or exponent if -fsanitize=shift-base or
-fsanitize=shift-exponent is specified, respectively.
Tim Northover [Mon, 9 Mar 2015 21:40:42 +0000 (21:40 +0000)]
ARM: use ABI-specified alignment for byval parameters.
When passing a type with large alignment byval, we were specifying the type's
alignment rather than the alignment that the backend is actually capable of
producing (ABIAlign).
This would be OK (if odd) assuming the backend dealt with it prooperly,
unfortunately it doesn't and trying to pass types with "byval align 16" can
cause it to set fp incorrectly and trash the stack during the prologue. I'll be
fixing that in a separate patch, but Clang should still be emitting IR that's
as close to its intent as possible.
[PATCH Sema Objective-C]. Patch to warn on missing designated initializer
override where at least a declaration of a designated initializer is in a super
class and not necessarily in the current class. rdar://19653785.