Sam McCall [Thu, 9 Nov 2017 10:37:39 +0000 (10:37 +0000)]
[Tooling] Use FixedCompilationDatabase when `compile_flags.txt` is found.
Summary:
This is an alternative to JSONCompilationDatabase for simple projects that
don't use a build system such as CMake.
(You can also drop one in ~, to make your tools use e.g. C++11 by default)
There's no facility for varying flags per-source-file or per-machine.
Possibly this could be accommodated backwards-compatibly using cpp, but even if
not the simplicity seems worthwhile for the cases that are addressed.
Tested with clangd, works great! (requires clangd restart)
Craig Topper [Thu, 9 Nov 2017 04:10:46 +0000 (04:10 +0000)]
[X86] Rename the VEX scalar fma builtins to end with a '3' to match gcc
I think we need to use different builtins for the FMA4 instructions since those instructions zero the upper bits and FMA3 instructions pass the bits through.
So this moves the existing builtins to be the FMA3 versions. New versions will be added for FMA4.
Vedant Kumar [Thu, 9 Nov 2017 02:33:40 +0000 (02:33 +0000)]
[Coverage] Emit deferred regions in headers
There are some limitations with emitting regions in macro expansions
because we don't gather file IDs within the expansions. Fix the check
that prevents us from emitting deferred regions in expansions to make an
exception for headers, which is something we can handle.
Vedant Kumar [Thu, 9 Nov 2017 02:33:38 +0000 (02:33 +0000)]
[Coverage] Emit a gap area after if conditions
The area immediately after the closing right-paren of an if condition
should have a count equal to the 'then' block's count. Use a gap region
to set this count, so that region highlighting for the 'then' block
remains precise.
This solves a problem we have with wrapped segments. Consider:
1| if (false)
2| foo();
Without a gap area starting after the condition, the wrapped segment
from line 1 would make it look like line 2 is executed, when it's not.
Alex Lorenz [Wed, 8 Nov 2017 22:44:34 +0000 (22:44 +0000)]
[ObjC] Fix function signature handling for blocks literals with attributes
Block literals can have a type with attributes in its signature, e.g.
ns_returns_retained. The code that inspected the type loc of the block when
declaring its parameters didn't account for this fact, and only looked through
paren type loc. This commit ensures that getAsAdjusted is used instead of
IgnoreParens to find the block's FunctionProtoTypeLoc. This ensures that
block parameters are declared correctly in the block and avoids the
'undeclared identifier' error.
Alex Lorenz [Wed, 8 Nov 2017 21:33:15 +0000 (21:33 +0000)]
[ObjC] Boxed strings should use the nullability from stringWithUTF8String's return type
Objective-C NSString has a class method stringWithUTF8String that creates a new
NSString from a C string. Objective-C box expression @(...) can be used to
create an NSString instead of invoking the stringWithUTF8String method directly
(The compiler lowers it down to the invocation though). This commit ensures that
the type of @(string-value) gets the same nullability attributes as the return
type of stringWithUTF8String to ensure that the diagnostics are consistent
between the two.
Add CoreOption flag to "-coverage" option to make it available for clang-cl
Summary:
The -coverage option is not a CoreOption, so it is not available to clang-cl.
This patch adds the CoreOption flag to "-coverage" to allow it to be used with clang-cl.
Masud Rahman [Wed, 8 Nov 2017 19:17:27 +0000 (19:17 +0000)]
[bindings] fix TLS test failure
Since cfe commit r237337, '__declspec(thread)' and 'thread_local' have
been the same since MSVC 2015. i.e. they are both considered to supply
a dynamic TLS kind, not a static TLS kind.
This test originally did not specify which version of MS compatibility
to assume. As a result, the test was brittle, since changing the
default compatibility version could break the test.
This commit adds a specific version when building up the flags used to
parse the translation unit, and tests both versions.
Artem Dergachev [Wed, 8 Nov 2017 17:27:58 +0000 (17:27 +0000)]
[analyzer] Fix a crash on logical operators with vectors.
Do not crash when trying to compute x && y or x || y where x and y are
of a vector type.
For now we do not seem to properly model operations with vectors. In particular,
operations && and || on a pair of vectors are not short-circuit, unlike regular
logical operators, so even our CFG is incorrect.
Avoid the crash, add respective FIXME tests for later.
Clang's completion output is non-deterministic, causing test failures
with turned on LLVM_REVERSE_ITERATION.
The workaround is to use CHECK-DAGs for now, will remove them when
PR35244 gets fixed.
Ilya Biryukov [Wed, 8 Nov 2017 10:39:09 +0000 (10:39 +0000)]
Avoid printing some redundant name qualifiers in completion
Summary:
Adjusted PrintingPolicy inside code completion to avoid printing some
redundant name qualifiers.
Before this change, typedefs that were written unqualified in source
code were printed with qualifiers in completion. For example, in the
following code
struct foo {
typedef int type;
type method();
};
completion item for `method` had return type of `foo::type`, even
though the original code used `type` without qualifiers.
After this change, the completion item has return type `type`, as
originally written in the source code.
Note that this change does not suppress qualifiers written by the
user. For example, in the following code
typedef int type;
struct foo {
typedef int type;
::type method(foo::type);
};
completion item for `method` has return type of `::type` and
parameter type of `foo::type`, as originally written in the source
code.
Justin Lebar [Tue, 7 Nov 2017 22:10:54 +0000 (22:10 +0000)]
[NVPTX] Implement __nvvm_atom_add_gen_d builtin.
Summary:
This just seems to have been an oversight. We already supported the f64
atomic add with an explicit scope (e.g. "cta"), but not the scopeless
version.
George Rokos [Tue, 7 Nov 2017 18:27:04 +0000 (18:27 +0000)]
Clang/libomptarget map interface flag renaming - NFC patch
This patch renames some of the flag names of the clang/libomptarget map interface. The old names are slightly misleading, whereas the new ones describe in a better way what each flag is about.
Only the macros within the enumeration are renamed, there is no change in functionality therefore there are no updated regression tests.
Devin Coughlin [Mon, 6 Nov 2017 22:12:19 +0000 (22:12 +0000)]
[analyzer] Model correct dispatch_once() 'done' value in BodyFarm
The analyzer's BodyFarm models dispatch_once() by comparing the passed-in
predicate against a known 'done' value. If the predicate does not have that
value, the model updates the predicate to have that value and executes the
passed in block.
Unfortunately, the current model uses the wrong 'done' value: 1 instead of ~0.
This interferes with libdispatch's static inline function _dispatch_once(),
which enables a fast path if the block has already been executed. That function
uses __builtin_assume() to tell the compiler that the done flag is set to ~0 on
exit. When r302880 added modeling of __builtin_assume(), this caused the
analyzer to assume 1 == ~0. This in turn caused the analyzer to never explore any code after a call to dispatch_once().
This patch regains the missing coverage by updating BodyFarm to use the correct
'done' value.
Dave Lee [Mon, 6 Nov 2017 21:18:05 +0000 (21:18 +0000)]
Vary Windows toolchain selection by -fuse-ld
Summary:
This change allows binutils to be used for linking with MSVC. Currently, when
using an MSVC target and `-fuse-ld=bfd`, the driver produces an invalid linker
invocation.
Craig Topper [Mon, 6 Nov 2017 21:00:49 +0000 (21:00 +0000)]
[X86] Replace the mask cmpeq/cmple/cmplt/cmpgt/cmpge/cmpneq intrinsics with macros that just pass the right comparison predicate value to the regular cmp intrinsic. Remove mask cmpeq/cmpgt builtins that are now unused.
This shortens the intrinsic headers a little and allows us to get rid of the cmpeq and cmpgt handling from CGBuiltin.cpp.
Martell Malone [Sat, 4 Nov 2017 02:07:59 +0000 (02:07 +0000)]
[Driver] MinGW: Remove custom linker detection
This is a re-apply of rL313082 which was reverted in rL313088
In rL289668 the ability to specify the default linker at compile time
was added but because the MinGW driver used custom detection we could
not take advantage of this new CMAKE flag CLANG_DEFAULT_LINKER.
rL289668 added no test cases and the mingw driver was either overlooked
or purposefully skipped because it has some custom linker tests
Removing them here because they are covered by the generic case.
Richard Trieu [Sat, 4 Nov 2017 01:20:50 +0000 (01:20 +0000)]
Update ODR hashing tests
Add a mix of postive and negative tests to check that wrong Decls won't be
flagged in the diagnostic. Split the check everything test and moved the
pieces closer to where the related tests are.
Alex Lorenz [Fri, 3 Nov 2017 18:11:22 +0000 (18:11 +0000)]
[refactor][extract] insert semicolons into extracted/inserted code
when needed
This commit implements the semicolon insertion logic into the extract
refactoring. The following rules are used:
- extracting expression: add terminating ';' to the extracted function.
- extracting statements that don't require terminating ';' (e.g. switch): add
terminating ';' to the callee.
- extracting statements with ';': move (if possible) the original ';' from the
callee and add terminating ';'.
- otherwise, add ';' to both places.
[Driver] Add Scudo as a possible -fsanitize= option
Summary:
This change adds Scudo as a possible Sanitizer option via -fsanitize=.
This allows for easier static & shared linking of the Scudo library, it allows
us to enforce PIE (otherwise the security of the allocator is moot), and check
for incompatible Sanitizers combo.
In its current form, Scudo is not compatible with any other Sanitizer, but the
plan is to make it work in conjunction with UBsan (-fsanitize=scudo,undefined),
which will require additional work outside of the scope of this change.
[clang-format] Sort using-declarations case sensitively with a special case for '_'
Summary:
This makes clang-format sort using declarations case-sensitive with the
exception that '_' comes just before 'A'. This is better than the current case
insensitive version, because it groups uppercase names in the same namespace
together.
Erik Verbruggen [Fri, 3 Nov 2017 09:40:07 +0000 (09:40 +0000)]
[preamble] Also record the "skipping" state of the preprocessor
When a preamble ends in a conditional preprocessor block that is being
skipped, the preprocessor needs to continue skipping that block when
the preamble is used.
Justin Lebar [Fri, 3 Nov 2017 02:30:00 +0000 (02:30 +0000)]
[CUDA] Mark CUDA as a no-errno platform.
Summary:
CUDA doesn't support errno at all, so this is the right thing -- or at
least, in the right direction.
But also, this unbreaks the CUDA test-suite math tests [0] after D39481.
__cuda_cmath.h forwards nexttoward() to __builtin_nexttoward, which,
after that patch, was lowering to a libcall that doesn't exist in NVPTX.
David Blaikie [Thu, 2 Nov 2017 22:28:50 +0000 (22:28 +0000)]
Modular Codegen: Don't home always_inline functions
Since they'll likely (not always - if the address is taken, etc) be
inlined away, even at -O0, separately provided weak definitions are
likely to be unused so skip all of that.
David Blaikie [Thu, 2 Nov 2017 21:55:40 +0000 (21:55 +0000)]
Modular Codegen: Don't home/modularize static functions in headers
Consistent with various workarounds in the backwards compatible modules
that allow static functions in headers to exist, be deduplicated to some
degree, and not generally fail right out of the gate... do the same with
modular codegen as there are enough cases (including in libstdc++ and in
LLVM itself - though I cleaned up the easy ones) that it's worth
supporting as a migration/backcompat step.
Simply create a separate, internal linkage function in each object that
needs it. If an available_externally/modularized function references a
static function, but the modularized function is eventually dropped and
not inlined, the static function will be dropped as unreferenced.
Sanjay Patel [Thu, 2 Nov 2017 20:39:26 +0000 (20:39 +0000)]
[CodeGen] fix const-ness of builtin equivalents of <math.h> and <complex.h> functions that might set errno
This just makes const-ness of the builtins match const-ness of their lib function siblings.
We're deferring fixing some of these that are obviously wrong to follow-up patches.
Hopefully, the bugs are visible in the new test file (added at rL317220).
As the description in Builtins.def says: "e = const, but only when -fmath-errno=0".
This is step 2 of N to fix builtins and math calls as discussed in D39204.
Eric Liu [Thu, 2 Nov 2017 12:48:48 +0000 (12:48 +0000)]
Fix clang-format CLion integration bug.
CLion's Sax parser threw this error:
Failed to parse clang-format XML replacements. Input: <?xml version='1.0'?>
<replacements xml:space='preserve' incomplete_format='true' line=89>
[...]
[org.xml.sax.SAXParseException; lineNumber: 2; columnNumber: 66; Open quote is expected for attribute "line" associated with an element type "replacements".]
Shoaib Meenai [Thu, 2 Nov 2017 05:04:00 +0000 (05:04 +0000)]
[libclang] Add dummy libclang-headers target
`LLVM_DISTRIBUTION_COMPONENTS` assumes that each component has both
`component` and `install-component` targets. Add a dummy no-op target
for `libclang-headers` to placate this check.
Shoaib Meenai [Thu, 2 Nov 2017 01:11:40 +0000 (01:11 +0000)]
[clang-rename] Use add_clang_tool
`add_clang_tool` includes a call to `add_clang_executable`, but it also
sets up the install rule, and adds an `install-*` target. The latter is
required for using `LLVM_DISTRIBUTION_COMPONENTS`.
[clang-format] Make parseUnaryOperator non-recursive, NFCI
Summary:
This patch makes the implementation of parseUnaryOperator non-recursive. We had
a problem with a file starting with tens of thousands of +'es and -'es which
caused clang-format to stack overflow.
projects / clang / log