Neil Hickey [Wed, 14 Dec 2016 13:18:48 +0000 (13:18 +0000)]
Fixing cast condition for removing casts from builtin FPClassification.
The function SemaBuiltinFPClassification removed superfluous float to double
casts, this was changed to also remove float to float casts but this isn't
valid in all cases, for example when doing an rvaluetolvalue cast. Added a
check to only remove if this was a conventional floating cast.
Added additional tests into SemaOpenCL/extensions to cover these cases
Stephan Bergmann [Wed, 14 Dec 2016 11:57:17 +0000 (11:57 +0000)]
Replace APFloatBase static fltSemantics data members with getter functions
At least the plugin used by the LibreOffice build
(<https://wiki.documentfoundation.org/Development/Clang_plugins>) indirectly
uses those members (through inline functions in LLVM/Clang include files in turn
using them), but they are not exported by utils/extract_symbols.py on Windows,
and accessing data across DLL/EXE boundaries on Windows is generally
problematic.
Paul Robinson [Wed, 14 Dec 2016 02:06:11 +0000 (02:06 +0000)]
[PS4] Undo dialect tweak for Objective-C.
In r267772, we had set the PS4's default dialect for both C and
Objective-C to gnu99. Make that change only for C; we don't really
support Objective-C/C++ so there's no point fiddling the dialect.
Richard Smith [Wed, 14 Dec 2016 00:03:17 +0000 (00:03 +0000)]
Remove custom handling of array copies in lambda by-value array capture and
copy constructors of classes with array members, instead using
ArrayInitLoopExpr to represent the initialization loop.
This exposed a bug in the static analyzer where it was unable to differentiate
between zero-initialized and unknown array values, which has also been fixed
here.
CodeGen: Start using inrange annotations on vtable getelementptr.
This annotation allows the optimizer to split vtable groups, as permitted by
a change to the Itanium ABI [1] that prevents compilers from adjusting virtual
table pointers between virtual tables.
CodeGen: New vtable group representation: struct of vtable arrays.
In a future change, this representation will allow us to use the new inrange
annotation on getelementptr to allow the optimizer to split vtable groups.
Reid Kleckner [Tue, 13 Dec 2016 19:48:32 +0000 (19:48 +0000)]
Align EvalInfo in ExprConstant to avoid PointerUnion assertions
32-bit MSVC doesn't provide more than 4 byte stack alignment by default.
This conflicts with PointerUnion's attempt to make assertions about
alignment. This fixes the problem by explicitly asking the compiler for
8 byte alignment.
Reid Kleckner [Tue, 13 Dec 2016 18:58:09 +0000 (18:58 +0000)]
__uuidof() and declspec(uuid("...")) should be allowed on enumeration types
Although not specifically mentioned in the documentation, MSVC accepts
__uuidof(…) and declspec(uuid("…")) attributes on enumeration types in
addition to structs/classes. This is meaningful, as such types *do* have
associated UUIDs in ActiveX typelibs, and such attributes are included
by default in the wrappers generated by their #import construct, so they
are not particularly unusual.
clang currently rejects the declspec with a –Wignored-attributes
warning, and errors on __uuidof() with “cannot call operator __uuidof on
a type with no GUID” (because it rejected the uuid attribute, and
therefore finds no value). This is causing problems for us while trying
to use clang-tidy on a codebase that makes heavy use of ActiveX.
I believe I have found the relevant places to add this functionality,
this patch adds this case to clang’s implementation of these MS
extensions. patch is against r285994 (or actually the git mirror 80464680ce).
Both include an update to test/Parser/MicrosoftExtensions.cpp to
exercise the new functionality.
This is my first time contributing to LLVM, so if I’ve missed anything
else needed to prepare this for review just let me know!
Artem Dergachev [Tue, 13 Dec 2016 17:19:18 +0000 (17:19 +0000)]
[analyzer] Detect ObjC properties that are both (copy) and Mutable.
When an Objective-C property has a (copy) attribute, the default setter
for this property performs a -copy on the object assigned.
Calling -copy on a mutable NS object such as NSMutableString etc.
produces an immutable object, NSString in our example.
Hence the getter becomes type-incorrect.
[Modules] Make header inclusion order from umbrella dirs deterministic
Sort the headers by name before adding the includes in
collectModuleHeaderIncludes. This makes the include order for building
umbrellas deterministic across different filesystems and also guarantees
that the ASTWriter always dump top headers in the same order.
There's currently no good way to test for this behavior.
[CrashReproducer] Collect PCH included via -include-pch
Collect the necessary input PCH files.
Do not try to validate the AST before copying it out because if the
crash is in this path, we won't be able to collect it. Instead only
check if it's a file containg an AST.
Bring back note about not supporting global register variables.
This was accidentally removed in r260506, even though we only support
non-allocatable global register variables. The general (allocatable) case
is explicitly not supported.
This adds a way for us to version any UBSan handler by itself.
The patch overrides D21289 for a better implementation (we're able to
rev up a single handler).
After this, then we can land a slight modification of D19667+D19668.
We probably don't want to keep all the versions in compiler-rt (maybe we
want to deprecate on one release and remove the old handler on the next
one?), but with this patch we will loudly fail to compile when mixing
incompatible handler calls, instead of silently compiling and then
providing bad error messages.
Michal Gorny [Mon, 12 Dec 2016 15:07:43 +0000 (15:07 +0000)]
[Driver] Fix finding multilib gcc install on Gentoo (with gcc-config)
Fix the gcc-config code to support multilib gcc installs properly. This
solves two problems: -mx32 using the 64-bit gcc directory (due to matching
installation triple), and -m32 not respecting gcc-config at all (due to
mismatched installation triple).
In order to fix the former issue, split the multilib scan out of
Generic_GCC::GCCInstallationDetector::ScanLibDirForGCCTriple() (the code
is otherwise unchanged), and call it for each installation found via
gcc-config.
In order to fix the latter issue, split the gcc-config processing out of
Generic_GCC::GCCInstallationDetector::init() and repeat it for all
triples, including extra and biarch triples. The only change
in the gcc-config code itself is adding the call to multilib scan.
Convert the gentoo_linux_gcc_multi_version_tree test input to multilib
x86_64+32+x32 install, and add appropriate tests to linux-header-search
and linux-ld.
Haojian Wu [Mon, 12 Dec 2016 14:12:10 +0000 (14:12 +0000)]
[StaticAnalysis] Remove unnecessary parameter in CallGraphNode::addCallee.
Summary:
Remove the CallGraph in addCallee as it is not used in addCallee.
It decouples addCallee from CallGraph, so that we can use CallGraphNode
within our customized CallGraph.
Daniel Jasper [Mon, 12 Dec 2016 12:42:29 +0000 (12:42 +0000)]
clang-format: Separate out a language kind for ObjC.
While C(++) and ObjC are generally formatted the same way and can be
mixed, people might want to choose different styles based on the
language. This patch recognizes .m and .mm files as ObjC and also
implements a very crude detection of whether or not a .h file contains
ObjC code. This can be improved over time.
Also move most of the ObjC tests into their own test file to keep file
size maintainable.
I made the wrong assumption that execution would continue after an error Diag
which led to unneeded complex code.
This patch aligns with the better implementation of ToolChain::GetRuntimeLibType.
Richard Smith [Mon, 12 Dec 2016 02:53:20 +0000 (02:53 +0000)]
Add two new AST nodes to represent initialization of an array in terms of
initialization of each array element:
* ArrayInitLoopExpr is a prvalue of array type with two subexpressions:
a common expression (an OpaqueValueExpr) that represents the up-front
computation of the source of the initialization, and a subexpression
representing a per-element initializer
* ArrayInitIndexExpr is a prvalue of type size_t representing the current
position in the loop
This will be used to replace the creation of explicit index variables in lambda
capture of arrays and copy/move construction of classes with array elements,
and also C++17 structured bindings of arrays by value (which inexplicably allow
copying an array by value, unlike all of C++'s other array declarations).
No uses of these nodes are introduced by this change, however.
Include headermaps (.hmap files) in the .cache directory and
add VFS entries. All headermaps are known after HeaderSearch
setup, collect them right after.
Craig Topper [Sun, 11 Dec 2016 01:26:52 +0000 (01:26 +0000)]
[AVX-512] Remove masking from 512-bit vpermil builtins. The backend now has versions without masking so wrap it with select.
This will allow the backend to constant fold these to generic shuffle vectors like 128-bit and 256-bit without having to working about handling masking.
Craig Topper [Sat, 10 Dec 2016 23:09:52 +0000 (23:09 +0000)]
[AVX-512] Remove masking from 512-bit pshufb builtin. The backend now has a version without masking so wrap it with select.
This will allow the backend to constant fold these to generic shuffle vectors like 128-bit and 256-bit without having to working about handling masking.
Devin Coughlin [Sat, 10 Dec 2016 01:16:09 +0000 (01:16 +0000)]
[analyzer] Improve VirtualCallChecker diagnostics and move into optin package.
The VirtualCallChecker is in alpha because its interprocedural diagnostics
represent the call path textually in the diagnostic message rather than with a
path sensitive diagnostic.
This patch turns off the AST-based interprocedural analysis in the checker so
that no call path is needed and improves with diagnostic text. With these
changes, the checker is ready to be moved into the optin package.
Ultimately the right fix is to rewrite this checker to be path sensitive -- but
there is still value in enabling the checker for intraprocedural analysis only
The interprocedural mode can be re-enabled with an -analyzer-config flag.
Vedant Kumar [Sat, 10 Dec 2016 00:54:13 +0000 (00:54 +0000)]
[clang-format] Another attempt at python 3 compatibility
The entries in vim.current.buffer appear to be decoded strings, which
means that python3 won't allow invoking 'decode' on them. Keep the old
behavior when running under python2, but skip the error-inducing decode
step with python3..
Vedant Kumar [Fri, 9 Dec 2016 23:48:18 +0000 (23:48 +0000)]
[ubsan] Treat ObjC's BOOL as if its range is always {0, 1}
On some Apple platforms, the ObjC BOOL type is defined as a signed char.
When performing instrumentation for -fsanitize=bool, we'd like to treat
the range of BOOL like it's always {0, 1}. While we can't change clang's
IRGen for char-backed BOOL's due to ABI compatibility concerns, we can
teach ubsan to catch potential abuses of this type.
Graydon Hoare [Fri, 9 Dec 2016 21:45:49 +0000 (21:45 +0000)]
[modules] Add optional out-param to ASTReader::ReadAST for imported submodules.
Summary:
The Swift frontend is acquiring the ability to load non-module PCH files containing
bridging definitions from C/ObjC. As part of this work, it needs to know which submodules
were imported by a PCH in order to wrap them in local Swift modules. This information
is collected by ASTReader::ReadAST in a local vector, but is currently kept private.
The change here is just to make the type of the vector elements public, and provide
an optional out-parameter to the ReadAST method to provide the vector's contents to
a caller after a successful read.
Reid Kleckner [Fri, 9 Dec 2016 21:10:43 +0000 (21:10 +0000)]
Remove special error recovery for ::(id)
The code pattern used to implement the token rewriting hack doesn't
interact well with token caching in the pre-processor. As a result,
clang would crash on 'int f(::(id));' while doing a tenative parse of
the contents of the outer parentheses. The original code from PR11852
still doesn't crash the compiler.
This error recovery also often does the wrong thing with member function
pointers. The test case from the original PR doesn't recover the right
way either:
void S::(*pf)() = S::f; // should be 'void (S::*pf)()'
Instead we were recovering as 'void S::*pf()', which is still wrong.
If we still think that users mistakenly parenthesize identifiers in
nested name specifiers, we should change clang to intentionally parse
that form with an error, rather than doing a token rewrite.
Fixes PR26623, but I think there will be many more bugs like this around
token rewriting in the parser.
Reid Kleckner [Fri, 9 Dec 2016 19:47:58 +0000 (19:47 +0000)]
Improve error message when referencing a non-tag type with a tag
Other compilers accept invalid code here that we reject, and we need a
better error message to try to convince users that the code is really
incorrect. Consider:
class Foo {
typedef MyIterHelper<Foo> iterator;
friend class iterator;
};
Previously our wording was "elaborated type refers to a typedef".
"elaborated type" isn't widely known terminology, so the new diagnostic
says "typedef 'iterator' cannot be referenced with class specifier".
Yaxun Liu [Fri, 9 Dec 2016 19:01:11 +0000 (19:01 +0000)]
Add support for non-zero null pointer for C and OpenCL
In amdgcn target, null pointers in global, constant, and generic address space take value 0 but null pointers in private and local address space take value -1. Currently LLVM assumes all null pointers take value 0, which results in incorrectly translated IR. To workaround this issue, instead of emit null pointers in local and private address space, a null pointer in generic address space is emitted and casted to local and private address space.
Tentative definition of global variables with non-zero initializer will have weak linkage instead of common linkage since common linkage requires zero initializer and does not have explicit section to hold the non-zero value.
Virtual member functions getNullPointer and performAddrSpaceCast are added to TargetCodeGenInfo which by default returns ConstantPointerNull and emitting addrspacecast instruction. A virtual member function getNullPointerValue is added to TargetInfo which by default returns 0. Each target can override these virtual functions to get target specific null pointer and the null pointer value for specific address space, and perform specific translations for addrspacecast.
Wrapper functions getNullPointer is added to CodegenModule and getTargetNullPointerValue is added to ASTContext to facilitate getting the target specific null pointers and their values.
This change has no effect on other targets except amdgcn target. Other targets can provide support of non-zero null pointer in a similar way.
This change only provides support for non-zero null pointer for C and OpenCL. Supporting for other languages will be added later incrementally.
Richard Smith [Fri, 9 Dec 2016 18:49:13 +0000 (18:49 +0000)]
DR1295 and cleanup for P0135R1: Make our initialization code more directly
mirror the description in the standard. Per DR1295, this means that binding a
const / rvalue reference to a bit-field no longer "binds directly", and per
P0135R1, this means that we materialize a temporary in reference binding
after adjusting cv-qualifiers and before performing a derived-to-base cast.
In C++11 onwards, this should have fixed the last case where we would
materialize a temporary of the wrong type (with a subobject adjustment inside
the MaterializeTemporaryExpr instead of outside), but we still have to deal
with that possibility in C++98, unless we want to start using xvalues to
represent materialized temporaries there too.
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers.
Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font.
In the past, \c command was used, unfortunately it applied to only one word.
<c> .. </c> has the same meaning, but applies to all words in between the tags.
Nico Weber [Fri, 9 Dec 2016 17:32:52 +0000 (17:32 +0000)]
Don't assert when redefining a built-in macro in a PCH, PR29119
PCH files store the macro history for a given macro, and the whole history list
for one identifier is given to the Preprocessor at once via
Preprocessor::setLoadedMacroDirective(). This contained an assert that no macro
history exists yet for that identifier. That's usually true, but it's not true
for builtin macros, which are created in Preprocessor() before flags and pchs
are processed. Luckily, ASTWriter stops writing macro history lists at builtins
(see shouldIgnoreMacro() in ASTWriter.cpp), so the head of the history list was
missing for builtin macros. So make the assert weaker, and splice the history
list to the existing single define for builtins.
Reid Kleckner [Fri, 9 Dec 2016 17:14:05 +0000 (17:14 +0000)]
Store decls in prototypes on the declarator instead of in the AST
This saves two pointers from FunctionDecl that were being used for some
rare and questionable C-only functionality. The DeclsInPrototypeScope
ArrayRef was added in r151712 in order to parse this kind of C code:
enum e {x, y};
int f(enum {y, x} n) {
return x; // should return 1, not 0
}
The challenge is that we parse 'int f(enum {y, x} n)' it its own
function prototype scope that gets popped before we build the
FunctionDecl for 'f'. The original change was doing two questionable
things:
1. Saving all tag decls introduced in prototype scope on a TU-global
Sema variable. This is problematic when you have cases like this, where
'x' and 'y' shouldn't be visible in 'f':
void f(void (*fp)(enum { x, y } e)) { /* no x */ }
This patch fixes that, so now 'f' can't see 'x', which is consistent
with GCC.
2. Storing the decls in FunctionDecl in ActOnFunctionDeclarator so that
they could be used in ActOnStartOfFunctionDef. This is just an
inefficient way to move information around. The AST lives forever, but
the list of non-parameter decls in prototype scope is short lived.
Moving these things to the Declarator solves both of these issues.
[CrashReproducer] Provide a clean dir path for -fmodules-cache-path
The most common workflow with module reproducers involves deleting the
module cache before running the script. This happens because leftovers
from the crash are present in the cache and could trigger unrelated and
confusing errors, misleading from the initial reproduction intent.
Change this to point to a clean path but leave the leftovers untouched.
When -fmodules is on, the reproducer invocation currently leave paths
for include-like flags as is. If the path is relative, the reproducer
doesn't know how to access that file during reproduction time because
the VFS cannot reason about relative paths.
Expand relative paths to absolute ones when creating the reproducer
command line. This allows, for example, the reproducer to work for
crashes while building clang with modules; this wasn't possible before
because building clang requires using relative inc dir from within the
build directory.
Paul Robinson [Fri, 9 Dec 2016 01:20:40 +0000 (01:20 +0000)]
Specify -std=gnu++98 on some Rewriter tests. NFC.
Rewriter tests rewrite Objective-C++ to C++, and then compile the
result. The rewritten result sometimes doesn't work with C++11. As
we want to allow Clang's default dialect to become C++11, we need to
make sure the tests will still pass.
Improved doxygen comments for fxsrintrin.h and mmintrin.h intrinsics by taagging parameter names with \a doxygen command to display parameters in italics.
David Gross [Thu, 8 Dec 2016 21:15:17 +0000 (21:15 +0000)]
[DebugInfo] Relax test case for long double constants.
Summary:
D27549 (partial fix for PR26619) emits a constant value in the debug
metadata for a floating-point static const that does not exceed 64
bits in size. The regression test accompanying that fix assumes that
a long double exceeds 64 bits in size and hence does not get a
constant value in the debug metadata. However, for some targets --
such as "--target=hexagon-unknown-elf" -- a long double does not
exceed 64 bits in size, and hence the test fails.
As a temporary fix, modify the regression test to no longer inspect
the debug metadata for a long double.
David Gross [Thu, 8 Dec 2016 20:02:46 +0000 (20:02 +0000)]
[DebugInfo] Add support for __fp16, float, and double constants.
Summary:
Partial fix for PR26619.
Prior to this change, a DIGlobalVariable corresponding to a static
const was marked with an expression corresponding to its constant
value only if it is of integral type. With this change, we now do the
same if it is of __fp16, float, or double type (that is,
floating-point types that do not exceed 64 bits in size, and hence are
supported easily by the existing LLVM machinery for creating constant
expressions in debug info).
Artem Belevich [Thu, 8 Dec 2016 19:38:13 +0000 (19:38 +0000)]
[CUDA] Ignore implicit target attributes during function template instantiation.
Some functions and templates are treated as __host__ __device__ even
when they don't have explicitly specified target attributes.
What's worse, this treatment may change depending on command line
options (-fno-cuda-host-device-constexpr) or
#pragma clang force_cuda_host_device.
Combined with strict checking for matching function target that comes
with D25809(r288962), it makes it hard to write code which would
explicitly instantiate or specialize some functions regardless of
pragmas or command line options in effect.
This patch changes the way we match target attributes of base template
vs attributes used in explicit instantiation or specialization so that
only explicitly specified attributes are considered. This makes base
template selection behave consistently regardless of pragma of command
line options that may affect CUDA target.
Improved doxygen comments for __wmmintrin_pclmul.h and ammintrin.h intrinsics by taagging parameter names with \a doxygen command to display parameters in italics.
Alex Lorenz [Thu, 8 Dec 2016 15:09:40 +0000 (15:09 +0000)]
[CodeCompletion] Provide Objective-C class property completion results
This commit provides class property code completion results. It supports
explicit and implicit class properties, but the special block completion is done
only for explicit properties right now.
Alex Lorenz [Thu, 8 Dec 2016 14:46:05 +0000 (14:46 +0000)]
[Sema] Avoid "case value not in enumerated type" warning for C++11 opaque enums
This commit ensures that the switch warning "case value not in enumerated type"
isn't shown for opaque enums. We don't know the actual list of values in opaque
enums, so that warning is incorrect.
Artem Dergachev [Thu, 8 Dec 2016 14:05:48 +0000 (14:05 +0000)]
[analyzer] Add dispatch_data_create as a special case in RetainCountChecker.
This function receives a callback block. The analyzer suspects that this block
may be used to take care of releasing the libdispatch object returned from
the function. In fact, it doesn't - it only releases the raw data buffer.
Inform the analyzer about that. Fixes the resulting false negatives.
Richard Smith [Thu, 8 Dec 2016 03:24:55 +0000 (03:24 +0000)]
[c++1z] P0490R0, NB comment GB 20: if std::tuple_size<T> is complete, use the
tuple-like interpretation of decomposition declaration even if there is no
::value member. We already did this, anticipating this resolution, just update
comments and tweak a testcase.
We continue to support dynamic exception specifications in C++1z as an
extension, but produce an error-by-default warning when we encounter one. This
allows users to opt back into the feature with a warning flag, and implicitly
opts system headers back into the feature should they happen to use it.
There is one semantic change implied by P0003R5 but not implemented here:
violating a throw() exception specification should now call std::terminate
directly instead of calling std::unexpected(), but since P0003R5 also removes
std::unexpected() and std::set_unexpected, and the default unexpected handler
calls std::terminate(), a conforming C++1z program cannot tell that we are
still calling it. The upside of this strategy is perfect backwards
compatibility; the downside is that we don't get the more efficient 'noexcept'
codegen for 'throw()'.
David L. Jones [Wed, 7 Dec 2016 23:41:58 +0000 (23:41 +0000)]
Refactor how the MSVC toolchain searches for a compatibility version.
Summary:
The MSVC toolchain and Clang driver combination currently uses a fairly complex
sequence of steps to determine the MS compatibility version to pass to cc1.
There is some oddness in this sequence currently, with some code which inspects
flags in the toolchain, and some code which inspects the triple and local
environment in the driver code.
This change is an attempt to consolidate most of this logic so that
Win32-specific code lives in MSVCToolChain.cpp. I'm not 100% happy with the
split, so any suggestions are welcome.
There are a few things you might want to watch for for specifically:
- On all platforms, if MSVC compatibility flags are provided (and valid), use
those.
- The fallback sequence should be the same as before, but is now consolidated
into MSVCToolChain::getMSVCVersion:
- Otherwise, try to use the Triple.
- Otherwise, on Windows, check the executable.
- Otherwise, on Windows or with --fms-extensions, default to 18.
- Otherwise, we can't determine the version.
- MSVCToolChain::ComputeEffectiveTriple no longer calls the base
ToolChain::ComputeEffectiveClangTriple. The only thing it would change for
Windows the architecture, which we don't care about for the compatibility
version.
- I'm not sure whether this is philosophically correct (but it should
be easy to add back to MSVCToolChain::getMSVCVersionFromTriple if not).
- Previously, Tools.cpp just called getTriple() anyhow, so it doesn't look
like the effective triple was always being used previously anyhow.
David L. Jones [Wed, 7 Dec 2016 23:39:44 +0000 (23:39 +0000)]
Add more tests for MSVC version handling.
Summary:
This change adds more test cases for the default MSVC compatibility version:
1. When -fms-extensions is supplied, but -fmsc-version and
-fms-compatibility-version are not.
2. With the target triple specifies an MSVC environment, but no other
-fms* flags.
projects / clang / log