platform/upstream/llvm.git
5 years ago[llvm-objcopy] Remove %p format specifiers
Eugene Leviant [Thu, 30 May 2019 09:09:01 +0000 (09:09 +0000)]
[llvm-objcopy] Remove %p format specifiers

On 32-bit machines %p expects 32 bit values, however
addresses in llvm-objcopy are always 64 bits.

llvm-svn: 362074

5 years ago[AArch64][SVE2] Asm: support SVE2 vector splice (constructive)
Cullen Rhodes [Thu, 30 May 2019 08:51:39 +0000 (08:51 +0000)]
[AArch64][SVE2] Asm: support SVE2 vector splice (constructive)

Summary:
The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

Reviewed By: SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D62530

llvm-svn: 362073

5 years ago[AArch64][SVE2] Asm: support SVE2 load instructions
Cullen Rhodes [Thu, 30 May 2019 08:44:27 +0000 (08:44 +0000)]
[AArch64][SVE2] Asm: support SVE2 load instructions

Summary:
Patch adds support for the following instructions:
    * LDNT1SB, LDNT1B, LDNT1SH, LDNT1H, LDNT1SW, LDNT1W, LDNT1D

The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

Reviewed By: SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D62528

llvm-svn: 362072

5 years ago[AArch64][SVE2] Asm: support FCVTX/FLOGB instructions
Cullen Rhodes [Thu, 30 May 2019 08:35:12 +0000 (08:35 +0000)]
[AArch64][SVE2] Asm: support FCVTX/FLOGB instructions

Summary:

Patch completes SVE2 support for:

    SVE Floating Point Unary Operations - Predicated Group

The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

Reviewed By: SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D62526

llvm-svn: 362071

5 years ago[AArch64][SVE2] Asm: add ext (immediate offset, constructive) instruction
Cullen Rhodes [Thu, 30 May 2019 08:25:17 +0000 (08:25 +0000)]
[AArch64][SVE2] Asm: add ext (immediate offset, constructive) instruction

Summary:
The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

Reviewed By: chill

Differential Revision: https://reviews.llvm.org/D62518

llvm-svn: 362070

5 years agoMake CompileUnit::GetSupportFiles return a const list
Pavel Labath [Thu, 30 May 2019 08:21:25 +0000 (08:21 +0000)]
Make CompileUnit::GetSupportFiles return a const list

There's no reason for anyone to modify a list from outside of a symbol
file (as that would break a lot of invariants that symbol files depend
on).

Make the function return a const FileSpecList and fix up a couple of
places that were needlessly binding non-const references to the result
of this function.

llvm-svn: 362069

5 years ago[ARM] Add an MVE execution domain
Sjoerd Meijer [Thu, 30 May 2019 08:07:06 +0000 (08:07 +0000)]
[ARM] Add an MVE execution domain

MVE architecturally specifies a 'beat' system in which a vector
instruction executed now will complete its actual operation over the
next four cycles, so it can overlap with the execution of the previous
and next MVE instruction.

This makes it generally an advantage to avoid moving values back and
forth between MVE registers and anywhere else, if there's any sensible
way to do the same processing in whatever register type the values
already occupied.

That's just what the 'execution domain' system is supposed to achieve.
So here we add a new execution domain which will contain all the MVE
vector instructions when they are added.

Patch by: Simon Tatham

Differential Revision: https://reviews.llvm.org/D60703

llvm-svn: 362068

5 years agoasm goto: fix out-of-bounds read of Constraints after rC362045
Fangrui Song [Thu, 30 May 2019 08:03:02 +0000 (08:03 +0000)]
asm goto: fix out-of-bounds read of Constraints after rC362045

When parsing goto labels, Names and Exprs are expanded but Constraints
is not, this may cause a out-of-bounds read later in:

// GCCAsmStmt::GCCAsmStmt
// `constraints` has only `NumExprs - NumLabels` elements
  Constraints = new (C) StringLiteral*[NumExprs];
  std::copy(constraints, constraints + NumExprs, Constraints);

llvm-svn: 362067

5 years ago[TableGen] AsmMatcher: allow repeated input operands
Sjoerd Meijer [Thu, 30 May 2019 07:38:09 +0000 (07:38 +0000)]
[TableGen] AsmMatcher: allow repeated input operands

If an assembly instruction has to mention an input operand name twice,
for example the MVE VMOV instruction that accesses two lanes of the
same vector by writing 'vmov r1, r2, q0[3], q0[1]', then the obvious
way to write its AsmString is to include the same operand (here $Qd)
twice. But this causes the AsmMatcher generator to omit that
instruction completely from the match table, on the basis that the
generator isn't clever enough to deal with the duplication.

But you need to have _some_ way of dealing with an instruction like
this - and in this case, where the mnemonic is shared with many other
instructions that the AsmMatcher does handle, it would be very painful
to take it out of the AsmMatcher system completely.

A nicer way is to add a custom AsmMatchConverter routine, and let that
deal with the problem if the autogenerated converter can't. But that
doesn't work, because TableGen leaves the instruction out of its table
_even_ if you provide a custom converter.

Solution: this change, which makes TableGen relax the restriction on
duplicated operands in the case where there's a custom converter.

Patch by: Simon Tatham

Differential Revision: https://reviews.llvm.org/D60695

llvm-svn: 362066

5 years ago[CMake] Use find_package(LLVM) instead of LLVMConfig
Petr Hosek [Thu, 30 May 2019 07:34:39 +0000 (07:34 +0000)]
[CMake] Use find_package(LLVM) instead of LLVMConfig

This addresses an issues introduced in r362047.

Differential Revision: https://reviews.llvm.org/D62640

llvm-svn: 362065

5 years ago[TableGen] New default operand "undef_tied_input"
Sjoerd Meijer [Thu, 30 May 2019 07:30:37 +0000 (07:30 +0000)]
[TableGen] New default operand "undef_tied_input"

This is a new special identifier which you can use as a default in
OperandWithDefaultOps. The idea is that you use it for an input
operand of an instruction that's tied to an output operand, and its
semantics are that (in the default case) the input operand's value is
not used at all.

The detailed effect is that when instruction selection emits the
instruction in the form of a pre-regalloc MachineInstr, it creates an
IMPLICIT_DEF node to use as that input.

If you're creating an MCInst with explicit register names, then the
right handling would be to set the input operand to the same register
as the output one (honouring the tie) and to add the 'undef' flag
indicating that that register is deemed to acquire a new don't-care
definition just before we read it. But I haven't done that in this
commit, because there was no need to - no Tablegen backend seems to
autogenerate default fields in an MCInst.

Patch by: Simon Tatham

Differential Revision: https://reviews.llvm.org/D60696

llvm-svn: 362064

5 years ago[lldb-server] Support 'g' packets
Pavel Labath [Thu, 30 May 2019 07:25:22 +0000 (07:25 +0000)]
[lldb-server] Support 'g' packets

Differential Revision: https://reviews.llvm.org/D62221
Patch by Guilherme Andrade <guiandrade@google.com>.

llvm-svn: 362063

5 years ago[AST] asm goto labels don't have constraints, don't try to copy them.
Benjamin Kramer [Thu, 30 May 2019 07:21:08 +0000 (07:21 +0000)]
[AST] asm goto labels don't have constraints, don't try to copy them.

Found by asan.

llvm-svn: 362062

5 years ago[runtimes] Use _LIBCPP_HAS_COMMENT_LIB_PRAGMA in all relevant files
Petr Hosek [Thu, 30 May 2019 06:57:27 +0000 (06:57 +0000)]
[runtimes] Use _LIBCPP_HAS_COMMENT_LIB_PRAGMA in all relevant files

These two sources were omitted in r362055.

llvm-svn: 362061

5 years ago[LoopVectorize] Precommit tests for D62510. NFC
Craig Topper [Thu, 30 May 2019 06:48:13 +0000 (06:48 +0000)]
[LoopVectorize] Precommit tests for D62510. NFC

llvm-svn: 362060

5 years agoMark CodeGen/asm-goto.c as x86 specific after r362045
Fangrui Song [Thu, 30 May 2019 06:48:13 +0000 (06:48 +0000)]
Mark CodeGen/asm-goto.c as x86 specific after r362045

llvm-svn: 362059

5 years ago[libcxx][libcxxabi] Remove the unused CMake checks
Petr Hosek [Thu, 30 May 2019 06:08:56 +0000 (06:08 +0000)]
[libcxx][libcxxabi] Remove the unused CMake checks

These seemed to have been used in the past but were since removed
by the add_compile_flags_if_supported functions that combine these
these checks and adding the flag, but the original checks were never
removed.

Differential Revision: https://reviews.llvm.org/D62566

llvm-svn: 362058

5 years ago[runtimes] Use -Wunknown-pragmas for the pragma check
Petr Hosek [Thu, 30 May 2019 05:38:06 +0000 (05:38 +0000)]
[runtimes] Use -Wunknown-pragmas for the pragma check

This is a follow up to r362055, we need -Wunknown-pragmas otherwise
the check is going to succeed it the pragma isn't supported.

llvm-svn: 362057

5 years ago[LV] Inform about exactly reason of loop illegality
Florian Hahn [Thu, 30 May 2019 05:03:12 +0000 (05:03 +0000)]
[LV] Inform about exactly reason of loop illegality

Currently, only the following information is provided by LoopVectorizer
in the case when the CF of the loop is not legal for vectorization:

 LV: Can't vectorize the instructions or CFG
    LV: Not vectorizing: Cannot prove legality.

But this information is not enough for the root cause analysis; what is
exactly wrong with the loop should also be printed:

 LV: Not vectorizing: The exiting block is not the loop latch.

Patch by Pavel Samolysov.

Reviewers: mkuper, hsaito, rengolin, fhahn

Reviewed By: fhahn

Differential Revision: https://reviews.llvm.org/D62311

llvm-svn: 362056

5 years ago[runtimes] Check if pragma comment(lib, ...) is supported first
Petr Hosek [Thu, 30 May 2019 04:40:21 +0000 (04:40 +0000)]
[runtimes] Check if pragma comment(lib, ...) is supported first

This fixes the issue introduced by r362048 where we always use
pragma comment(lib, ...) for dependent libraries when the compiler
is Clang, but older Clang versions don't support this pragma so
we need to check first if it's supported before using it.

llvm-svn: 362055

5 years agoAdd the `objc_class_stub` attribute.
John McCall [Thu, 30 May 2019 04:09:01 +0000 (04:09 +0000)]
Add the `objc_class_stub` attribute.

Swift requires certain classes to be not just initialized lazily on first
use, but actually allocated lazily using information that is only available
at runtime.  This is incompatible with ObjC class initialization, or at least
not efficiently compatible, because there is no meaningful class symbol
that can be put in a class-ref variable at load time.  This leaves ObjC
code unable to access such classes, which is undesirable.

objc_class_stub says that class references should be resolved by calling
a new ObjC runtime function with a pointer to a new "class stub" structure.
Non-ObjC compilers (like Swift) can simply emit this structure when ObjC
interop is required for a class that cannot be statically allocated,
then apply this attribute to the `@interface` in the generated ObjC header
for the class.

This attribute can be thought of as a generalization of the existing
`objc_runtime_visible` attribute which permits more efficient class
resolution as well as supporting the additon of categories to the class.
Subclassing these classes from ObjC is currently not allowed.

Patch by Slava Pestov!

llvm-svn: 362054

5 years ago[X86] Add ENQCMD instructions
Pengfei Wang [Thu, 30 May 2019 03:59:16 +0000 (03:59 +0000)]
[X86] Add ENQCMD instructions

For more details about these instructions, please refer to the latest
ISE document:
https://software.intel.com/en-us/download/intel-architecture-instruction-set-extensions-programming-reference.

Patch by Tianqing Wang (tianqing)

Differential Revision: https://reviews.llvm.org/D62281

llvm-svn: 362053

5 years ago[Driver] Render target options (e.g. -fuse-init-array) for -fembed-bitcode
Fangrui Song [Thu, 30 May 2019 02:30:04 +0000 (02:30 +0000)]
[Driver] Render target options (e.g. -fuse-init-array) for -fembed-bitcode

Modern ELF platforms use -fuse-init-array to emit .init_array instead of
.ctors .  ld.bfd and gold --ctors-in-init-array merge .init_array and
.ctors into .init_array but lld doesn't do that.

If crtbegin*.o crtend*.o don't provide .ctors/.dtors, such .ctors in
user object files can lead to crash (see PR42002. The first and the last
elements in .ctors/.dtors are ignored - they are traditionally provided
by crtbegin*.o crtend*.o).

Call addClangTargetOptions() to ensure -fuse-init-array is rendered on
modern ELF platforms. On Hexagon, this renders -target-feature
+reserved-r19 for -ffixed-r19.

Reviewed By: compnerd

Differential Revision: https://reviews.llvm.org/D62509

llvm-svn: 362052

5 years ago[PowerPC] Set the default PLT mode on musl to Secure PLT
Fangrui Song [Thu, 30 May 2019 02:13:15 +0000 (02:13 +0000)]
[PowerPC] Set the default PLT mode on musl to Secure PLT

The musl libc only supports Secure PLT.

Patch by A. Wilcox!

Reviewed By: jhibbits

Differential Revision: https://reviews.llvm.org/D59185

llvm-svn: 362051

5 years ago[Driver] -static-pie: add -z text
Fangrui Song [Thu, 30 May 2019 01:55:43 +0000 (01:55 +0000)]
[Driver] -static-pie: add -z text

This matches gcc -static-pie. The intention is to prevent dynamic
relocations in read-only segments.

In ld.bfd and gold, -z notext is the default. If text relocations are needed:

* -z notext: allow and emit DF_TEXTREL.
  DF_TEXTREL is not emitted if there is no text relocation.
* -z text: error

In lld, -z text is the default (this change is a no-op).

* -z text: error on text relocations
* -z notext: allow text relocations, and emit DF_TEXTREL no matter whether
  text relocations exist.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D62606

llvm-svn: 362050

5 years agoDWARFDebugInfoEntry: delete unused Extract() and rename FastExtract() to Extract()
Fangrui Song [Thu, 30 May 2019 01:51:16 +0000 (01:51 +0000)]
DWARFDebugInfoEntry: delete unused Extract() and rename FastExtract() to Extract()

The function Extract() is almost a duplicate of FastExtract() but is not used.
Delete it and rename FastExtract() to Extract().

Reviewed By: JDevlieghere

Differential Revision: https://reviews.llvm.org/D62593

llvm-svn: 362049

5 years ago[runtimes] Support ELF dependent libraries feature
Petr Hosek [Thu, 30 May 2019 01:34:41 +0000 (01:34 +0000)]
[runtimes] Support ELF dependent libraries feature

As of r360984, LLD supports dependent libraries feature for ELF.
libunwind, libc++abi and libc++ have library dependencies: libdl librt
and libpthread, which means that when libunwind and libc++ are being
statically linked (using -static-libstdc++ flag), user has to manually
specify -ldl -lpthread which is onerous.

This change includes the lib pragma to specify the library dependencies
directly in the source that uses those libraries. This doesn't make any
difference when using linkers that don't support dependent libraries.
However, when using LLD that has dependent libraries feature, users no
longer have to manually specifying library dependencies when using
static linking, linker will pick the library automatically.

Differential Revision: https://reviews.llvm.org/D62090

llvm-svn: 362048

5 years ago[CMake] Set LLVM_PATH in the runtimes build
Petr Hosek [Thu, 30 May 2019 01:24:31 +0000 (01:24 +0000)]
[CMake] Set LLVM_PATH in the runtimes build

This avoids using llvm-config for inferring various paths within the
runtimes build. We also set LLVM_INCLUDE_DIR variable that's used by
these builds and move assignment of LLVM_BINARY_DIR and LLVM_LIBRARY_DIR
to the same location for consistency.

Differential Revision: https://reviews.llvm.org/D62637

llvm-svn: 362047

5 years ago[BitstreamWriter][NFC] Remove obsolete comment.
Jan Korous [Thu, 30 May 2019 01:08:38 +0000 (01:08 +0000)]
[BitstreamWriter][NFC] Remove obsolete comment.

The Abbv parameter was just a raw pointer when the comment was written.

llvm-svn: 362046

5 years agoclang support gnu asm goto.
Jennifer Yu [Thu, 30 May 2019 01:05:46 +0000 (01:05 +0000)]
clang support gnu asm goto.
Syntax:
  asm [volatile] goto ( AssemblerTemplate
                      :
                      : InputOperands
                      : Clobbers
                      : GotoLabels)

https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html

New llvm IR is "callbr" for inline asm goto instead "call" for inline asm
For:
asm goto("testl %0, %0; jne %l1;" :: "r"(cond)::label_true, loop);
IR:
callbr void asm sideeffect "testl $0, $0; jne ${1:l};", "r,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %0, i8* blockaddress(@foo, %label_true), i8* blockaddress(@foo, %loop)) #1
          to label %asm.fallthrough [label %label_true, label %loop], !srcloc !3

asm.fallthrough:

Compiler need to generate:
1> a dummy constarint 'X' for each label.
2> an unique fallthrough label for each asm goto stmt " asm.fallthrough%number".

Diagnostic
1> duplicate asm operand name are used in output, input and label.
2> goto out of scope.

llvm-svn: 362045

5 years ago[crashlog] Add a missing call to decode.
Davide Italiano [Thu, 30 May 2019 00:35:43 +0000 (00:35 +0000)]
[crashlog] Add a missing call to decode.

<rdar://problem/51139357>

llvm-svn: 362044

5 years ago[Commands] Remove commented out code
Alex Langford [Wed, 29 May 2019 23:25:44 +0000 (23:25 +0000)]
[Commands] Remove commented out code

llvm-svn: 362042

5 years ago[llvm-objcopy][MachO] Print an error message on use of unsupported options
Seiya Nuta [Wed, 29 May 2019 22:21:12 +0000 (22:21 +0000)]
[llvm-objcopy][MachO] Print an error message on use of unsupported options

Summary:
It is better to print an error message instead of silently ignoring unsupported options.

As mentioned in https://reviews.llvm.org/D57045, this is not the best solution and we should print which flag is not supported at some time.

Reviewers: alexshap, rupprecht, jhenderson, jakehehrlich

Reviewed By: alexshap, rupprecht, jakehehrlich

Subscribers: jakehehrlich, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62578

llvm-svn: 362040

5 years agoMove the pybool logic from CMake to Python for simplicity
Reid Kleckner [Wed, 29 May 2019 21:45:43 +0000 (21:45 +0000)]
Move the pybool logic from CMake to Python for simplicity

This is how similar booleans are handled, and seems to work for me
locally.

llvm-svn: 362039

5 years agoCodeView - add static data members to global variable debug info.
Amy Huang [Wed, 29 May 2019 21:45:34 +0000 (21:45 +0000)]
CodeView - add static data members to global variable debug info.

Summary:
Add static data members to IR debug info's list of global variables
so that they are emitted as S_CONSTANT records.

Related to https://bugs.llvm.org/show_bug.cgi?id=41615.

Reviewers: rnk

Subscribers: aprantl, cfe-commits, llvm-commits, thakis

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D62167

llvm-svn: 362038

5 years ago[llvm-pdbutil] Dump inline call site line table annotations
Reid Kleckner [Wed, 29 May 2019 21:26:25 +0000 (21:26 +0000)]
[llvm-pdbutil] Dump inline call site line table annotations

This ports and improves on some existing llvm-readobj -codeview dumping
functionality that llvm-pdbutil lacked.

Helpful for comparing inline line tables between MSVC and clang.

llvm-svn: 362037

5 years agoUse correct format specifier to silence -Wformat warning.
Richard Trieu [Wed, 29 May 2019 21:25:15 +0000 (21:25 +0000)]
Use correct format specifier to silence -Wformat warning.

llvm-svn: 362035

5 years ago[ObjC] Fix encoding of ObjC pointer types that are pointers to typedefs
Akira Hatanaka [Wed, 29 May 2019 21:23:30 +0000 (21:23 +0000)]
[ObjC] Fix encoding of ObjC pointer types that are pointers to typedefs

clang was encoding pointers to typedefs as if they were pointers to
structs because that is apparently what gcc is doing.

For example:

```
@class Class1;

typedef NSArray<Class1 *> MyArray;

void foo1(void) {
  const char *s0 = @encode(MyArray *); // "^{NSArray=#}"
  const char *s1 = @encode(NSArray<Class1 *> *); // "@"
}
```

This commit removes the code that was there to make clang compatible
with gcc and make clang emit the correct encoding for ObjC pointers,
which is "@".

rdar://problem/50563529

Differential Revision: https://reviews.llvm.org/D61974

llvm-svn: 362034

5 years agoFix Xcode project lldb unit test target so it compiles.
Greg Clayton [Wed, 29 May 2019 21:22:54 +0000 (21:22 +0000)]
Fix Xcode project lldb unit test target so it compiles.

llvm-svn: 362033

5 years ago[Target] Sink some asserts into Process::GetLanguageRuntime
Alex Langford [Wed, 29 May 2019 21:07:53 +0000 (21:07 +0000)]
[Target] Sink some asserts into Process::GetLanguageRuntime

llvm-svn: 362032

5 years agoLoopVersioningLICM: Respect convergent and noduplicate
Matt Arsenault [Wed, 29 May 2019 20:47:59 +0000 (20:47 +0000)]
LoopVersioningLICM: Respect convergent and noduplicate

llvm-svn: 362031

5 years ago[analyzer] Remove EndPath function as it is dead code
Csaba Dabis [Wed, 29 May 2019 20:47:27 +0000 (20:47 +0000)]
[analyzer] Remove EndPath function as it is dead code

Summary: -

Reviewers: george.karpenkov

Reviewed By: george.karpenkov

Subscribers: baloghadamsoftware, cfe-commits, xazax.hun, szepet, a.sidorin,
             mikhail.ramalho, Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D53720

llvm-svn: 362030

5 years agoRevert "IR: add optional type to 'byval' function parameters"
Tim Northover [Wed, 29 May 2019 20:46:38 +0000 (20:46 +0000)]
Revert "IR: add optional type to 'byval' function parameters"

The IRLinker doesn't delve into the new byval attribute when mapping types, and
this breaks LTO.

llvm-svn: 362029

5 years agoRevert "LLVM IR: update Clang tests for byval being a typed attribute."
Tim Northover [Wed, 29 May 2019 20:45:32 +0000 (20:45 +0000)]
Revert "LLVM IR: update Clang tests for byval being a typed attribute."

The underlying LLVM change couldn't cope with llvm-link and broke LTO builds.

llvm-svn: 362028

5 years ago[analyzer] ConditionBRVisitor: Boolean support
Csaba Dabis [Wed, 29 May 2019 20:34:29 +0000 (20:34 +0000)]
[analyzer] ConditionBRVisitor: Boolean support

Summary: -

Reviewers: NoQ, george.karpenkov

Reviewed By: NoQ, george.karpenkov

Subscribers: cfe-commits, xazax.hun, baloghadamsoftware, szepet, a.sidorin,
             mikhail.ramalho, Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D58207

llvm-svn: 362027

5 years ago[analyzer] ConditionBRVisitor: MemberExpr support
Csaba Dabis [Wed, 29 May 2019 20:29:02 +0000 (20:29 +0000)]
[analyzer] ConditionBRVisitor: MemberExpr support

Summary: -

Reviewers: NoQ, george.karpenkov

Reviewed By: NoQ

Subscribers: cfe-commits, xazax.hun, baloghadamsoftware, szepet, a.sidorin,
             mikhail.ramalho, Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D58206

llvm-svn: 362026

5 years ago[analyzer] ConditionBRVisitor: Remove duplicated code
Csaba Dabis [Wed, 29 May 2019 20:18:07 +0000 (20:18 +0000)]
[analyzer] ConditionBRVisitor: Remove duplicated code

Summary: -

Reviewers: NoQ, george.karpenkov

Reviewed By: NoQ

Subscribers: cfe-commits, xazax.hun, baloghadamsoftware, szepet, a.sidorin,
             mikhail.ramalho, Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D58199

llvm-svn: 362025

5 years ago[analyzer] ConditionBRVisitor: Test 'add-pop-up-notes=false'
Csaba Dabis [Wed, 29 May 2019 20:13:29 +0000 (20:13 +0000)]
[analyzer] ConditionBRVisitor: Test 'add-pop-up-notes=false'

Summary: -

Reviewers: NoQ, alexfh

Reviewed By: alexfh

Subscribers: cfe-commits, xazax.hun, baloghadamsoftware, szepet, a.sidorin,
             mikhail.ramalho, Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D61060

llvm-svn: 362023

5 years ago[LoopIdiomRecognize][NFC] Use DEBUG_TYPE, add LLVM_DEBUG() to runOnNoncountableLoop()
Roman Lebedev [Wed, 29 May 2019 20:11:53 +0000 (20:11 +0000)]
[LoopIdiomRecognize][NFC] Use DEBUG_TYPE, add LLVM_DEBUG() to runOnNoncountableLoop()

Split off from D61144

llvm-svn: 362022

5 years ago[ARC] Cleanup ARCAsmPrinter.
Pete Couperus [Wed, 29 May 2019 20:07:35 +0000 (20:07 +0000)]
[ARC] Cleanup ARCAsmPrinter.

Summary:
Remove unused getTargetStreamer.
Remove unused headers.

Reviewers: dantrushin

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62549

llvm-svn: 362021

5 years ago[analyzer] ConditionBRVisitor: Enhance to write out more information
Csaba Dabis [Wed, 29 May 2019 20:06:09 +0000 (20:06 +0000)]
[analyzer] ConditionBRVisitor: Enhance to write out more information

Summary:
Add extra messages to the bug report to inform the user why the analyzer
`Taking true/false branch`.

Reviewers: NoQ, george.karpenkov

Reviewed By: NoQ

Subscribers: gerazo, gsd, dkrupp, whisperity, baloghadamsoftware, xazax.hun,
             eraman, szepet, a.sidorin, mikhail.ramalho, Szelethus,
             donat.nagy, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D53076

llvm-svn: 362020

5 years agoUpdateTestChecks: Lanai triple support
Roman Lebedev [Wed, 29 May 2019 20:03:00 +0000 (20:03 +0000)]
UpdateTestChecks: Lanai triple support

Summary:
The assembly structure most resembles the SPARC pattern:
```
        .globl  f6                      ! -- Begin function f6
        .p2align        2
        .type   f6,@function
f6:                                     ! @f6
        .cfi_startproc
! %bb.0:
        st      %fp, [--%sp]
<...>
        ld      -8[%fp], %fp
.Lfunc_end0:
        .size   f6, .Lfunc_end0-f6
        .cfi_endproc
                                        ! -- End function
```
Test being affected by upcoming patch, so regenerate it.

Reviewers: RKSimon, jpienaar

Reviewed By: RKSimon

Subscribers: jyknight, fedor.sergeev, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62545

llvm-svn: 362019

5 years agogn build: Make it possible to build with coverage information
Nico Weber [Wed, 29 May 2019 20:00:36 +0000 (20:00 +0000)]
gn build: Make it possible to build with coverage information

Differential Revision: https://reviews.llvm.org/D62508

llvm-svn: 362018

5 years ago[DAGCombiner] Replace gathers with a zero mask with the passthru value
Benjamin Kramer [Wed, 29 May 2019 19:24:19 +0000 (19:24 +0000)]
[DAGCombiner] Replace gathers with a zero mask with the passthru value

These can be created by the legalizer when splitting a larger gather.

See https://llvm.org/PR42055 for a motivating example.

Differential Revision: https://reviews.llvm.org/D62613

llvm-svn: 362015

5 years ago[analyzer] [NFC] PathDiagnostic: Create PathDiagnosticPopUpPiece
Csaba Dabis [Wed, 29 May 2019 19:21:59 +0000 (19:21 +0000)]
[analyzer] [NFC] PathDiagnostic: Create PathDiagnosticPopUpPiece

Summary:
This new piece is similar to our macro expansion printing in HTML reports:
On mouse-hover event it pops up on variables. Similar to note pieces it
supports `plist` diagnostics as well.

It is optional, on by default: `add-pop-up-notes=true`.

Extra: In HTML reports `background-color: LemonChiffon` was too light,
changed to `PaleGoldenRod`.

Reviewers: NoQ, alexfh

Reviewed By: NoQ

Subscribers: cfe-commits, gerazo, gsd, george.karpenkov, alexfh, xazax.hun,
             baloghadamsoftware, szepet, a.sidorin, mikhail.ramalho,
             Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D60670

llvm-svn: 362014

5 years agoLLVM IR: update Clang tests for byval being a typed attribute.
Tim Northover [Wed, 29 May 2019 19:13:29 +0000 (19:13 +0000)]
LLVM IR: update Clang tests for byval being a typed attribute.

Since byval is now a typed attribute it gets sorted slightly differently by
LLVM when the order of attributes is being canonicalized. This updates the few
Clang tests that depend on the old order.

llvm-svn: 362013

5 years agoIR: add optional type to 'byval' function parameters
Tim Northover [Wed, 29 May 2019 19:12:48 +0000 (19:12 +0000)]
IR: add optional type to 'byval' function parameters

When we switch to opaque pointer types we will need some way to describe
how many bytes a 'byval' parameter should occupy on the stack. This adds
a (for now) optional extra type parameter.

If present, the type must match the pointee type of the argument.

Note to front-end maintainers: if this causes test failures, it's probably
because the "byval" attribute is printed after attributes without any parameter
after this change.

llvm-svn: 362012

5 years ago[analyzer] print() JSONify chain: Generic stmt_id
Csaba Dabis [Wed, 29 May 2019 18:58:41 +0000 (18:58 +0000)]
[analyzer] print() JSONify chain: Generic stmt_id

Summary: Some environment create less statements so make them generic.
llvm-svn: 362011

5 years agomac: Make ubsan test config look more like asan test config
Nico Weber [Wed, 29 May 2019 18:54:28 +0000 (18:54 +0000)]
mac: Make ubsan test config look more like asan test config

In particular, don't call get_target_flags_for_arch() since that
will cause an error in some situations:

If DARWIN_iossim_ARCHS=i386;x86_64, DARWIN_osx_ARCHS=x86_64, and
DARWIN_iossym_SYSROOT isn't set (due to the simulator sysroot not being
available), then config-ix.cmake won't add i386 to COMPILER_RT_SUPPORTED_ARCH
but ubsan's test/CMakeLists.txt would call get_target_flags_for_arch()
with i386, which would then run into the error in
get_target_flags_for_arch().

Having these conditions isn't ideal. The background here is that we
configure our mac-hosted trunk bots all the same (so they all have the
same DARWIN_*_archs, and we don't easily know if a mac host bot is
targeting mac or ios at the place where we call cmake), but only the
ios-targeting bots have ios sysroots available.

This will hopefully unbreak that use case without impacting anything
else -- and it makes ubsan and asan test setup more alike.

llvm-svn: 362010

5 years ago[analyzer] SATestBuild.py: Use driver for analyzing single-file tests.
Artem Dergachev [Wed, 29 May 2019 18:49:31 +0000 (18:49 +0000)]
[analyzer] SATestBuild.py: Use driver for analyzing single-file tests.

Don't bother coming up with a -cc1 run-line ourselves.

This, in particular, gets rid of a macOS-specific code path.

llvm-svn: 362009

5 years ago[analyzer] print() JSONify: SVal implementation
Csaba Dabis [Wed, 29 May 2019 18:38:52 +0000 (18:38 +0000)]
[analyzer] print() JSONify: SVal implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: cfe-commits, szepet, rnkovacs, a.sidorin, mikhail.ramalho,
             Szelethus, donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62497

llvm-svn: 362008

5 years ago[CMake] [Runtimes] Set *_STANDALONE_BUILD
Chris Bieneman [Wed, 29 May 2019 18:37:49 +0000 (18:37 +0000)]
[CMake] [Runtimes] Set *_STANDALONE_BUILD

Summary:
The runtimes use `*_STANDALONE_BUILD=OFF` to signify that clang is an in-tree target. This is not the case with the runtime builds, so we really need this set to `ON`.

In order to resolve the issues phosek was having with checks, we should use checks that don't link. We can use compiler-rt's `try_compile_only` as a basis for that.

This patch is *required* to be able to run the runtime libraries check-* targets.

Reviewers: smeenai, phosek, compnerd

Reviewed By: phosek

Subscribers: mgorny, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62410

llvm-svn: 362007

5 years ago[InstCombine] Optimize always overflowing signed saturating add/sub
Nikita Popov [Wed, 29 May 2019 18:37:13 +0000 (18:37 +0000)]
[InstCombine] Optimize always overflowing signed saturating add/sub

Based on the overflow direction information added in D62463, we can
now fold always overflowing signed saturating add/sub to signed min/max.

Differential Revision: https://reviews.llvm.org/D62544

llvm-svn: 362006

5 years agoFix an unused-variable error.
Haojian Wu [Wed, 29 May 2019 18:36:54 +0000 (18:36 +0000)]
Fix an unused-variable error.

llvm-svn: 362005

5 years ago[WebAssembly] Support VPtr sanitizer for Emscripten
Thomas Lively [Wed, 29 May 2019 18:31:50 +0000 (18:31 +0000)]
[WebAssembly] Support VPtr sanitizer for Emscripten

Summary:
After https://github.com/emscripten-core/emscripten/pull/8651, Emscripten
supports the full UBSan runtime. This includes the VPtr sanitizer.

This diff allows clang to generate code that uses the VPtr sanitizer for
Emscripten.

Patch by Guanzhong Chen

Reviewers: tlively, aheejin

Reviewed By: aheejin

Subscribers: dschuff, sbc100, jgravelle-google, sunfish, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62559

llvm-svn: 362004

5 years ago[analyzer] print() JSONify: CFG implementation
Csaba Dabis [Wed, 29 May 2019 18:29:31 +0000 (18:29 +0000)]
[analyzer] print() JSONify: CFG implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: cfe-commits, szepet, rnkovacs, a.sidorin, mikhail.ramalho,
             donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62496

llvm-svn: 362003

5 years ago[analyzer] print() JSONify: Decl revision
Csaba Dabis [Wed, 29 May 2019 18:21:14 +0000 (18:21 +0000)]
[analyzer] print() JSONify: Decl revision

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: cfe-commits, szepet, rnkovacs, a.sidorin, mikhail.ramalho,
             donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62495

llvm-svn: 362002

5 years agoAMDGPU: Return address lowering
Aakanksha Patil [Wed, 29 May 2019 18:20:11 +0000 (18:20 +0000)]
AMDGPU: Return address lowering

The patch computes the return address for the current function.

Differential revision: https://reviews.llvm.org/D59666

llvm-svn: 362001

5 years ago[analyzer][AST] print() JSONify: Stmt implementation
Csaba Dabis [Wed, 29 May 2019 18:17:18 +0000 (18:17 +0000)]
[analyzer][AST] print() JSONify: Stmt implementation

Summary:
This patch also adds a function called `JsonFormat()` which:
- Flattens the string so removes the new-lines.
- Escapes double quotes.

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: cfe-commits, szepet, rnkovacs, a.sidorin, mikhail.ramalho,
             donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62494

llvm-svn: 362000

5 years ago[Target] Introduce Process::GetLanguageRuntimes
Alex Langford [Wed, 29 May 2019 18:08:22 +0000 (18:08 +0000)]
[Target] Introduce Process::GetLanguageRuntimes

Summary:
Currently there's not really a good way to iterate over the language runtimes a
process has. This is sometimes desirable (as seen in my change to Thread).
Additionally, there's not really a good reason to iterate over every available
language, but rather only over languages for which we have a plugin loaded.

Reviewers: JDevlieghere, davide, jingham

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D62562

llvm-svn: 361999

5 years agolit: modernize the lit configuration for the lit tests
Stella Stamenova [Wed, 29 May 2019 18:07:39 +0000 (18:07 +0000)]
lit: modernize the lit configuration for the lit tests

Summary: This also normalizes the config feature that represents the windows platform to "system-windows" as opposed to having both "windows" and "system-windows"

Reviewers: asmith, probinson

Subscribers: delcypher, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D61798

llvm-svn: 361998

5 years ago[analyzer] print() JSONify: getNodeLabel implementation
Csaba Dabis [Wed, 29 May 2019 18:05:53 +0000 (18:05 +0000)]
[analyzer] print() JSONify: getNodeLabel implementation

Summary: This patch also rewrites the ProgramPoint printing.

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: cfe-commits, szepet, rnkovacs, a.sidorin, mikhail.ramalho,
             donat.nagy, dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62346

llvm-svn: 361997

5 years ago[Index] Compute correct symbol kind for variable templates
Ilya Biryukov [Wed, 29 May 2019 17:49:30 +0000 (17:49 +0000)]
[Index] Compute correct symbol kind for variable templates

Summary:
The index library itself seems to never pass variable templates as
input, however clangd does.

Reviewers: kadircet

Reviewed By: kadircet

Subscribers: jkorous, arphaman, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62579

llvm-svn: 361996

5 years agoFixed source header [NFC]
Greg Clayton [Wed, 29 May 2019 17:25:03 +0000 (17:25 +0000)]
Fixed source header [NFC]

llvm-svn: 361995

5 years ago[CUDA][HIP] Skip setting `externally_initialized` for static device variables.
Michael Liao [Wed, 29 May 2019 17:23:27 +0000 (17:23 +0000)]
[CUDA][HIP] Skip setting `externally_initialized` for static device variables.

Summary:
- By declaring device variables as `static`, we assume they won't be
  addressable from the host side. Thus, no `externally_initialized` is
  required.

Reviewers: yaxunl

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62603

llvm-svn: 361994

5 years agoYet another attempt to fix buildbot after r361949
Eugene Leviant [Wed, 29 May 2019 17:14:48 +0000 (17:14 +0000)]
Yet another attempt to fix buildbot after r361949

Looks like %p format specifier of createStringError behaves
differently on different platforms

llvm-svn: 361993

5 years agoAdd curley brackets to case statement that declares a variable.
Erich Keane [Wed, 29 May 2019 17:04:48 +0000 (17:04 +0000)]
Add curley brackets to case statement that declares a variable.

llvm-svn: 361992

5 years ago[X86] Fix machineverifier error on avx512f-256-set0.mir
Craig Topper [Wed, 29 May 2019 17:02:27 +0000 (17:02 +0000)]
[X86] Fix machineverifier error on avx512f-256-set0.mir

Previously the pass ran the entire llc pipeline which caused the IR to be recodegened.

This commit restricts it to just running the postrapseudos pass and checking the results of that instead of the final assembly.

llvm-svn: 361991

5 years agoCallSiteSplitting: Respect convergent and noduplicate
Matt Arsenault [Wed, 29 May 2019 16:59:48 +0000 (16:59 +0000)]
CallSiteSplitting: Respect convergent and noduplicate

llvm-svn: 361990

5 years ago[ThinLTO] Use original alias visibility when importing
Teresa Johnson [Wed, 29 May 2019 16:50:46 +0000 (16:50 +0000)]
[ThinLTO] Use original alias visibility when importing

Summary:
When we import an alias, we do so by making a clone of the aliasee. Just
as this clone uses the original alias name and linkage, it should also
use the same visibility (not the aliasee's visibility). Otherwise,
linker behavior is affected (e.g. if the aliasee was hidden, but the
alias is not, the resulting imported clone should not be hidden,
otherwise the linker will make the final symbol hidden which is
incorrect).

Reviewers: wmi

Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62535

llvm-svn: 361989

5 years ago[NFC][InstCombine] Add a unary FNeg test to fsub.ll.
Cameron McInally [Wed, 29 May 2019 16:50:14 +0000 (16:50 +0000)]
[NFC][InstCombine] Add a unary FNeg test to fsub.ll.

llvm-svn: 361988

5 years agoAdd more information to the log timer dump
Antonio Afonso [Wed, 29 May 2019 16:31:32 +0000 (16:31 +0000)]
Add more information to the log timer dump

Summary:
The `log timer dump` is showing the time of the function itself minus any function that is called from this one that also happens to be timed. However, this is really not obvious and it also makes it hard to understand the time spent in total and also which children are actually taking the time.
To get a better reading of the timer dump I added the total, children (which I named child) and also the hit count. I used these timers to figure out a performance issue and only after adding this things were more clear to me.

It looks like this:
```
(lldb) log timer dump
35.447713617 sec (total: 35.449s; child: 0.001s; count: 1374) for void SymbolFileDWARF::Index()
29.717921481 sec (total: 29.718s; child: 0.000s; count: 8230500) for const lldb_private::ConstString &lldb_private::Mangled::GetDemangledName(lldb::LanguageType) const
21.049508865 sec (total: 24.683s; child: 3.633s; count: 1399) for void lldb_private::Symtab::InitNameIndexes()
...
```

Reviewers: clayborg, teemperor, labath, espindola, xiaobai

Reviewed By: labath, xiaobai

Subscribers: emaste, mgorny, arichardson, eraman, MaskRay, jdoerfert, labath, davide, teemperor, aprantl, erik.pilkington, jfb, abidh, lldb-commits

Tags: #lldb

Differential Revision: https://reviews.llvm.org/D61235

llvm-svn: 361987

5 years ago[clang-format] Allow configuring list of function-like macros that resolve to a type
Francois Ferrand [Wed, 29 May 2019 16:30:47 +0000 (16:30 +0000)]
[clang-format] Allow configuring list of function-like macros that resolve to a type

Summary:
Adds a `TypenameMacros` configuration option that causes certain identifiers to be handled in a way similar to `typeof()`.

This is enough to:
- Prevent misinterpreting declarations of pointers to such types as expressions (`STACK_OF(int) * foo` -> `STACK_OF(int) *foo`),
- Avoid surprising line breaks in variable/struct field declarations (`STACK_OF(int)\nfoo;` -> `STACK_OF(int) foo;`, see https://bugs.llvm.org/show_bug.cgi?id=30353).

Reviewers: Typz, krasimir, djasper

Reviewed By: Typz

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D57184

llvm-svn: 361986

5 years agoPartial revert of revert of r361827: Add constrained intrinsic tests for powerpc64le.
Kevin P. Neal [Wed, 29 May 2019 16:29:31 +0000 (16:29 +0000)]
Partial revert of revert of r361827: Add constrained intrinsic tests for powerpc64le.

The powerpc64-"nonle" tests are removed. They fail because of a bug that
Drew is currently working on that affects multiple targets.

Submitted by: Drew Wock <drew.wock@sas.com>
Reviewed by: Hal Finkel, Kevin P. Neal
Approved by: Hal Finkel
Differential Revision: http://reviews.llvm.org/D62388

llvm-svn: 361985

5 years ago[NFC] clang-format: Use LLVM style in NamespaceEndCommentsFixerTest
Francois Ferrand [Wed, 29 May 2019 16:22:43 +0000 (16:22 +0000)]
[NFC] clang-format: Use LLVM style in NamespaceEndCommentsFixerTest

As pointed out in https://reviews.llvm.org/D37813#inline-555026, the
code which is formatted does not match LLVM formatting style.

Technically this is not a problem since these tests bypass most of the
formatter, but it can be misleading.

llvm-svn: 361984

5 years ago[analyzer] print() JSONify: Program state implementation
Csaba Dabis [Wed, 29 May 2019 16:22:21 +0000 (16:22 +0000)]
[analyzer] print() JSONify: Program state implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62087

llvm-svn: 361983

5 years ago[analyzer] print() JSONify: Checker messages implementation
Csaba Dabis [Wed, 29 May 2019 16:02:33 +0000 (16:02 +0000)]
[analyzer] print() JSONify: Checker messages implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62086

llvm-svn: 361982

5 years ago[NFC][libcxx] Remove trailing whitespace
Louis Dionne [Wed, 29 May 2019 16:01:36 +0000 (16:01 +0000)]
[NFC][libcxx] Remove trailing whitespace

It's incredibly annoying when trying to create diffs

llvm-svn: 361981

5 years ago[analyzer] print() JSONify: Constructing objects implementation
Csaba Dabis [Wed, 29 May 2019 15:58:26 +0000 (15:58 +0000)]
[analyzer] print() JSONify: Constructing objects implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62085

llvm-svn: 361980

5 years ago[analyzer] print() JSONify: Type information implementation
Csaba Dabis [Wed, 29 May 2019 15:53:12 +0000 (15:53 +0000)]
[analyzer] print() JSONify: Type information implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62083

llvm-svn: 361979

5 years ago[analyzer] print() JSONify: Constraints implementation
Csaba Dabis [Wed, 29 May 2019 15:43:26 +0000 (15:43 +0000)]
[analyzer] print() JSONify: Constraints implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62082

llvm-svn: 361978

5 years ago[WebAssembly] Move direct call tracking from member to local. NFC.
Sam Clegg [Wed, 29 May 2019 15:41:08 +0000 (15:41 +0000)]
[WebAssembly] Move direct call tracking from member to local. NFC.

This data structure is only needed temporarily while symbols are being
created.

This is a followup on rL361678.

Differential Revision: https://reviews.llvm.org/D62548

llvm-svn: 361977

5 years ago[analyzer] print() JSONify: Environment implementation
Csaba Dabis [Wed, 29 May 2019 15:36:58 +0000 (15:36 +0000)]
[analyzer] print() JSONify: Environment implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62081

llvm-svn: 361976

5 years ago[WebAssembly] Fix signatures of undefined function in LTO object which are not called...
Sam Clegg [Wed, 29 May 2019 15:36:42 +0000 (15:36 +0000)]
[WebAssembly] Fix signatures of undefined function in LTO object which are not called directly.

We recently added special handling for function that are not called
directly but failed to add testing for the LTO case.

See https://reviews.llvm.org/D62153

Differential Revision: https://reviews.llvm.org/D62561

llvm-svn: 361975

5 years ago[CodeComplete] Add semicolon when completing patterns for 'static_assert' and 'typedef
Ilya Biryukov [Wed, 29 May 2019 15:32:17 +0000 (15:32 +0000)]
[CodeComplete] Add semicolon when completing patterns for 'static_assert' and 'typedef

This is a trivial follow-up to r360042, which added semicolons to other
pattern completions, so sending without review.

llvm-svn: 361974

5 years ago[NFC][InstCombine] Add unary FNeg tests to fpcast.ll and fpextend.ll
Cameron McInally [Wed, 29 May 2019 15:29:35 +0000 (15:29 +0000)]
[NFC][InstCombine] Add unary FNeg tests to fpcast.ll and fpextend.ll

llvm-svn: 361973

5 years ago[analyzer] print() JSONify: Store implementation
Csaba Dabis [Wed, 29 May 2019 15:25:19 +0000 (15:25 +0000)]
[analyzer] print() JSONify: Store implementation

Summary: -

Reviewers: NoQ, xazax.hun, ravikandhadai, baloghadamsoftware, Szelethus

Reviewed By: NoQ

Subscribers: szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy,
             dkrupp

Tags: #clang

Differential Revision: https://reviews.llvm.org/D61912

llvm-svn: 361972

5 years ago[NFC][InstCombine] Add unary FNeg tests to fsub.ll known-never-nan.ll
Cameron McInally [Wed, 29 May 2019 15:21:28 +0000 (15:21 +0000)]
[NFC][InstCombine] Add unary FNeg tests to fsub.ll known-never-nan.ll

llvm-svn: 361971

5 years agoAdd additional constraints on midpoint(pointer, pointer). Fixes PR#42037.
Marshall Clow [Wed, 29 May 2019 15:17:55 +0000 (15:17 +0000)]
Add additional constraints on midpoint(pointer, pointer). Fixes PR#42037.

llvm-svn: 361970

5 years ago[clangd] Map typedefs to the same LSP completion kind as VSCode
Ilya Biryukov [Wed, 29 May 2019 15:10:19 +0000 (15:10 +0000)]
[clangd] Map typedefs to the same LSP completion kind as VSCode

For consistency and, more importantly, to get a nicer icon for those in VSCode.

llvm-svn: 361969