This, in instcombine, allows conversions to i8/i16/i32 (very
common cases) even if the resulting type is not legal according
to the data layout. This can often open up extra combine
opportunities.

Differential Revision: https://reviews.llvm.org/D42424

llvm-svn: 324174

Fixed build issue when building with g++-4.8 (specialization after instantiation).

llvm-svn: 324173

From the discussion in D41835 it looks possible the change will be backed out, 
but for now let's fix the RISCV tests.

llvm-svn: 324172

llvm-svn: 324171

Previously, RISCV32TargetInfo or RISCV64TargetInfo were created 
unconditionally. Use LinuxTargetInfo<RISCV??TargetInfo> to ensure that the 
proper OS-specific defines are present.

This patch only adds logic to instantiate LinuxTargetInfo and leaves a TODO, 
as I'm reluctant to add logic for other targets (e.g. FreeBSD, RTEMS) until 
I've produced and tested at least one binary for that OS+target combo.

Thanks to @mgrang to reporting the issue.

llvm-svn: 324170

Splitting basic blocks into multiple statements if there are now
additional scalar dependencies gives more freedom to the scheduler, but
more statements also means higher compile-time complexity. Switch to
finer statement granularity, the additional compile time should be
limited by the number of operations quota.

The regression tests are written for the -polly-stmt-granularity=bb
setting, therefore we add that flag to those tests that break with the
new default. Some of the tests only fail because the statements are
named differently due to a basic block resulting in multiple statements,
but which are removed during simplification of statements without
side-effects. Previous commits tried to reduce this effect, but it is
not completely avoidable.

Differential Revision: https://reviews.llvm.org/D42151

llvm-svn: 324169

Do not add a "_last" suffix to the statement name if there is no (other)
main statement for a basic block. In other words, it becomes the main
statement itself. This further reduces the statement naming difference
between -polly-stmt-granularity=bb and
-polly-stmt-granularity=scalar-indep.

llvm-svn: 324168

Due to Buildbot failures - most likely that's because target triples were not
specified in the tests, even though the checker behaves differently with
different target triples.

llvm-svn: 324167

This is a security check which is disabled by default but will be enabled
whenever the user consciously enables the security package. If mmap()ed memory
is both writable and executable, it makes it easier for the attacker to execute
arbitrary code when contents of this memory are compromised. Some applications
require such mmap()s though, such as different sorts of JIT.

Patch by David Carlier!

Differential Revision: https://reviews.llvm.org/D42645

llvm-svn: 324166

llvm-svn: 324165

When Clang encounters an already invalid class declaration, it can
emit incorrect diagnostics about the exception specification on
some of its members. This patch temporarily works around that
incorrect diagnostic.

The clang bug was introduced in r324062.

llvm-svn: 324164

Reviewers: kcc

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D42874

llvm-svn: 324163

[analyzer] Do not infer nullability inside function-like macros, even when macro is explicitly returning NULL

We already suppress such reports for inlined functions, we should then
get the same behavior for macros.
The underlying reason is that the same macro, can be called from many
different contexts, and nullability can only be expected in _some_ of
them.
Assuming that the macro can return null in _all_ of them sometimes leads
to a large number of false positives.

E.g. consider the test case for the dynamic cast implementation in
macro: in such cases, the bug report is unwanted.

Tracked in rdar://36304776

Differential Revision: https://reviews.llvm.org/D42404

llvm-svn: 324161

We could in principle support such pack expansion, using techniques similar to
what we do for pack expansion of lambdas, but it's not clear it's worthwhile.
For now at least, cleanly reject these cases rather than crashing.

llvm-svn: 324160

in debugserver.  This is already set this way in the lldb
project files but not in debugserver.  Updating for
consistency.

llvm-svn: 324158

Group all synthetic symbols in the in single struct to match
the ELF linker.

This change is part of a larger change to add more linker
symbols such as `_end` and `_edata`.

Differential Revision: https://reviews.llvm.org/D42866

llvm-svn: 324157

into a std::string so we don't run off the end of the array when
there is no nul byte in ProcessElfCore::parseLinuxNotes.  
Found with ASAN testing.

<rdar://problem/37134319> 

Differential revision: https://reviews.llvm.org/D42828

llvm-svn: 324156

llvm-svn: 324155

llvm-svn: 324154

Clang previously reported an empty union as having a unique object
representation. This was incorrect and was fixed in a recent Clang commit.

This patch fixes the libc++ tests.

llvm-svn: 324153

We want to check that we can assign to an lvalue here, not a prvalue.

llvm-svn: 324152

parsing a trailing-return-type of a (function pointer) variable declaration.

llvm-svn: 324151

GNU linkers have this option.

Differential Revision: https://reviews.llvm.org/D42858

llvm-svn: 324150

Summary:
Small NFC change to change the name of the function used getting and setting
the alignment of a memset.

llvm-svn: 324148

This is running pre-legalize, we should try to use target independent nodes. This will give the best opportunity for target independent optimizations.

llvm-svn: 324147

Differential Revision: https://reviews.llvm.org/D42823

llvm-svn: 324146

Differential Revision: https://reviews.llvm.org/D42865

llvm-svn: 324145

Without extra instructions and uses, swapMayExposeCSEOpportunities() would change
the icmp (as seen in the check lines), so we were not actually testing patterns 
that should be handled by D41480.

llvm-svn: 324143

llvm-svn: 324142

Differential Revision: https://reviews.llvm.org/D42859

llvm-svn: 324141

Use NULL instead of nullptr in invalid-pointer-pairs-compare-null.cc; nullptr is not available on older Darwin systems.

llvm-svn: 324136

llvm-svn: 324135

Summary:
Clang incorrectly reports empty unions as having a unique object representation. However, this is not correct since `sizeof(EmptyUnion) == 1` AKA it has 8 bits of padding. Therefore it should be treated the same as an empty struct and report `false`.

@erichkeane also suggested this fix should be merged into the 6.0 release branch, so the initial release of `__has_unique_object_representations` is as bug-free as possible. 

Reviewers: erichkeane, rsmith, aaron.ballman, majnemer

Reviewed By: erichkeane

Subscribers: cfe-commits, erichkeane

Differential Revision: https://reviews.llvm.org/D42863

llvm-svn: 324134

llvm-svn: 324133

libprotobuf-mutator accepts protobufs with missing fields, which means
clang-proto-fuzzer does as well.  clang-proto-to-cxx should match this
behavior.

llvm-svn: 324132

Summary:
This is an alternative approach to D42014 after some
investigation by stephanemoore@ and myself.

Previously, the format parameter `BinPackParameters` controlled both
C function parameter list bin-packing and Objective-C protocol conformance
list bin-packing.

We found in the Google style, some teams were changing
`BinPackParameters` from its default (`true`) to `false` so they could
lay out Objective-C protocol conformance list items one-per-line
instead of bin-packing them into as few lines as possible.

To allow teams to use one-per-line Objective-C protocol lists without
changing bin-packing for other areas like C function parameter lists,
this diff introduces a new LibFormat parameter
`ObjCBinPackProtocolList` to control the behavior just for ObjC
protocol conformance lists.

The new parameter is an enum which defaults to `Auto` to keep the
previous behavior (delegating to `BinPackParameters`).

Depends On D42649

Test Plan: New tests added. make -j12 FormatTests && ./tools/clang/unittests/Format/FormatTests

Reviewers: jolesiak, stephanemoore, djasper

Reviewed By: stephanemoore

Subscribers: Wizard, hokein, cfe-commits, klimek

Differential Revision: https://reviews.llvm.org/D42650

llvm-svn: 324131

[X86] Add avx512 command line to ptest.ll to demonstrate that 512-bit vectors are not handled by LowerVectorAllZeroTest.

llvm-svn: 324130

Partially revert r324124 [X86] Add tests for missed opportunities to use ptest for all ones comparison.

Turns out I misunderstood the flag behavior of PTEST because I read the documentation for KORTEST which is different than PTEST/KTEST and made a bad assumption.

Keep the test rename though cause that's useful.

llvm-svn: 324129

Summary:
Previously, Clang only emitted label names in assert builds.
However there is a CC1 option -discard-value-names that should have been used to control emission instead.
    
This patch removes the NDEBUG preprocessor block and instead allows LLVM to handle removing the names in accordance with the option.


Reviewers: erichkeane, aaron.ballman, majnemer

Reviewed By: aaron.ballman

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D42829

llvm-svn: 324127

The "sleep(5)" sometimes times out on our bots, causing the test to fail. Let's use pthread_join.

Differential Revision: https://reviews.llvm.org/D42862

llvm-svn: 324126