Alexander Musman [Mon, 21 Sep 2015 14:41:00 +0000 (14:41 +0000)]
Fix assertion in inline assembler IR gen
Several inputs may not refer to one output constraint in inline assembler
insertions, clang was failing on assertion on such test case.
[ARM] Handle +t2dsp feature as an ArchExtKind in ARMTargetParser.def
Currently, the availability of DSP instructions (ACLE 6.4.7) is handled in
a hand-rolled tricky condition block in lib/Basic/Targets.cpp, with a FIXME:
attached.
http://reviews.llvm.org/D12937 moved the handling of +t2dsp over to
ARMTargetParser.def in LLVM, to be in line with other architecture extensions.
This is the corresponding patch to clang, to clear the FIXME: and update
the tests.
This changes getARMFloatABI to use the ToolChain and Args instead of Driver,
Args, Triple. Although this pushes the Triple calculation/parsing into the
function itself, it enables the use of the function for a future change. The
reason to sink the triple calculation here is to avoid threading the Triple
through multiple layers in a future change.
Rather than using re-calculating the effective triple, thread the already
calculated value down into AddARMTargetArgs. This avoids both recreating the
triple, as well as re-parsing the triple as it was already done in the previous
frame.
Simon Pilgrim [Sat, 19 Sep 2015 15:12:38 +0000 (15:12 +0000)]
[X86][SSE] Replace 128-bit SSE41 PMOVSX intrinsics with native IR
128-bit vector integer sign extensions correctly lower to the pmovsx instructions even for debug builds.
This patch removes the builtins and reimplements the _mm_cvtepi*_epi* intrinsics __using builtin_shufflevector (to extract the bottom most subvector) and __builtin_convertvector (to actually perform the sign extension).
[Modules] More descriptive diagnostics for misplaced import directive
If an import directive was put into wrong context, the error message was obscure,
complaining on misbalanced braces. To get more descriptive messages, annotation
tokens related to modules are processed where they must not be seen.
Use an enumeration for the Floating Point ABIs supported on MIPS. This is
replicating the ARM change to avoid string based tracking of the floating point
ABI. NFC.
Richard Trieu [Fri, 18 Sep 2015 23:18:39 +0000 (23:18 +0000)]
Split off the binary literal warning into a subgroup of C++14 warnings
Binary literals predate C++14, but they are listed as a C++14 extension since
this was the first time they were standardized in the language. Move the
warning into a subgroup so it can be selectively disabled when checking for
other C++14 features.
Adrian Prantl [Fri, 18 Sep 2015 23:01:45 +0000 (23:01 +0000)]
CGDebugInfo: Make creating a skeleton CU in getOrCreateModuleRef optional.
We don't want a skeleton CU when generating debug info for the module
itself.
Summary:
Name `Out` refers to the parameter. It is moved into the member `Out`
in ctor-init. Dereferencing null pointer will crash clang, if user
passes '-analyzer-viz-egraph-ubigraph' argument.
Chris Bieneman [Fri, 18 Sep 2015 21:15:54 +0000 (21:15 +0000)]
[CMake] Refactoring and cleaning up clang symlink generation.
Generation of clang symlinks now uses add_clang_symlink macro which uses add_llvm_symlink. Also the list of symlinks to generate is configurable via CLANG_LINKS_TO_CREATE.
This re-lands r248015, with fixes for clang symlinks to always be generated as part of the clang and install-clang targets.
Adrian Prantl [Fri, 18 Sep 2015 21:06:14 +0000 (21:06 +0000)]
Debug Info: Use the full module name as a key when caching DIModules.
The signature may not have been computed at the time the module reference
is generated (e.g.: in the future while emitting debug info for a clang
module). Using the full module name is safe because each clang module may
only have a single definition.
LLVM r217812 made it so that clang-cl implicitly creates bigobj files when
needed, independent of this flag. It looks like cl has this flag to produce obj
flags compatible with MSVS 2003's linker by default, something we don't care
about. Since clang-cl always has /bigobj behavior, don't warn that the flag is
unused, just ignore it silently.
Chris Bieneman [Fri, 18 Sep 2015 18:09:15 +0000 (18:09 +0000)]
[CMake] Refactoring and cleaning up clang symlink generation.
Generation of clang symlinks now uses add_clang_symlink macro which uses add_llvm_symlink. Also the list of symlinks to generate is configurable via CLANG_LINKS_TO_CREATE.
Use an enumeration and change the use of the FloatABI from a string to the
enumeration. This avoids the use of string values to represent an enumeration.
NFC.
Charles Davis [Thu, 17 Sep 2015 20:55:33 +0000 (20:55 +0000)]
Support __builtin_ms_va_list.
Summary:
This change adds support for `__builtin_ms_va_list`, a GCC extension for
variadic `ms_abi` functions. The existing `__builtin_va_list` support is
inadequate for this because `va_list` is defined differently in the Win64
ABI vs. the System V/AMD64 ABI.
Rename AST node matchers to match the AST node names directly. Part of this rename also splits recordDecl() (which used to match CXXRecordDecl) into recordDecl() (that matches RecordDecl) and cxxRecordDecl (that matches CXXRecordDecl). Also adds isStruct(), isUnion(), and isClass() narrowing matchers for RecordDecl objects.
[analyzer] Add generateErrorNode() APIs to CheckerContext.
The analyzer trims unnecessary nodes from the exploded graph before reporting
path diagnostics. However, in some cases it can trim all nodes (including the
error node), leading to an assertion failure (see
https://llvm.org/bugs/show_bug.cgi?id=24184).
This commit addresses the issue by adding two new APIs to CheckerContext to
explicitly create error nodes. Unless the client provides a custom tag, these
APIs tag the node with the checker's tag -- preventing it from being trimmed.
The generateErrorNode() method creates a sink error node, while
generateNonFatalErrorNode() creates an error node for a path that should
continue being explored.
The intent is that one of these two methods should be used whenever a checker
creates an error node.
This commit updates the checkers to use these APIs. These APIs
(unlike addTransition() and generateSink()) do not take an explicit Pred node.
This is because there are not any error nodes in the checkers that were created
with an explicit different than the default (the CheckerContext's Pred node).
It also changes generateSink() to require state and pred nodes (previously
these were optional) to reduce confusion.
Additionally, there were several cases where checkers did check whether a
generated node could be null; we now explicitly check for null in these places.
This commit also includes a test case written by Ying Yi as part of
http://reviews.llvm.org/D12163 (that patch originally addressed this issue but
was reverted because it introduced false positive regressions).
[WinEH] Pass the catch adjectives to catchpad directly
This avoids building a fake LLVM IR global variable just to ferry an i32
down into LLVM codegen. It also puts a nail in the coffin of using MS
ABI C++ EH with landingpads, since now we'll assert in the lpad code
when flags are present.
Anton Yartsev [Wed, 16 Sep 2015 18:12:15 +0000 (18:12 +0000)]
[analyzer] Improved behavior if Clang was not found, part II
- scan-build help: display 'Could not query Clang for the list of available checkers.' + the reason why it happened so if clang was not found.
- display requested/forced help in case of --use-analyzer=Xcode.
[analyzer] SATestBuild.py: Move additional checkers logic so SATestAdd.py can use it as well.
Move the logic looking for additional checkers in the SA_ADDITIONAL_CHECKERS
environmental variable from SATestBuild's main() to runScanBuild(). This allows
SATestAdd.py to use the variable as well. Without it, we won't include
additional checkers when building reference results for the build bot.
Fixed testing callback emission order to account for vptr.
Poison vtable in either complete or base dtor, depending on
if virtual bases exist. If virtual bases exist, poison in
complete dtor. Otherwise, poison in base.
Remove commented-out block.
Don't crash when passing &@selector to a _Nonnull parameter. Fixes PR24774.
The root cause here is that ObjCSelectorExpr is an rvalue, yet it can have its
address taken. That's kind of awkward, but fixing this is awkward in other
ways, see https://llvm.org/bugs/show_bug.cgi?id=24774#c16 . For now, just
fix the crash.
Chris Bieneman [Tue, 15 Sep 2015 21:52:42 +0000 (21:52 +0000)]
[CMake] Add cache scripts for Apple-style clang builds.
Summary:
These CMake cache scripts are my first pass at replicating Apple's packaging logic from autoconf. They can be used on any Darwin machine to approximate an Apple Clang build.
The included README file includes documentation and a sample CMake invocation.
Piotr Padlewski [Tue, 15 Sep 2015 21:46:55 +0000 (21:46 +0000)]
Decorating vptr load & stores with !invariant.group
Adding !invariant.group to vptr load/stores for devirtualization purposes.
For more goto:
http://lists.llvm.org/pipermail/cfe-dev/2015-July/044227.html
projects / clang / log