Eric Fiselier [Mon, 6 Mar 2017 23:38:15 +0000 (23:38 +0000)]
[coroutines] Add DependentCoawaitExpr and fix re-building CoroutineBodyStmt.
Summary:
The changes contained in this patch are:
1. Defines a new AST node `CoawaitDependentExpr` for representing co_await expressions while the promise type is still dependent.
2. Correctly detect and transform the 'co_await' operand to `p.await_transform(<expr>)` when possible.
3. Change the initial/final suspend points to build during the initial parse, so they have the correct operator co_await lookup results.
4. Fix transformation of the CoroutineBodyStmt so that it doesn't re-build the final/initial suspends.
@rsmith: This change is a little big, but it's not trivial for me to split it up. Please let me know if you would prefer this submitted as multiple patches.
Eric Fiselier [Mon, 6 Mar 2017 22:52:28 +0000 (22:52 +0000)]
[coroutines] Improve diagnostics when building implicit constructs.
Previously when a coroutine was building the implicit setup/destroy
constructs it would emit diagostics about failures on the first co_await/co_return/co_yield
it encountered. This was confusing because that construct may not itself be ill-formed.
This patch moves the diagnostics to the function start instead.
Reid Kleckner [Mon, 6 Mar 2017 22:18:34 +0000 (22:18 +0000)]
Don't assume cleanup emission preserves dominance in expr evaluation
Summary:
Because of the existence branches out of GNU statement expressions, it
is possible that emitting cleanups for a full expression may cause the
new insertion point to not be dominated by the result of the inner
expression. Consider this example:
struct Foo { Foo(); ~Foo(); int x; };
int g(Foo, int);
int f(bool cond) {
int n = g(Foo(), ({ if (cond) return 0; 42; }));
return n;
}
Before this change, result of the call to 'g' did not dominate its use
in the store to 'n'. The early return exit from the statement expression
branches to a shared cleanup block, which ends in a switch between the
fallthrough destination (the assignment to 'n') or the function exit
block.
This change solves the problem by spilling and reloading expression
evaluation results when any of the active cleanups have branches.
I audited the other call sites of enterFullExpression, and they don't
appear to keep and Values live across the site of the cleanup, except in
ARC code. I wasn't able to create a test case for ARC that exhibits this
problem, though.
Gor Nishanov [Mon, 6 Mar 2017 21:12:54 +0000 (21:12 +0000)]
[coroutines] Add co_return statement emission
Summary:
Added co_return statement emission.
Tweaked coro-alloc.cpp test to use co_return to trigger coroutine processing instead of co_await, since this change starts emitting the body of the coroutine and await expression handling has not been upstreamed yet.
Vassil Vassilev [Mon, 6 Mar 2017 16:54:40 +0000 (16:54 +0000)]
[modules] Make GNUMode a normal language option to fix module compilation.
GNUMode shouldn't be a benign language option because it influences the
resulting AST when checking for the existence of GNUMode-specific macro
"linux" (e.g. by having code inside #ifdef linux). This patch marks it as a
normal language option so it gets correctly passed to the compiler invocation
for the used modules.
The added test case illustrated this because it compiles without modules, but
fails when using modules.
[clang-format] Make NamespaceEndCommentFixer add at most one comment
Summary:
Until now, NamespaceEndCommentFixer was adding missing comments for every run,
which results in multiple end comments for:
```
namespace {
int i;
int j;
}
#if A
int a = 1;
#else
int a = 2;
#endif
```
result before:
```
namespace {
int i;
int j;
}// namespace // namespace
#if A
int a = 1;
#else
int a = 2;
#endif
```
result after:
```
namespace {
int i;
int j;
}// namespace
#if A
int a = 1;
#else
int a = 2;
#endif
```
Alex Lorenz [Mon, 6 Mar 2017 15:58:34 +0000 (15:58 +0000)]
[Sema][ObjC] Warn about 'performSelector' calls with selectors
that return record or vector types
The performSelector family of methods from Foundation use objc_msgSend to
dispatch the selector invocations to objects. However, method calls to methods
that return record types might have to use the objc_msgSend_stret as the return
value won't find into the register. This is also supported by this sentence from
performSelector documentation: "The method should not have a significant return
value and should take a single argument of type id, or no arguments". This
commit adds a new warning that warns when a selector which corresponds to a
method that returns a record type is passed into performSelector.
[XRay] [clang] Allow logging the first argument of a function call.
Summary:
Functions with the "xray_log_args" attribute will tell LLVM to emit a special
XRay sled for compiler-rt to copy any call arguments to your logging handler.
Vedant Kumar [Mon, 6 Mar 2017 05:28:22 +0000 (05:28 +0000)]
[ubsan] Extend the nonnull arg check to ObjC
UBSan's nonnull argument check applies when a parameter has the
"nonnull" attribute. The check currently works for FunctionDecls, but
not for ObjCMethodDecls. This patch extends the check to work for ObjC.
[index] C++: Improve handling of typedefs as base names in C++ class declarations
Report the typedef as reference, and desugar it to report the underlying class as an
implicit 'base' reference.
Reporting the underlying base class for 'base' relations matches the ObjC handling and
leads to a simpler model.
Richard Trieu [Sat, 4 Mar 2017 02:42:41 +0000 (02:42 +0000)]
Handle null QualType better in Stmt::Profile
If the QualType is null, calling ASTContext::getCanonicalType on it will lead
to an assert. This was found while testing a new use for Stmt::Profile, so
there is no test case for this.
Richard Trieu [Sat, 4 Mar 2017 00:08:58 +0000 (00:08 +0000)]
[ODRHash] Add support for detecting different method properties.
Now print diagnostics for static, virtual, inline, volatile, and const
differences in methods. Also use DeclarationName instead of IdentifierInfo
for additional robustness in diagnostic printing.
This change adds an arch-specific subdirectory in <ResourceDir>/lib/<OS>
to the linker search path. This path also gets added as '-rpath' for
native compilation if a runtime is linked in as a shared object. This
allows arch-specific libraries to be installed alongside clang.
Dominic Chen [Fri, 3 Mar 2017 18:02:02 +0000 (18:02 +0000)]
Reland 4: [analyzer] NFC: Update test infrastructure to support multiple constraint managers
Summary: Replace calls to %clang/%clang_cc1 with %clang_analyze_cc1 when invoking static analyzer, and perform runtime substitution to select the appropriate constraint manager, per D28952.
Aleksei Sidorin [Fri, 3 Mar 2017 16:58:53 +0000 (16:58 +0000)]
[Analyzer] Terminate analysis on OpenMP code instead of assertion crash
* ExprEngine assumes that OpenMP statements should never appear in CFG.
However, current CFG doesn't know anything about OpenMP and passes
such statements as CFG nodes causing "UNREACHABLE executed!" crashes.
Since there is no OpenMP implementation in ExprEngine or CFG,
we stop the analysis on OpenMP statements to avoid crashes.
[PowerPC] Enable -fomit-frame-pointer by default for PPC
As is the case on platforms like Mips, X86 and SystemZ, the -fomit-frame-pointer
should be enabled by default on PPC when optimizing at -O1 and above. This
brings the behaviour of LLVM on PPC in line with GCC.
Dave Bartolomeo [Fri, 3 Mar 2017 00:08:55 +0000 (00:08 +0000)]
Fix msc-version.c test to handle _MSC_VER=1910
Previously, VC++ has always set _MSC_VER to a four-digit value with the two least significant digits set to zero. Visual Studio 2017, however, sets _MSC_VER=1910, and we expect to update the least significant digit as we release major updates for VS 2017. This patch fixes the msc-version.c test to handle non-zero values in the two least significant digits of _MSC_VER.
Dominic Chen [Thu, 2 Mar 2017 23:05:45 +0000 (23:05 +0000)]
Reland 3: [analyzer] NFC: Update test infrastructure to support multiple constraint managers
Summary: Replace calls to %clang/%clang_cc1 with %clang_analyze_cc1 when invoking static analyzer, and perform runtime substitution to select the appropriate constraint manager, per D28952.
Dominic Chen [Thu, 2 Mar 2017 22:45:24 +0000 (22:45 +0000)]
Reland 2: [analyzer] NFC: Update test infrastructure to support multiple constraint managers
Summary: Replace calls to %clang/%clang_cc1 with %clang_analyze_cc1 when invoking static analyzer, and perform runtime substitution to select the appropriate constraint manager, per D28952.
Serialization: use the PCH chain to check PCH mode
When we are deciding whether we are creating a PCH or a module, we would
check if the ModuleMgr had any elements to switch into PCH mode.
However, when creating a module, the size may be 1. This would result
in us going down the wrong path.
This was found by cross-compiling the swift standard library. Use the
PCH chain length instead to identify the PCH mode.
Unfortunately, I have not yet been able to create a simple test case for
this, but have verified that this fixes the swift standard library
construction.
Thanks to Adrian Prantl for help and discussions with this change!
[clang-format] Use number of unwrapped lines for short namespace
Summary:
This patch makes the namespace comment fixer use the number of unwrapped lines
that a namespace spans to detect it that namespace is short, thus not needing
end comments to be added.
This is needed to ensure clang-format is idempotent. Previously, a short namespace
was detected by the original source code lines. This has the effect of requiring two
runs for this example:
```
namespace { class A; }
```
after first run:
```
namespace {
class A;
}
```
after second run:
```
namespace {
class A;
} // namespace
```
Many things were wrong:
- We didn't always allow wrapping after "as", which can be necessary.
- We used to Undestand the identifier after "as" as a start of a name.
- We didn't properly parse the structure of the expression with "as"
having the precedence of relational operators
While parsing the DECL_VAR for 'struct FBox', clang recursively read all the
dep decls until it finds the DECL_CXX_RECORD forward declaration for 'struct
FVector'. Then, it resumes all the way up back to DECL_VAR handling in
`ReadDeclRecord`, where it checks if `isConsumerInterestedIn` for the decl.
One of the condition for `isConsumerInterestedIn` to return false is if the
VarDecl is imported from a module `D->getImportedOwningModule()`, because it
will get emitted when we import the relevant module. However, before checking
if it comes from a module, clang checks if `Ctx.DeclMustBeEmitted(D)`, which
triggers the emission of 'struct FBox'. Since one of its fields is still
incomplete, it crashes.
Instead, check if `D->getImportedOwningModule()` is true before calling
`Ctx.DeclMustBeEmitted(D)`.
Alex Lorenz [Wed, 1 Mar 2017 18:06:25 +0000 (18:06 +0000)]
Introduce an 'external_source_symbol' attribute that describes the origin
and the nature of a declaration
This commit adds an external_source_symbol attribute to Clang. This attribute
specifies that a declaration originates from an external source and describes
the nature of that source. This attribute will be used to improve IDE features
like 'jump-to-definition' for mixed-language projects or project that use
auto-generated code.
Devin Coughlin [Wed, 1 Mar 2017 17:48:39 +0000 (17:48 +0000)]
[analyzer] pr32088: Don't destroy the temporary if its initializer causes return.
In the following code involving GNU statement-expression extension:
struct S {
~S();
};
void foo() {
const S &x = ({ return; S(); });
}
function 'foo()' returns before reference x is initialized. We shouldn't call
the destructor for the temporary object lifetime-extended by 'x' in this case,
because the object never gets constructed in the first place.
The real problem is probably in the CFG somewhere, so this is a quick-and-dirty
hotfix rather than the perfect solution.
Eric Liu [Wed, 1 Mar 2017 13:14:01 +0000 (13:14 +0000)]
Introducing clang::tooling::AtomicChange for refactoring tools.
Summary:
An AtomicChange is used to create and group a set of source edits, e.g.
replacements or header insertions. Edits in an AtomicChange should be related,
e.g. replacements for the same type reference and the corresponding header
insertion/deletion.
An AtomicChange is uniquely identified by a key position and will either be
fully applied or not applied at all. The key position should be the location
of the key syntactical element that is being changed, e.g. the call to a
refactored method.
Daniel Jasper [Wed, 1 Mar 2017 10:47:52 +0000 (10:47 +0000)]
clang-format: Ignore contents of #ifdef SWIG .. #endif blocks.
Those blocks are used if C++ code is SWIG-wrapped (see swig.org) and
usually do not contain C++ code. Also cleanup the implementation of for #if 0
and #if false a bit.
Akira Hatanaka [Wed, 1 Mar 2017 06:11:25 +0000 (06:11 +0000)]
[Sema] Add variable captured by a block to the enclosing lambda's
potential capture list.
Fix Sema::getCurLambda() to return the innermost lambda scope when there
is a block enclosed in the lambda. Previously, the method would return a
nullptr in such cases, which would prevent a variable captured by the
enclosed block to be added to the lambda scope's potential capture list.
Richard Trieu [Wed, 1 Mar 2017 03:07:55 +0000 (03:07 +0000)]
Add warning for inconsistent overrides on destructor.
The exisiting warning for inconsistent overrides does not include the destructor
as it was noted in review that it was too noisy. Instead, add to a separate
warning group that is off by default for users who want consistent warnings
between methods and destructors.
Paul Robinson [Wed, 1 Mar 2017 01:01:10 +0000 (01:01 +0000)]
[PS4] Set our default dialect to C++11. NFC for other targets.
Reapplies r296209 now that r296549 has fixed what really seems to be
the last problematic test.
Christof Douma [Tue, 28 Feb 2017 09:09:53 +0000 (09:09 +0000)]
[ARM] Don't pass -arm-execute-only to cc1as
The option -mexecute-only is translated into the backend option
-arm-execute-only. But this option only makes sense for the compiler and
the assembler does not recognize it. This patch stops clang from passing
this option to the assembler.
Dominic Chen [Tue, 28 Feb 2017 00:02:36 +0000 (00:02 +0000)]
Reland: [analyzer] NFC: Update test infrastructure to support multiple constraint managers
Summary: Replace calls to %clang/%clang_cc1 with %clang_analyze_cc1 when invoking static analyzer, and perform runtime substitution to select the appropriate constraint manager, per D28952.
Nico Weber [Mon, 27 Feb 2017 22:59:58 +0000 (22:59 +0000)]
clang-format: Don't leave behind temp files in -i mode on Windows, PR26125, reloaded
Second attempt after http://llvm.org/viewvc/llvm-project?rev=296166&view=rev
In the first attempt, Code (the memory buffer backing the input file) was reset
before overwriteChangedFiles() was called, but overwriteChangedFiles() still
reads from it. This time, load the whole input file into memory instead of
using mmap when formatting in-place.
(Since the test is identical to what was in the repo before chapuni's revert,
svn diff doesn't show it – see the above link for the test.)
Adrian Prantl [Mon, 27 Feb 2017 21:30:05 +0000 (21:30 +0000)]
PR32042: Create inlined debug info for EmitInlinedInheritingCXXConstructorCall.
When clang emits an inheriting C++ constructor it may inline code
during the CodeGen phase. This patch ensures that any debug info in
this inlined code gets a proper inlined location. Otherwise we can end
up with invalid debug info metadata, since all inlined local variables
and function arguments would be reparented into the call site.
Analogous to ApplyInlineLocation this patch introduces a
ApplyInlineDebugLocation scoped helper to facilitate entering an
inlined scope and cleaning up afterwards.
This fixes one of the issues discovered in PR32042.
Summary:
This patch adds a NamespaceEndCommentsFixer TokenAnalyzer for clang-format,
which fixes end namespace comments.
It currently supports inserting and updating existing wrong comments.
Example source:
```
namespace A {
int i;
}
namespace B {
int j;
} // namespace A
```
after formatting:
```
namespace A {
int i;
} // namespace A
Dominic Chen [Mon, 27 Feb 2017 02:36:15 +0000 (02:36 +0000)]
[analyzer] NFC: Update test infrastructure to support multiple constraint managers
Summary: Replace calls to %clang/%clang_cc1 with %clang_analyze_cc1 when invoking static analyzer, and perform runtime substitution to select the appropriate constraint manager, per D28952.
Oren Ben Simhon [Sun, 26 Feb 2017 11:58:15 +0000 (11:58 +0000)]
[X86] DAZ Macros Relocation
The DAZ feature introduces the denormal zero support for x86.
Currently the definitions are located under SSE3 header, however there are some SSE2 targets that support the feature as well.
[index] Add 'Parameter' symbol kind and 'Local' symbol property to distinguish function-local symbols
Parameters have a 'child' relation to their function/method.
Also add an option '-include-locals' to 'c-index-test core' to enable indexing of function-local symbols.
Original patch from Nathan Hawes with some changes by me.
https://reviews.llvm.org/D30304
Richard Smith [Sat, 25 Feb 2017 23:53:05 +0000 (23:53 +0000)]
C++ DR1611, 1658, 2180: implement "potentially constructed subobject" rules for special member functions.
Essentially, as a base class constructor does not construct virtual bases, such
a constructor for an abstract class does not need the corresponding base class
construction to be valid, and likewise for destructors.
This creates an awkward situation: clang will sometimes generate references to
the complete object and deleting destructors for an abstract class (it puts
them in the construction vtable for a derived class). But we can't generate a
"correct" version of these because we can't generate references to base class
constructors any more (if they're template specializations, say, we might not
have instantiated them and can't assume any other TU will emit a copy).
Fortunately, we don't need to, since no correct program can ever invoke them,
so instead emit symbols that just trap.
We should stop emitting references to these symbols, but still need to emit
definitions for compatibility.
[driver] Pass a resource dir without the '/../' part.
This get the resource dir string to match with the one from libclang (which is not adding '/../'),
and allows clang to accept a modules-enabled PCH that was created by libclang.
projects / clang / log