Tim Northover [Wed, 12 Feb 2014 12:04:59 +0000 (12:04 +0000)]
ARM NEON: fix range checking on immediates.
Previously, range checking on the __builtin_neon_XYZ_v Clang intrinsics didn't
take account of the type actually passed to the call, which meant a request
like "vext_s16(a, b, 7)" was allowed through (TableGen was conservative and
allowed 0-7 for all types). This caused an assert in the backend because the
lane doesn't make sense.
Dmitri Gribenko [Wed, 12 Feb 2014 10:33:14 +0000 (10:33 +0000)]
Add an option to allow Clang verify source files for a module only once during
the build
When Clang loads the module, it verifies the user source files that the module
was built from. If any file was changed, the module is rebuilt. There are two
problems with this:
1. correctness: we don't verify system files (there are too many of them, and
stat'ing all of them would take a lot of time);
2. performance: the same module file is verified again and again during a
single build.
This change allows the build system to optimize source file verification. The
idea is based on the fact that while the project is being built, the source
files don't change. This allows us to verify the module only once during a
single build session. The build system passes a flag,
-fbuild-session-timestamp=, to inform Clang of the time when the build started.
The build system also requests to enable this feature by passing
-fmodules-validate-once-per-build-session. If these flags are not passed, the
behavior is not changed. When Clang verifies the module the first time, it
writes out a timestamp file. Then, when Clang loads the module the second
time, it finds a timestamp file, so it can compare the verification timestamp
of the module with the time when the build started. If the verification
timestamp is too old, the module is verified again, and the timestamp file is
updated.
Jonathan Roelofs [Wed, 12 Feb 2014 03:21:20 +0000 (03:21 +0000)]
Add Multilib selection machinery
This patch improves the support for picking Multilibs from gcc installations.
It also provides a better approximation for the flags '-print-multi-directory'
and '-print-multi-lib'.
This reverts r201203 (i.e. re-applying r201202 with small fixes in
unittests/CMakeLists.txtto make the build bots happy).
Jonathan Roelofs [Wed, 12 Feb 2014 01:29:25 +0000 (01:29 +0000)]
Add Multilib selection machinery
This patch improves the support for picking Multilibs from gcc installations.
It also provides a better approximation for the flags '-print-multi-directory'
and '-print-multi-lib'.
David Majnemer [Wed, 12 Feb 2014 00:43:02 +0000 (00:43 +0000)]
MS ABI: vptr injection should obey alignment requirements
vptr injection must inject padding equivalent to the alignment of the
most aligned non-virtual subobject, not the alignment of the enclosing
record.
To fascilitate this change, don't let record layout observe the
alignment of the record until we've injected our vptrs. Also, do not
allow the alignment of vbases to affect required alignment until just
before we insert the vtordisp field.
Expose the name of the checker producing each diagnostic message.
Summary:
In clang-tidy we'd like to know the name of the checker producing each
diagnostic message. PathDiagnostic has BugType and Category fields, which are
both arbitrary human-readable strings, but we need to know the exact name of the
checker in the form that can be used in the CheckersControlList option to
enable/disable the specific checker.
This patch adds the CheckName field to the CheckerBase class, and sets it in
the CheckerManager::registerChecker() method, which gets them from the
CheckerRegistry.
Checkers that implement multiple checks have to store the names of each check
in the respective registerXXXChecker method.
David Blaikie [Tue, 11 Feb 2014 21:37:47 +0000 (21:37 +0000)]
Remove bad debug info test.
Another test that's testing through assembly from Clang which is
problematic. An attempt to upgrade this to just be an IR-generation test
was unsuccessful (I was unable to get this test to fail) but it looks
like there's other, better, test coverage in this area
(test/CodeGenObjC/debuginfo-properties.m) anyway.
David Blaikie [Tue, 11 Feb 2014 21:16:44 +0000 (21:16 +0000)]
Remove bad debug info test.
This test case doesn't belong in Clang (it's testing IndVarSimplify) but
in an effort to reproduce the test case this was intended to cover (by
essentially reverting r134441) I wasn't able to reproduce the failure
this test case should've produced. So I haven't ported this down to
LLVM, instead I'm just deleting it.
I suspect the test is just underconstrained, but I've no great interest
in trying hard to fix it right now - if anyone else wants to, I'd be
more than welcome to that.
Jordan Rose [Tue, 11 Feb 2014 17:27:59 +0000 (17:27 +0000)]
'nonnull(1)' on a block parameter should apply to the block's argument.
Thanks to r199467, __attribute__((nonnull)) (without arguments) can apply
directly to parameters, instead of being applied to the whole function.
However, the old form of nonnull (with an argument index) could also apply
to the arguments of function and block pointers, and both of these can be
passed as parameters.
Now, if 'nonnull' with an argument is found on a parameter, /and/ the
parameter is a function or block pointer, it is handled the old way.
Dmitri Gribenko [Tue, 11 Feb 2014 14:34:14 +0000 (14:34 +0000)]
libclang: audit all APIs that accept a CXTranslationUnit to make sure that
checks for invalid translation unit are in place. Also, while there, add log
output for this case.
Tim Northover [Tue, 11 Feb 2014 11:27:44 +0000 (11:27 +0000)]
AArch64: share slgihtly more NEON implementation with ARM.
The s64/u64 vcvt conversion operations are actually pretty much identical to
the s32/u32 ones in implementation, and can be shared with just one extra
variable.
Robert Lytton [Tue, 11 Feb 2014 10:34:51 +0000 (10:34 +0000)]
XCore target: add section information.
Xcore target ABI requires const data that is externally visible
to be handled differently if it has C-language linkage rather than
C++ language linkage.
Oliver Stannard [Tue, 11 Feb 2014 09:25:50 +0000 (09:25 +0000)]
AAPCS: Do not split structs after CPRC allocated on stack
According to the AAPCS, we can split structs between GPRs and the stack,
except for when an argument has already been allocated on the stack. This
can occur when a large number of floating-point arguments fill up the VFP
registers, and are alllocated on the stack before the general-purpose argument
registers are full.
Ted Kremenek [Tue, 11 Feb 2014 06:29:29 +0000 (06:29 +0000)]
Fix PCH deserialization bug with local static symbols being treated as local extern.
This triggered a miscompilation of code using Boost's function_template.hpp
when it was included inside a PCH file. A local static within
that header would be treated as local extern, resulting in the wrong
mangling. This only occurred during PCH deserialization.
Josh Magee [Tue, 11 Feb 2014 01:35:14 +0000 (01:35 +0000)]
[stackprotector] Add command line option -fstack-protector-strong
This option has the following effects:
* It adds the sspstrong IR attribute to each function within the CU.
* It defines the macro __SSP_STRONG__ with the value of 2.
David Majnemer [Mon, 10 Feb 2014 19:50:15 +0000 (19:50 +0000)]
MS ABI: Add support for #pragma pointers_to_members
Introduce a notion of a 'current representation method' for
pointers-to-members.
When starting out, this is set to 'best case' (representation method is
chosen by examining the class, selecting the smallest representation
that would work given the class definition or lack thereof).
This pragma allows the translation unit to dictate exactly what
representation to use, similar to how the inheritance model keywords
operate.
Tim Northover [Mon, 10 Feb 2014 16:20:36 +0000 (16:20 +0000)]
ARM: move vshll NEON implementation to common code
Now that both ARM backends use the same implementation for vshll operations,
the code can be shared. This is also a necessary LLVM/Clang interface update.
NAKAMURA Takumi [Mon, 10 Feb 2014 10:51:09 +0000 (10:51 +0000)]
Fix Clang install rules to not set permissions on include/
The CMake install(DIRECTORY) command documents that it sets permissions
on directories it is asked to install. Since the <prefix>/include
directory may not be exclusive to the LLVM/Clang installation, we should
not ask CMake to manage permissions of that directory for us. Instead,
give only our own include/clang and include/clang-c subdirectories to
the install(DIRECTORY) command.
Kaelyn Uhrain [Sun, 9 Feb 2014 21:47:04 +0000 (21:47 +0000)]
PR18685: Ignore class template specializations as potential
nested-name-specifiers for typos unless the typo already has
a nested-name-specifier that is a template specialization.
David Blaikie [Sun, 9 Feb 2014 06:54:23 +0000 (06:54 +0000)]
Improve diagnostic for using non-class/namespace/scoped enum in a nested name specifier.
Rather than simply saying "X is not a class or namespace", clarify what
X is by providing the aka type in the case where X is a type, or
pointing to the named declaration if there's an unambiguous one to refer
to. In the ambiguous case, the ambiguities are already enumerated
(though could be clarified by describing what kind of entities they are)
Included a few FIXMEs in tests where some further improvements could be
made.
Richard Smith [Sun, 9 Feb 2014 00:54:43 +0000 (00:54 +0000)]
PR16519, PR18009: When checking a partial specialization for uses of its own
template parameters, don't look for parameters of outer templates. If a problem
is found in a default template argument, point the diagnostic at the partial
specialization (with a note pointing at the default argument) instead of
pointing it at the default argument and leaving it unclear which partial
specialization os problematic.
John McCall [Sat, 8 Feb 2014 03:26:05 +0000 (03:26 +0000)]
type_info objects are not unnamed_addr: the ABI requires us to
unique them and permits the implementation of dynamic_cast (and
anything else which knows it's working with a complete class
type) to compare their addresses directly.
Reid Kleckner [Sat, 8 Feb 2014 02:40:20 +0000 (02:40 +0000)]
Move the -fms-compatibility using decl check after real access checking
Summary:
This avoids false positives from -Wmicrosoft when name lookup would
normally succeed in standard C++. This triggered on a common CRTP
pattern in clang, where a derived class would have a private using decl
to pull in members of a dependent base:
class Verifier : InstVisitor<Verifier> {
private:
using InstVisitor<Verifier>::visit;
...
void anything() {
visit(); // warned here
}
};
Real access checks pass here because we're in the context of the
Verifier, but the -Wmicrosoft extension was just looking for the private
access specifier.
Richard Smith [Fri, 7 Feb 2014 22:51:16 +0000 (22:51 +0000)]
PR16638, DR1552: the exception specification on an implicitly-declared
'operator delete' or 'operator delete[]' is an explicit exception
specification. Therefore we should diagnose 'void operator delete(void*)'
instead of 'void operator delete(void*) noexcept'.
This diagnostic remains an ExtWarn, since in practice people don't always
include the exception specification in such a declaration.
Ben Langmuir [Fri, 7 Feb 2014 17:31:11 +0000 (17:31 +0000)]
Stat system dependencies when using -verify-pch
We don't stat the system headers to check for stalenes during regular
PCH loading for performance reasons. When explicitly saying
-verify-pch, we want to check all the dependencies - user or system.
Oliver Stannard [Fri, 7 Feb 2014 11:25:57 +0000 (11:25 +0000)]
Fix AAPCS compliance for HFAs containing doubles and long doubles
An HFA is defined as a struct containing floating point values of the
same machine type. In the 32-bit ABI, double and long double have the
same machine type, so a struct with a mixture of these types must be an
HFA (assuming it meets the other criteria).
Revert r194097: "With this patch -Wwrite-strings is still implemented with the terrible
hack of passing -fconst-strings to -cc1"
Passing or not a language option based on diagnostic settings is a bad idea, it breaks
using a PCH that was compiled with different diagnostic settings.
Also add a test case to make sure we don't regress.
David Majnemer [Fri, 7 Feb 2014 00:43:07 +0000 (00:43 +0000)]
MS ABI: Don't be so hasty to judge an inheritance model
If we are in the middle of defining the class, don't attempt to
validate previously annotated declarations. We may not have seen base
specifiers or virtual method declarations yet.
Richard Smith [Thu, 6 Feb 2014 23:35:16 +0000 (23:35 +0000)]
Temporary fix for PR18473: Don't try to evaluate the initializer for a
type-dependent variable, even if the initializer isn't value-dependent. This
happens for ParenListExprs composed of non-value-dependent subexpressions, for
instance.
We should really give ParenListExprs (and InitListExprs) the type of the
initialized entity if they're used to represent a dependent initialization (and
if so, set them to be type-, value- and instantiation-dependent).
Ben Langmuir [Thu, 6 Feb 2014 18:53:25 +0000 (18:53 +0000)]
Move -verify-pch to use VerifyJobAction
Use the verify hook rather than the compile hook to represent the
-verify-pch action, and move the exising --verify-debug-info action
into its own subclass of VerifyJobAction. Incidentally change the name
printed by -ccc-print-phases for --verify-debug-info.
David Majnemer [Thu, 6 Feb 2014 10:59:19 +0000 (10:59 +0000)]
MS ABI: Tweak pointer-to-member mangling/inheritance model selection
Properly determine the inheritance model when dealing with nullptr:
- If a nullptr template argument is being checked against
pointer-to-member parameter, nail down an inheritance model.
N.B. We will chose an inheritance model even if we won't ultimately
choose the template to instantiate! Cooky, right?
- Null pointer-to-datamembers have a virtual base table offset of -1,
not zero. Previously, we chose an offset of 0.
NAKAMURA Takumi [Thu, 6 Feb 2014 07:15:59 +0000 (07:15 +0000)]
check-clang: Introduce the feature "utf8-capable-terminal".
clang/test/FixIt/fixit-unicode-with-utf8-output.c has begun complained since LLVM r200885.
Although it is changes for StringRef, it brought LLVM_ON_WIN32 to Support/Locale.cpp.
Before r200885, LLVM_ON_WIN32 was undefined in Locale.cpp!
Serge Pavlov [Thu, 6 Feb 2014 03:49:11 +0000 (03:49 +0000)]
Allow transformation of VariableArray to ConstantArray.
In the following code:
struct A { static const int sz; };
template<class T> void f() { T arr[A::sz]; }
the array 'arr' is represented as a variable size array in the template.
If 'A::sz' gets value below in the translation unit, the array in
instantiation can turn into constant size array.
Richard Smith [Thu, 6 Feb 2014 01:31:33 +0000 (01:31 +0000)]
DR101, PR12770: If a function is declared in the same context as a
using-declaration, and they declare the same function (either because
the using-declaration is in the same namespace as the declaration it
imports, or because they're both extern "C"), they do not conflict.
Manman Ren [Wed, 5 Feb 2014 20:40:15 +0000 (20:40 +0000)]
PGO: instrumentation based profiling sets function attributes.
We collect a maximal function count among all functions in the pgo data file.
For functions that are hot, we set its InlineHint attribute. For functions that
are cold, we set its Cold attribute.
We currently treat functions with >= 30% of the maximal function count as hot
and functions with <= 1% of the maximal function count are treated as cold.
These two numbers are from preliminary tuning on SPEC.
This commit should not affect non-PGO builds and should boost performance on
instrumentation based PGO.
projects / clang / log