Inline asm IR input register constraints don't have early clobber
modifiers on them. If we have a matching output constraint with
an early clobber make sure we don't propagate that to the input
constraint.
Steven Wu [Fri, 6 Feb 2015 18:08:29 +0000 (18:08 +0000)]
-iframework option should be forwarded to linker
Summary:
-iframework option is used to specified System framework path so the
path specified should be passed to linker as -F option
rdar://problem/18234544
Ben Langmuir [Fri, 6 Feb 2015 17:34:28 +0000 (17:34 +0000)]
Make FileEntry::closeFile public
If you request that the file manager not close your file immediately
after reading, it's useful to be able to close it later to prevent a
file descriptor leak.
OpenCL: handle shift operator with vector operands
Introduce a number of checks:
1. If LHS is a scalar, then RHS cannot be a vector.
2. Operands must be of integer type.
3. If both are vectors, then the number of elements must match.
Relax the requirement for "usual arithmetic conversions":
When LHS is a vector, a scalar RHS can simply be expanded into a
vector; OpenCL does not require that its rank be lower than the LHS.
For example, the following code is not an error even if the implicit
type of the constant literal is "int".
char2 foo(char2 v) { return v << 1; }
Consolidate existing tests under CodeGenOpenCL, and add more tests
under SemaOpenCL.
Richard Smith [Fri, 6 Feb 2015 02:42:59 +0000 (02:42 +0000)]
[modules] If a module declares an entity and then imports another declaration
of that entity, ensure that the redeclaration chain is reordered properly on
reload. Otherwise, the result of name lookup for that entity may point to an
entity that is too old; if that's an injected friend name or the like, that
can result in the name not being found at all.
Tim Northover [Fri, 6 Feb 2015 01:25:07 +0000 (01:25 +0000)]
Preprocessor: support __BIGGEST_ALIGNMENT__ macro
For compatibility with GCC (and because it's generally helpful information
otherwise inaccessible to the preprocessor). This appears to be canonically the
alignment of max_align_t (e.g. on i386, __BIGGEST_ALIGNMENT__ is 4 even though
vector types will be given greater alignment).
Force an unknown target for the default case to make it work.
Summary:
This test was failing if LLVM_DEFAULT_TARGET_TRIPLE=x86_64-scei-ps4 due to
its behavior for exceptions.
Force an --unknown triple for the default case.
Reid Kleckner [Thu, 5 Feb 2015 18:56:03 +0000 (18:56 +0000)]
Re-land r228258 and make clang-cl's /EHs- disable -fexceptions again
After r228258, Clang started emitting C++ EH IR that LLVM wasn't ready
to deal with, even when exceptions were disabled with /EHs-. This time,
make /EHs- turn off -fexceptions while still emitting exceptional
constructs in functions using __try. Since Sema rejects C++ exception
handling constructs before CodeGen, landingpads should only appear in
such functions as the result of a __try.
Richard Smith [Thu, 5 Feb 2015 07:02:11 +0000 (07:02 +0000)]
PR22465: when performing list-initialization for a class type C, if we see an
initializer of the form {x}, where x is of type C or a type derived from C,
perform *non-list* initialization of the entity from x, but create a
CXXConstructExpr that knows that we used list-initialization syntax.
Plus some fixes to ensure we mangle correctly in this and related cases.
Alexey Bataev [Thu, 5 Feb 2015 05:57:51 +0000 (05:57 +0000)]
[OPENMP] Codegen for 'taskyield' directive
For 'taskyield' directive emit call to kmp_int32 __kmpc_omp_taskyield(ident_t *,
kmp_int32 global_tid, int end_part); runtime function call with end_part arg set
to 0 (it is ignored).
Differential Revision: http://reviews.llvm.org/D7047
Nico Weber [Thu, 5 Feb 2015 02:08:50 +0000 (02:08 +0000)]
Revert r228258.
It caused a chromium base unittest that tests throwing and catching SEH
exceptions to fail (http://crbug.com/455488) and I suspect it might also
be the cause of the chromium clang win 64-bit shared release builder timing
out during compiles. So revert to see if that's true.
Bill Schmidt [Thu, 5 Feb 2015 01:12:24 +0000 (01:12 +0000)]
[PowerPC] Revert workaround for TLS linker bug
In r227480, Ulrich Weigand introduced a workaround for a linker
optimization bug that can create mis-optimized code for accesses to
general-dynamic or local-dynamic TLS variables. The linker
optimization bug only occurred for Clang/LLVM because of some
inefficient code being generated for these TLS accesses. I have
recently corrected LLVM to produce the efficient code sequence
expected by the linkers, so this workaround is no longer needed.
Therefore this patch reverts r227480.
I've tested that the previous bootstrap failure no longer occurs with
the workaround reverted.
Anna Zaks [Thu, 5 Feb 2015 01:02:59 +0000 (01:02 +0000)]
[analyzer] Relax an assertion in VisitLvalArraySubscriptExpr
The analyzer thinks that ArraySubscriptExpr cannot be an r-value (ever).
However, it can be in some corner cases. Specifically, C forbids expressions
of unqualified void type from being l-values.
Note, the analyzer will keep modeling the subscript expr as an l-value. The
analyzer should be treating void* as a char array
(https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gcc/Pointer-Arith.html).
Anna Zaks [Thu, 5 Feb 2015 01:02:53 +0000 (01:02 +0000)]
[analyzer] Look for allocation site in the parent frames as well as the current one.
Instead of handling edge cases (mostly involving blocks), where we have difficulty finding
an allocation statement, allow the allocation site to be in a parent node.
Previously we assumed that the allocation site can always be found in the same frame
as allocation, but there are scenarios in which an element is leaked in a child
frame but is allocated in the parent.
Reid Kleckner [Wed, 4 Feb 2015 23:45:07 +0000 (23:45 +0000)]
Driver: Stop forcing frame pointer usage on Windows
Previously, we would use a frame pointer by default on non-Linux OSs. On
Linux, any optimization flags imply -fomit-frame-pointer. XCore always
defaulted to -fomit-frame-pointer.
Now x86 Windows matches our behavior on Linux. All other ISAs supported
by Windows (ARM, x64) use xdata information, and frame pointers aren't
useful. Frame pointers are now off by default for such targets, but can
be forced via -fno-omit-frame-pointer and code using alloca().
In fact, on Win64 our frame-pointer prologue is not describable with
UNWIND_INFO. This change is a workaround to avoid using the broken FP
using prologue for most functions. This is PR22467.
Richard Smith [Wed, 4 Feb 2015 23:37:59 +0000 (23:37 +0000)]
[modules] When using -E, we may try to merge decls despite having no Sema
object. In such a case, use the TU's DC for merging global decls rather than
giving up when we find there is no TU scope.
Ultimately, we should probably avoid all loading of decls when preprocessing,
but there are other reasonable use cases for loading an AST file with no Sema
object for which this is the right thing.
Reid Kleckner [Wed, 4 Feb 2015 22:37:07 +0000 (22:37 +0000)]
Implement IRGen for SEH __finally and AbnormalTermination
Previously we would simply double-emit the body of the __finally block,
but that doesn't work when it contains any kind of Decl, which we can't
double emit.
This fixes that by emitting the block once and branching into a shared
code region and then branching back out.
Thread Safety Analysis: support adopting of locks, as implemented in
std::lock_guard. If EXCLUSIVE_LOCKS_REQUIRED is placed on the constructor of
a SCOPED_LOCKABLE class, then that constructor is assumed to adopt the lock;
e.g. the lock must be held on construction, and will be released on destruction.
David Blaikie [Wed, 4 Feb 2015 19:47:54 +0000 (19:47 +0000)]
DebugInfo: Attribute cleanup code to the end of the scope, not the end of the function.
Now if you break on a dtor and go 'up' in your debugger (or you get an
asan failure in a dtor) during an exception unwind, you'll have more
context. Instead of all dtors appearing to be called from the '}' of the
function, they'll be attributed to the end of the scope of the variable,
the same as the non-exceptional dtor call.
This doesn't /quite/ remove all uses of CurEHLocation (which might be
nice to remove, for a few reasons) - it's still used to choose the
location for some other work in the landing pad. It'd be nice to
attribute that code to the same location as the exception calls within
the block and to remove CurEHLocation.
Jordan Rose [Wed, 4 Feb 2015 19:24:52 +0000 (19:24 +0000)]
[analyzer] RetainCountChecker: be forgiving when ivars are accessed directly.
A refinement of r204730, itself a refinement of r198953, to better handle
cases where an object is accessed both through a property getter and
through direct ivar access. An object accessed through a property should
always be treated as +0, i.e. not owned by the caller. However, an object
accessed through an ivar may be at +0 or at +1, depending on whether the
ivar is a strong reference. Outside of ARC, we don't always have that
information.
The previous attempt would clear out the +0 provided by a getter, but only
if that +0 hadn't already participated in other retain counting operations.
(That is, "self.foo" is okay, but "[[self.foo retain] autorelease]" is
problematic.) This turned out to not be good enough when our synthesized
getters get involved.
This commit drops the notion of "overridable" reference counting and instead
just tracks whether a value ever came from a (strong) ivar. If it has, we
allow one more release than we otherwise would. This has the added benefit
of being able to catch /some/ overreleases of instance variables, though
it's not likely to come up in practice.
We do still get some false negatives because we currently throw away
refcount state upon assigning a value into an ivar. We should probably
improve on that in the future, especially once we synthesize setters as
well as getters.
Alexey Samsonov [Wed, 4 Feb 2015 17:40:08 +0000 (17:40 +0000)]
Allow to specify multiple -fsanitize-blacklist= arguments.
Summary:
Allow user to provide multiple blacklists by passing several
-fsanitize-blacklist= options. These options now don't override
default blacklist from Clang resource directory, which is always
applied (which fixes PR22431).
-fno-sanitize-blacklist option now disables all blacklists that
were specified earlier in the command line (including the default
one).
This change depends on http://reviews.llvm.org/D7367.
Nico Weber [Wed, 4 Feb 2015 15:26:27 +0000 (15:26 +0000)]
clang-format: Add support for SEH __try / __except / __finally blocks.
This lets clang-format format
__try {
} __except(0) {
}
and
__try {
} __finally {
}
correctly. __try and __finally are keywords if `LangOpts.MicrosoftExt` is set,
so this turns this on. This also enables a few other keywords, but it
shouldn't overly perturb regular clang-format operation. __except is a
context-sensitive keyword, so `AdditionalKeywords` needs to be passed around to
a few more places.
Daniel Sanders [Wed, 4 Feb 2015 14:25:47 +0000 (14:25 +0000)]
Preserve early clobber flag when using named registers in inline assembly.
Summary:
Named registers with the constraint "=&r" currently lose the early clobber flag
and turn into "=r" when converted to LLVM-IR. This patch correctly passes it on.
OpenCL: handle ternary operator when the condition is a vector
When the condition is a vector, OpenCL specifies additional
requirements on the operand types, and also the operations
required to determine the result type of the operator. This is a
combination of OpenCL v1.1 s6.3.i and s6.11.6, and the semantics
remain unchanged in later versions of OpenCL.
Alexey Bataev [Wed, 4 Feb 2015 04:45:32 +0000 (04:45 +0000)]
[MSVC2012] Allow 'mutable' references
Some standard header files from MSVC2012 use 'mutable' on references, though it is directly prohibited by the standard.
Fix for http://llvm.org/PR22444
Differential Revision: http://reviews.llvm.org/D7370
Richard Smith [Wed, 4 Feb 2015 01:23:46 +0000 (01:23 +0000)]
[modules] If we already have a fake definition for one declaration of a class,
don't think we're providing a new one in an update record adding a definition
to another declaration of the same class.
Akira Hatanaka [Wed, 4 Feb 2015 00:27:13 +0000 (00:27 +0000)]
[inlineasm] Fix an incorrect warning about register constraint and modifier.
Previously, when the following piece of code was compiled, clang would
incorrectly warn that the size of "wide_two" does not match register size
specified by the constraint and modifier":
David Majnemer [Tue, 3 Feb 2015 23:04:06 +0000 (23:04 +0000)]
CodeGen: Copy-ctorm must obey the destination's alignment requirement
We would synthesize memcpy intrinsics when emitting calls to trivial C++
constructors but we wouldn't take into account the alignment of the
destination.
Reid Kleckner [Tue, 3 Feb 2015 22:52:35 +0000 (22:52 +0000)]
SEH: Track users of __try so we can pick a per-func EH personality
There are four major kinds of declarations that cause code generation:
- FunctionDecl (includes CXXMethodDecl etc)
- ObjCMethodDecl
- BlockDecl
- CapturedDecl
This patch tracks __try usage on FunctionDecls and diagnoses __try usage
in other decls. If someone wants to use __try from ObjC, they can use it
from a free function, since the ObjC code will need an ObjC-style EH
personality.
Eventually we will want to look through CapturedDecls and track SEH
usage on the parent FunctionDecl, if present.
David Blaikie [Tue, 3 Feb 2015 22:37:17 +0000 (22:37 +0000)]
DebugInfo: Ensure calls to functions with default arguments which themselves have default arguments, still have locations.
To handle default arguments in C++ in the debug info, we disable code
updating the debug location during the emission of default arguments.
This code was buggy in the case of default arguments which, themselves,
have default arguments - the inner default argument would re-enable
debug info when it was finished, but before the outer default argument
was finished.
This was already a bug, but got worse (because a crasher instead of just
a quality bug) with the recent improvements to debug info line quality
because... The ApplyDebugLocation scoped device would find the debug
info disabled and not save any debug location. But then in
~ApplyDebugLocation it would find the debug info had been enabled and
would then apply the no-location. Then the outer function call would be
emitted without any location. That's bad.
Arguably we could /also/ fix the ApplyDebugLocation to assert on this
situation (where debug info was disabled in the ctor and enabled in the
dtor, or the other way around) but this is at least the necessary fix
regardless.
(also, I imagine this disabling behavior might need to be in-place for
CGExprComplex and CGExprAgg too, maybe... ?)
And I seem to recall seeing some weird default arg stepping behavior
recently which might be related to this too... I'll have to look into
it.
Thread Safety Analysis: add support for before/after annotations on mutexes.
These checks detect potential deadlocks caused by inconsistent lock
ordering. The checks are implemented under the -Wthread-safety-beta flag.
This patch also replaces calls to getAttrs() with calls to attrs() throughout
ThreadSafety.cpp, which fixes the earlier issue that cause assert failures.
Ben Langmuir [Tue, 3 Feb 2015 19:28:37 +0000 (19:28 +0000)]
Make the default module cache user-specific
Appends the username to the first component (after the temp dir) of the
module cache path. If the username contains a character that shouldn't
go into a path (for now conservatively allow [a-zA-Z0-9_]), we fallback
to the user id.
Adrian Prantl [Tue, 3 Feb 2015 18:40:42 +0000 (18:40 +0000)]
Merge ArtificialLocation into ApplyDebugLocation and make a clear
distinction between the different use-cases. With the previous default
behavior we would occasionally emit empty debug locations in situations
where they actually were strictly required (= on invoke insns).
We now have a choice between defaulting to an empty location or an
artificial location.
Specifically, this fixes a bug caused by a missing debug location when
emitting C++ EH cleanup blocks from within an artificial function, such as
an ObjC destroy helper function.
Thread Safety Analysis: add support for before/after annotations on mutexes.
These checks detect potential deadlocks caused by inconsistent lock
ordering. The checks are implemented under the -Wthread-safety-beta flag.
Nico Weber [Tue, 3 Feb 2015 17:06:08 +0000 (17:06 +0000)]
Implement jump scope SEHmantic analysis.
Thou shall not jump into SEH blocks. Jumping out of SEH __try and __excepts
is A-ok. Jumping out of __finally blocks is B-ok (msvc doesn't error about it,
but warns that it has undefined behavior).
I've checked that clang's behavior with this patch matches msvc's behavior.
We don't have the warning on jumping out of a __finally yet, see the FIXME
in the test. clang also currently crashes on codegen for a jump out of a
__finally block, see PR22414 comment 7.
I also added a few tests for the interaction of indirect jumps and SEH blocks.
MSVC doesn't support indirect jumps, so there's no way to know if clang behave
the same way as msvc here. clang's behavior with this patch does make sense
to me, but maybe it could be argued that it should be more permissive (see
FIXME in the indirect jump tests -- shout if you have an opinion on this).
Daniel Jasper [Tue, 3 Feb 2015 09:45:52 +0000 (09:45 +0000)]
Add some overloads so that floating point literals can be AST matched properly.
I am not entirely sure whether the implemented sematics are ideal. In
particular, should floatLiteral(equals(0.5)) match "0.5f" and should
floatLiteral(equals(0.5f)) match "0.5". With the overloads in this
patch, the answer to both questions is yes, but I am happy to change
that.