[OpenMP] Prevent nesting of target constructs within target code execution regions.
Summary:
This patch enhances Sema to check for the following restriction:
OpenMP 4.5 [2.17 Nesting of Regions]
If a target, target update, target data, target enter data, or
target exit data construct is encountered during execution of a
target region, the behavior is unspecified.
Artem Dergachev [Mon, 1 Feb 2016 09:29:17 +0000 (09:29 +0000)]
[analyzer] Use a wider integer type for an array index.
Avoids unexpected overflows while performing pointer arithmetics in 64-bit code.
Moreover, neither PointerDiffType nor 'int' can be used as a common array index
type because arrays may have size (and indexes) more than PTRDIFF_MAX but less
than SIZE_MAX.
Craig Topper [Sun, 31 Jan 2016 04:20:03 +0000 (04:20 +0000)]
Replace usage of llvm::utostr_32 with just llvm::utostr. While this is less efficient, its unclear that the one place using the _32 version was doing so for efficiency.
This patch adds doxygen comments for the intrinsincs in the header file __wmmintrin_aes.h.
The doxygen comments are automatically generated based on Sony's intrinsics document.
Richard Trieu [Fri, 29 Jan 2016 23:51:16 +0000 (23:51 +0000)]
Improve -Wconstant-conversion
Switch the evaluation from isIntegerConstantExpr to EvaluateAsInt.
EvaluateAsInt will evaluate more types of expressions than
isIntegerConstantExpr.
Move one case from -Wsign-conversion to -Wconstant-conversion. The case is:
1) Source and target types are signed
2) Source type is wider than the target type
3) The source constant value is positive
4) The conversion will store the value as negative in the target.
Manman Ren [Fri, 29 Jan 2016 23:46:55 +0000 (23:46 +0000)]
Class Property: generate metadata for class properties in protocols.
The list of class properties is saved in
Old ABI: protocol->ext->class_properties (protocol->ext->size will be updated)
New ABI: protocol->class_properties (protocol->size will be updated)
Manman Ren [Fri, 29 Jan 2016 23:45:01 +0000 (23:45 +0000)]
Class Property: generate metadata for class properties in categories.
The list of class properties is saved in
Old ABI: category->class_properties (category->size will be updated as well)
New ABI: category->class_properties (a flag in objc_image_info to indicate
whether or not the list of class properties is present)
This patch adds doxygen comments for the intrinsincs in the header file __wmmintrin_pclmul.h. The doxygen comments are automatically generated based on Sony's intrinsics document.
- Include the position of the argument on which the nullability is violated
- Differentiate between a 'method' and a 'function' in the message wording
- Test for the error message text in the tests
- Fix a bug with setting 'IsDirectDereference' which resulted in regular dereferences assumed to have call context.
Nathan Wilson [Fri, 29 Jan 2016 04:43:59 +0000 (04:43 +0000)]
[Concepts] Implement a portion of Concepts TS[dcl.spec.concept]p5 and p6:
Diagnose if the return type of a function concept or declaration type of a
variable concept is not bool.
Justin Lebar [Thu, 28 Jan 2016 23:58:28 +0000 (23:58 +0000)]
[CUDA] Generate CUDA's printf alloca in its function's entry block.
Summary:
This is necessary to prevent llvm from generating stacksave intrinsics
around this alloca. NVVM doesn't have a stack, and we don't handle said
intrinsics.
Manman Ren [Thu, 28 Jan 2016 23:36:05 +0000 (23:36 +0000)]
Class Property: change PropertyMap to include isClassProperty.
PropertyMap used to map IdentifierInfo (name of the property) to
ObjcPropertyDecl *. Now that a class property can have the same name as
an instance property, we change PropertyMap to map a pair <IdentifierInfo *,
unsigned> to ObjcPropertyDecl *.
Also update a few places from iterating over instance_properties to
iterating over all properties.
Manman Ren [Thu, 28 Jan 2016 23:29:02 +0000 (23:29 +0000)]
Check for frontend errors after releasing the Builder.
Frontend can emit errors when releaseing the Builder. If there are errors before
or when releasing the Builder, we reset the module to stop here before invoking
the backend.
Before this commit, clang will continue to invoke the backend and backend can
crash.
Devin Coughlin [Thu, 28 Jan 2016 22:23:34 +0000 (22:23 +0000)]
[analyzer] Suppress nullability warnings in copy, mutableCopy, and init families.
There are multiple, common idioms of defensive nil-checks in copy,
mutableCopy, and init methods in ObjC. The analyzer doesn't currently have the
capability to distinguish these idioms from true positives, so suppress all
warnings about returns in those families. This is a pretty blunt suppression
that we should improve later.
Nico Weber [Thu, 28 Jan 2016 19:25:00 +0000 (19:25 +0000)]
Include RecordDecls from anonymous unions in the AST.
For
void f() {
union { int i; };
}
clang used to omit the RecordDecl from the anonymous union from the AST.
That's because the code creating it only called PushOnScopeChains(), which adds
it to the current DeclContext, which here is the function's DeclContext. But
RecursiveASTVisitor doesn't descent into all decls in a FunctionDecl.
Instead, for DeclContexts that contain statements, return the RecordDecl so
that it can be included in the DeclStmt containing the VarDecl for the union.
Interesting bits from the AST before this change:
|-FunctionDecl
| `-CompoundStmt
| |-DeclStmt
| | `-VarDecl 0x589cd60 <col:3> col:3 implicit used 'union (anonymous at test.cc:3:3)' callinit
After this change:
-FunctionDecl
| `-CompoundStmt
| |-DeclStmt
| | |-CXXRecordDecl 0x4612e48 <col:3, col:18> col:3 union definition
| | | |-FieldDecl 0x4612f70 <col:11, col:15> col:15 referenced i 'int'
| | `-VarDecl 0x4613010 <col:3> col:3 implicit used 'union (anonymous at test.cc:3:3)' callinit
This is now closer to how anonymous struct and unions are represented as
members of structs. It also enabled deleting some one-off code in the
template instantiation code.
Finally, it fixes a crash with ASTMatchers, see the included test case
(this fixes http://crbug.com/580749).
Manman Ren [Thu, 28 Jan 2016 18:49:28 +0000 (18:49 +0000)]
Class Property: class property and instance property can have the same name.
Add "enum ObjCPropertyQueryKind" to a few APIs that used to only take the name
of the property: ObjCPropertyDecl::findPropertyDecl,
ObjCContainerDecl::FindPropertyDeclaration,
ObjCInterfaceDecl::FindPropertyVisibleInPrimaryClass,
ObjCImplDecl::FindPropertyImplDecl, and Sema::ActOnPropertyImplDecl.
ObjCPropertyQueryKind currently has 3 values:
OBJC_PR_query_unknown, OBJC_PR_query_instance, OBJC_PR_query_class
This extra parameter specifies that we are looking for an instance property with
the given name, or a class property with the given name, or any property with
the given name (if both exist, the instance property will be returned).
1. Make test case more focused and robust by focusing on what to be tested (linkage, icall) -- make it easier to validate
2. Testing linkages of data and counter variables instead of names. Counters and data are more relavant to be tested.
Oliver Stannard [Thu, 28 Jan 2016 10:07:34 +0000 (10:07 +0000)]
Add backend dignostic printer for unsupported features
Re-commit of r258950 after fixing layering violation.
Add backend dignostic printer for unsupported features
The related LLVM patch adds a backend diagnostic type for reporting
unsupported features, this adds a printer for them to clang.
In the case where debug location information is not available, I've
changed the printer to report the location as the first line of the
function, rather than the closing brace, as the latter does not give the
user any information. This also affects optimisation remarks.
Craig Topper [Thu, 28 Jan 2016 05:22:54 +0000 (05:22 +0000)]
[Lex] Share some common code between decimal and octal parsing in NumericLiteralParser.
There were a couple slight variations between the two copies that I don't believe were intentional. For example, only one of the paths checked for digit separations proceeding a '.', but I think the lexer itself splits the token if a digit separator proceeds a period.
In OpenCL, `bool` vectors are a reserved type, and are therefore
illegal.
Outside of OpenCL, if we try to make an extended vector of N `bool`s,
Clang will lower it to an `[N x i1]`. LLVM has no ABI for bitvectors, so
lots of operations on such vectors are thoroughly broken. As a result,
this patch makes them illegal in everything else, as well. :)
Tim Northover [Wed, 27 Jan 2016 19:32:40 +0000 (19:32 +0000)]
ARMv7k: select ABI based on v7k Arch rather than watchos OS.
Various bits we'd like to use the new ABI actually compile with "-arch armv7k
-miphoneos-version-min=9.0". Not ideal, but also not ridiculous given how
slices work.
John McCall [Wed, 27 Jan 2016 18:32:30 +0000 (18:32 +0000)]
Emit calls to objc_unsafeClaimAutoreleasedReturnValue when
reclaiming a call result in order to ignore it or assign it
to an __unsafe_unretained variable. This avoids adding
an unwanted retain/release pair when the return value is
not actually returned autoreleased (e.g. when it is returned
from a nonatomic getter or a typical collection accessor).
This runtime function is only available on the latest Apple
OS releases; the backwards-compatibility story is that you
don't get the optimization unless your deployment target is
recent enough. Sorry.
Oliver Stannard [Wed, 27 Jan 2016 17:30:28 +0000 (17:30 +0000)]
Add backend dignostic printer for unsupported features
The related LLVM patch adds a backend diagnostic type for reporting
unsupported features, this adds a printer for them to clang.
In the case where debug location information is not available, I've
changed the printer to report the location as the first line of the
function, rather than the closing brace, as the latter does not give the
user any information. This also affects optimisation remarks.
Devin Coughlin [Wed, 27 Jan 2016 01:41:58 +0000 (01:41 +0000)]
[analyzer] ObjCDeallocChecker: Only operate on classes with retained properties.
Previously the ObjC Dealloc Checker only checked classes with ivars, not
retained properties, which caused three bugs:
- False positive warnings about a missing -dealloc method in classes with only
ivars.
- Missing warnings about a missing -dealloc method on classes with only
properties.
- Missing warnings about an over-released or under-released ivar associated with
a retained property in classes with only properties.
The fix is to check only classes with at least one retained synthesized
property.
This also exposed a bug when reporting an over-released or under-released
property that did not contain a synthesize statement. The checker tried to
associate the warning with an @synthesize statement that did not exist, which
caused an assertion failure in debug builds. The fix is to fall back to the
@property statement in this case.
James Y Knight [Wed, 27 Jan 2016 01:04:51 +0000 (01:04 +0000)]
Revert "Change of UserLabelPrefix default value from "_" to """
This reverts commit r258504.
This commit breaks (at least) sparc-rtems -- the OS (RTEMS) used to
override UserLabelPrefix to "", despite the arch (SPARC) having set it
to "_". Now, the OS doesn't override anymore, but the arch sets it to
"_", resulting in the wrong value. I expect this probably breaks other
OSes that overrode to "" before, as well. (Clearly we have some missing
test cases, here...)
Devin Coughlin [Tue, 26 Jan 2016 23:58:48 +0000 (23:58 +0000)]
[analyzer] Body farm: Look for property ivar in shadowing readwrite property.
After r251874, readonly properties that are shadowed by a readwrite property
in a class extension no longer have an instance variable, which caused the body
farm to not synthesize getters. Now, if a readonly property does not have an
instance variable look for a shadowing property and try to get the instance
variable from there.
Artem Belevich [Tue, 26 Jan 2016 23:37:29 +0000 (23:37 +0000)]
[CUDA] Implemented device-side support functions in <cmath>.
CUDA expects math functions in std:: namespace to work on device side.
In order to make it work with clang without allowing device-side code
generation for functions w/o appropriate target attributes, this patch
provides device-side implementations for <cmath> functions. Most of
them call global-scope math functions provided by CUDA headers. In few
cases we use clang builtins.
Tested out-of tree by compiling and running thrust's unit_tests.
https://github.com/thrust/thrust/tree/master/testing
David Majnemer [Tue, 26 Jan 2016 23:14:47 +0000 (23:14 +0000)]
[WinEH] Annotate calls to __RTtypeid with a funclet bundle
Clang's CodeGen has several paths which end up invoking or calling a
function. The one that we used for calls to __RTtypeid did not
appropriately annotate the call with a funclet bundle.
Chris Bieneman [Tue, 26 Jan 2016 21:30:40 +0000 (21:30 +0000)]
Remove autoconf support
Summary:
This patch is provided in preparation for removing autoconf on 1/26. The proposal to remove autoconf on 1/26 was discussed on the llvm-dev thread here: http://lists.llvm.org/pipermail/llvm-dev/2016-January/093875.html
"This is the way [autoconf] ends
Not with a bang but a whimper."
-T.S. Eliot
David Majnemer [Tue, 26 Jan 2016 19:30:26 +0000 (19:30 +0000)]
[MS ABI] Allow a member pointers' converted type to change
Member pointers in the MS ABI are tricky for a variety of reasons.
The size of a member pointer is indeterminate until the program reaches
a point where the representation is required to be known. However,
*pointers* to member pointers may exist without knowing the pointee
type's representation. In these cases, we synthesize an opaque LLVM
type for the pointee type.
However, we can be in a situation where the underlying member pointer's
representation became known mid-way through the program. To account for
this, we attempted to manicure CodeGen's type-cache so that we can
replace the opaque member pointer type with the real deal while leaving
the pointer types unperturbed. This, unfortunately, is a problematic
approach to take as we will violate CodeGen's invariants.
These violations are mostly harmless but let's do the right thing
instead: invalidate the type-cache if a member pointer's LLVM
representation changes.
Justin Lebar [Tue, 26 Jan 2016 17:47:20 +0000 (17:47 +0000)]
[CUDA] Add -fcuda-allow-variadic-functions.
Summary:
Turns out the variadic function checking added in r258643 was too strict
for some existing users; give them an escape valve. When
-fcuda-allow-variadic-functions is passed, the front-end makes no
attempt to disallow C-style variadic functions. Calls to va_arg are
still not allowed.
Chad Rosier [Tue, 26 Jan 2016 15:52:05 +0000 (15:52 +0000)]
[Driver] Make sure -fno-math-builtin option is being passed by the driver.
Support for the -fno-math-builtin option was added in r186899. The codegen side
is being tested in test/CodeGen/nomathbuiltin.c. The missing part was just
passing the option through the driver.
Xiuli Pan [Tue, 26 Jan 2016 04:03:48 +0000 (04:03 +0000)]
Recommit: R258773 [OpenCL] Pipe builtin functions
Fix arc patch fuzz error.
Summary:
Support for the pipe built-in functions for OpenCL 2.0.
The pipe builtin functions may have infinite kinds of element types, one approach
would be to just generate calls that would always use generic types such as void*.
This patch is based on bader's opencl support patch on SPIR-V branch.
Richard Trieu [Tue, 26 Jan 2016 02:51:55 +0000 (02:51 +0000)]
Fix -Wnull-conversion for long macros.
Move the function to get a macro name from DiagnosticRenderer.cpp to Lexer.cpp
so that other files can use it. Lexer now has two functions to get the
immediate macro name, the newly added one is better for diagnostic purposes.
Make -Wnull-conversion use this function for better NULL macro detection.
Xiuli Pan [Tue, 26 Jan 2016 02:06:04 +0000 (02:06 +0000)]
[OpenCL] Pipe builtin functions
Summary:
Support for the pipe built-in functions for OpenCL 2.0.
The pipe builtin functions may have infinite kinds of element types, one approach
would be to just generate calls that would always use generic types such as void*.
This patch is based on bader's opencl support patch on SPIR-V branch.
projects / clang / log