Daniel Sanders [Fri, 1 Aug 2014 10:29:21 +0000 (10:29 +0000)]
[mips] Defer va_arg expansion to the backend.
Summary:
This patch causes clang to emit va_arg instructions to the backend instead of
expanding them into an implementation itself. The backend already implements
va_arg since this is necessary for NaCl so this patch is removing redundant
code.
Together with the llvm patch (D4556) that accounts for the effect of endianness
on the expansion of va_arg, this fixes PR19612.
Richard Smith [Fri, 1 Aug 2014 01:56:39 +0000 (01:56 +0000)]
[modules] Remove IRGen special case for emitting implicit special members if
they're somehow missing a body. Looks like this was left behind when the loop
was generalized, and it's not been problematic before because without modules,
a used, implicit special member function declaration must be a definition.
This was resulting in us trying to emit a constructor declaration rather than
a definition, and producing a constructor missing its member initializers.
Richard Trieu [Fri, 1 Aug 2014 01:42:01 +0000 (01:42 +0000)]
Remove this pointer that is converted to bool. In well-defined contexts, the
this pointer is always non-null. If the this pointer is null, it is undefined
and the compiler may optimize it away by assuming it is non-null. The null
checks are pushed into the callers.
Richard Smith [Thu, 31 Jul 2014 23:46:44 +0000 (23:46 +0000)]
[modules] Maintain an AST invariant across module load/save: if any declaration
of a function has a resolved exception specification, then all declarations of
the function do.
We should probably improve the AST representation to make this implicit (perhaps
only store the exception specification on the canonical declaration), but this
fixes things for now.
The testcase for this (which used to assert) also exposes the actual bug I was
trying to reduce here: we sometimes fail to emit the body of an imported
special member function definition. Fix for that to follow.
Richard Smith [Thu, 31 Jul 2014 21:57:55 +0000 (21:57 +0000)]
Factor out exception specification information from
FunctionProtoType::ExtProtoInfo. Most of the users of these fields don't care
about the other ExtProtoInfo bits and just want to talk about the exception
specification.
Loop hint pragmas sometimes do not contain an identifier option (such as #pragma unroll(4)). Check explicitly that the token we stored was an identifier.
This patch is necessary to support constant expressions which replaces the integer value in the loop hint attribute with an expression. The integer value was also storing the pragma’s state for options like vectorize(enable/disable) and the pragma unroll and nounroll directive. The state variable is introduced to hold the state of those options/pragmas. This moves the validation of the state (keywords) from SemaStmtAttr handler to the loop hint annotation token handler.
Resubmit with changes to try to fix the build-bot issue.
Eli Bendersky [Thu, 31 Jul 2014 18:04:56 +0000 (18:04 +0000)]
Exposes a C API to name mangling for a given cursor.
Inspired by https://gist.github.com/tritao/2766291, and was previously discussed
on cfe-dev: http://lists.cs.uiuc.edu/pipermail/cfe-dev/2014-June/037577.html
Obective-C. Patch to fix the incorrect ObjcMessageExpr argument source ranges,
when arguments are structures or classes. PR16392.
patch by Karlis Senko
Automate attribute argument count semantic checking when there are variadic or optional arguments present. With this, the only time you should have to manually check attribute argument counts is when HasCustomParsing is set to true, or when you have variadic arguments that aren't really variadic (like ownership_holds and friends).
Updating the diagnostics in the launch_bounds test since they have been improved in that case. Adding a test for nonnull since it has little test coverage, but has truly variadic arguments.
Richard Smith [Thu, 31 Jul 2014 06:31:19 +0000 (06:31 +0000)]
PR18097: Support initializing an _Atomic(T) from an object of C++ class type T
or a class derived from T. We already supported this when initializing
_Atomic(T) from T for most (and maybe all) other reasonable values of T.
Updating a comment related to the implementation of -Woverloaded-virtual, and adding a FIXME to a test case. (Drive-by removal of trailing whitespace in the test case as well.)
This patch is necessary to support constant expressions which replaces the integer value in the loop hint attribute with an expression. The integer value was also storing the pragma’s state for options like vectorize(enable/disable) and the pragma unroll and nounroll directive. The state variable is introduced to hold the state of those options/pragmas. This moves the validation of the state (keywords) from SemaStmtAttr handler to the loop hint annotation token handler.
As defined in the SPIR 1.2 specification, this node behaves similarly to
kernel_arg_type but will print the underlying type name, e.g., without
typedefs.
Example:
typedef unsigned int myunsignedint;
would report:
'myunsignedint' in the kernel_arg_type node
'uint' in the kernel_arg_base_type node
Robert Khasanov [Wed, 30 Jul 2014 13:53:40 +0000 (13:53 +0000)]
[SKX] Enabling SKX target (Skylake server chip)
a) add SKX support to Clang driver;
b) add tests for SKX target and AVX512BW, AVX512DQ, AVX512VL features into clang driver tests
David Majnemer [Wed, 30 Jul 2014 08:42:33 +0000 (08:42 +0000)]
AST: Simplify some code
Iterator invalidation issues already force us to do one lookup and one
insert.
Don't use the particular bit-pattern of the 'Align' field to determine
whether or not we have already inserted into the TypeInfo DenseMap;
instead ask for an iterator to the TypeInfo entry.
Manuel Klimek [Wed, 30 Jul 2014 08:34:42 +0000 (08:34 +0000)]
Fix some cases of incorrect handling of lifetime extended temporaries.
MaterializeTemporaryExpr already contains information about the lifetime
of the temporary; if the lifetime is not the full statement, we do not
want to emit a destructor at the end of the full statement for it.
David Majnemer [Wed, 30 Jul 2014 08:20:03 +0000 (08:20 +0000)]
MS ABI: Mangle alias templates used as template-template arguments
A templated using declaration may be used as a template-template
argument.
Unfortunately, the VS "14" chooses '?' as the sole marker for the
argument. This is problematic because it presupposes the possibility of
using more than one template-aliases as arguments to the same template.
Richard Smith [Tue, 29 Jul 2014 21:20:12 +0000 (21:20 +0000)]
PR20473: Don't "deduplicate" string literals with the same value but different
lengths! In passing, simplify string literal deduplication by relying on LLVM
to deduplicate the underlying constant values.
Bob Wilson [Tue, 29 Jul 2014 20:17:52 +0000 (20:17 +0000)]
Support LIBRARY_PATH on all Darwin targets.
r197490 changed the behavior of LIBRARY_PATH to try to match GCC's behavior
for cross compilers and make clang work better on "bare metal" targets.
Unfortunately that change is breaking a number of MacPorts projects because
the LIBRARY_PATH environment variable is being ignored when compiling on a
64-bit host for a 32-bit target. Because the host and target architectures
differ, isCrossCompiling returns true. This does not make sense for Darwin,
where multiple architectures are supported natively via "fat" Mach-O slices
and where development is generally done against SDKs regardless. This patch
fixes the problem by overriding isCrossCompiling to return false for Darwin
toolchains.
Fix PR10177 where non-type template arguments to alias templates are not marked as used in dependent contexts. The fix actually forces non-dependent names to be checked at template definition time as expected from the standard.
Richard Smith [Tue, 29 Jul 2014 00:58:01 +0000 (00:58 +0000)]
[modules] Add testcase for a bug reduced from a selfhost issue. This bug was
never present in Clang trunk, but was present in some of my development work,
and it seems like a useful test to have.
Bob Wilson [Tue, 29 Jul 2014 00:23:18 +0000 (00:23 +0000)]
Fix up handling of ARM options for controlling strict alignment.
The -mstrict-align option was originally added in r167619 as a target-
independent option. It was then changed in r167623 to be implemented with an
ARM-specific backend option, even though the code remained in the
target-independent Clang::ConstructJob function. This means that if you used
the -mstrict-align option with a non-ARM target, you would still get the
-arm-strict-align option getting passed to the backend, which was harmless
but gross. The driver option was then replaced by the GCC-compatible
-m[no-]unaligned-access option (r189175) and modified to work with AArch64
(r208075). However, in the process, the help text for -mstrict-align was
incorrectly changed to show it as only being supported for AArch64. Even worse,
the logic for handling these options together with -mkernel was wrong for
AArch64, where -mkernel does not currently imply strict alignment.
This patch fixes up all of those things. Besides the obvious change to the
option help text, it moves the logic into the ARM and AArch64-specific parts
of the driver, so that the option will be correctly ignored for non-ARM
targets. <rdar://problem/17823697>
This isn't nearly as elaborate as the GCC inline asm which emits an
array of source locations, but it's very, very hard to trigger backend
diagnostics in MS inline asm because we parse it up front with good
source information, unlike GCC inline asm.
Currently I can trigger a "inline assembly requires more registers than
available" diagnostic with this code:
Change __INTx_TYPE__ to be always signed. This changes the value for
char-based types from "char" to "signed char". Adjust stdint.h to use
__INTx_TYPE__ directly without prefixing it with signed and to use
__UINTx_TYPE__ for unsigned ones.
Add another keyword-selection flag to CorrectionCandidateCallback.
The new flag, WantFunctionLikeCasts, covers a subset of the keywords
covered by WantTypeSpecifiers that can be used in casts that look like
function calls, e.g. "return long(5);", while excluding the keywords
like "enum" and "const" that would be included when WantTypeSpecifiers
is true but cannot be used in something that looks like a function call.
Thread Safety Analysis: Replace the old and broken SExpr with the new
til::SExpr. This is a large patch, with many small changes to pretty printing
and expression lowering to make the new SExpr representation equivalent in
functionality to the old.
[PowerPC] Support ELFv1/ELFv2 ABI selection via -mabi= option
While Clang now supports both ELFv1 and ELFv2 ABIs, their use is currently
hard-coded via the target triple: powerpc64-linux is always ELFv1, while
powerpc64le-linux is always ELFv2.
These are of course the most common scenarios, but in principle it is
possible to support the ELFv2 ABI on big-endian or the ELFv1 ABI on
little-endian systems (and GCC does support that), and there are some
special use cases for that (e.g. certain Linux kernel versions could
only be built using ELFv1 on LE).
This patch implements the Clang side of supporting this, based on the
LLVM commit 214072. The command line options -mabi=elfv1 or -mabi=elfv2
select the desired ABI if present. (If not, Clang uses the same default
rules as now.)
Specifically, the patch implements the following changes based on the
presence of the -mabi= option:
In the driver:
- Pass the appropiate -target-abi flag to the back-end
- Select the correct dynamic loader version (/lib64/ld64.so.[12])
In the preprocessor:
- Define _CALL_ELF to the appropriate value (1 or 2)
In the compiler back-end:
- Select the correct ABI in TargetInfo.cpp
- Select the desired ABI for LLVM via feature (elfv1/elfv2)
Richard Smith [Sun, 27 Jul 2014 05:12:49 +0000 (05:12 +0000)]
When looking for temporary dtors while building the CFG, do not walk into
lambda expressions (other than their capture initializers) nor blocks. Do walk
into default argument expressions and default initializer expressions.
These bugs were causing us to produce broken CFGs whenever a lambda expression
was used to initialize a libstdc++ std::function object!
projects / clang / log