platform/upstream/llvm.git
10 years agoPer comments on PR12350, move DR244 back from 'done' to 'partial' and add tests showi...
Richard Smith [Fri, 17 Oct 2014 01:03:17 +0000 (01:03 +0000)]
Per comments on PR12350, move DR244 back from 'done' to 'partial' and add tests showing what we get wrong.

llvm-svn: 220009

10 years ago[OCaml] Add Llvm.instr_clone.
Peter Zotov [Fri, 17 Oct 2014 01:02:40 +0000 (01:02 +0000)]
[OCaml] Add Llvm.instr_clone.

llvm-svn: 220008

10 years ago[LLVM-C] Add LLVMInstructionClone.
Peter Zotov [Fri, 17 Oct 2014 01:02:34 +0000 (01:02 +0000)]
[LLVM-C] Add LLVMInstructionClone.

llvm-svn: 220007

10 years agoCodeGen: Kill FillInNullDataMemberPointers
David Majnemer [Fri, 17 Oct 2014 01:00:43 +0000 (01:00 +0000)]
CodeGen: Kill FillInNullDataMemberPointers

It exists to handle the case where base subobjects are character arrays.
This never happens.

llvm-svn: 220006

10 years agoAST: Remove dead code from RecordLayoutBuilder
David Majnemer [Fri, 17 Oct 2014 01:00:41 +0000 (01:00 +0000)]
AST: Remove dead code from RecordLayoutBuilder

No functionality change intended.

llvm-svn: 220005

10 years ago[llvm-symbolizer] Introduce the -dsym-hint option.
Alexander Potapenko [Fri, 17 Oct 2014 00:50:19 +0000 (00:50 +0000)]
[llvm-symbolizer] Introduce the -dsym-hint option.

llvm-symbolizer will consult one of the .dSYM paths passed via -dsym-hint
if it fails to find the .dSYM bundle at the default location.

llvm-svn: 220004

10 years agoAdd a few more bug IDs for x'fail test cases
Enrico Granata [Fri, 17 Oct 2014 00:47:44 +0000 (00:47 +0000)]
Add a few more bug IDs for x'fail test cases

llvm-svn: 220003

10 years agotrying to fix the new test again, this time for the clang-cmake-armv7-a15 bot
Kostya Serebryany [Fri, 17 Oct 2014 00:47:30 +0000 (00:47 +0000)]
trying to fix the new test again, this time for the clang-cmake-armv7-a15 bot

llvm-svn: 220002

10 years agoMade multi-line test case actions possible in
Sean Callanan [Fri, 17 Oct 2014 00:39:37 +0000 (00:39 +0000)]
Made multi-line test case actions possible in
the inline test cases.  This makes them much
more readable.

llvm-svn: 220001

10 years ago[CMake] lld: Introduce ${cmake_2_8_12_INTERFACE} onto each target_link_libraries...
NAKAMURA Takumi [Fri, 17 Oct 2014 00:37:49 +0000 (00:37 +0000)]
[CMake] lld: Introduce ${cmake_2_8_12_INTERFACE} onto each target_link_libraries. [PR20254]

FIXME: Dependencies should be reorganized.
llvm-svn: 220000

10 years agoR600/SI: Simplify debug printing
Matt Arsenault [Fri, 17 Oct 2014 00:36:20 +0000 (00:36 +0000)]
R600/SI: Simplify debug printing

llvm-svn: 219999

10 years ago[libcxx] Fix SFINAE in <cmath>. Patch from K-Ballo.
Eric Fiselier [Fri, 17 Oct 2014 00:31:47 +0000 (00:31 +0000)]
[libcxx] Fix SFINAE in <cmath>. Patch from K-Ballo.

Delay instantiation of `__numeric_type` within <cmath>,
don't instantiate it when the `is_arithmetic` conditions do not hold as it causes
errors with user-defined types with ambiguous conversions. Fixes PR21083.

llvm-svn: 219998

10 years agoSanitizerBlacklist: blacklist functions by their source location.
Alexey Samsonov [Fri, 17 Oct 2014 00:20:19 +0000 (00:20 +0000)]
SanitizerBlacklist: blacklist functions by their source location.

This commit changes the way we blacklist functions in ASan, TSan,
MSan and UBSan. We used to treat function as "blacklisted"
and turned off instrumentation in it in two cases:

1) Function is explicitly blacklisted by its mangled name.
This part is not changed.

2) Function is located in llvm::Module, whose identifier is
contained in the list of blacklisted sources. This is completely
wrong, as llvm::Module may not correspond to the actual source
file function is defined in. Also, function can be defined in
a header, in which case user had to blacklist the .cpp file
this header was #include'd into, not the header itself.
Such functions could cause other problems - for instance, if the
header was included in multiple source files, compiled
separately and linked into a single executable, we could end up
with both instrumented and non-instrumented version of the same
function participating in the same link.

After this change we will make blacklisting decision based on
the SourceLocation of a function definition. If a function is
not explicitly defined in the source file, (for example, the
function is compiler-generated and responsible for
initialization/destruction of a global variable), then it will
be blacklisted if the corresponding global variable is defined
in blacklisted source file, and will be instrumented otherwise.

After this commit, the active users of blacklist files may have
to revisit them. This is a backwards-incompatible change, but
I don't think it's possible or makes sense to support the
old incorrect behavior.

I plan to make similar change for blacklisting GlobalVariables
(which is ASan-specific).

llvm-svn: 219997

10 years agoAppease the buildbots with the special case for non-set variables
Filipe Cabecinhas [Thu, 16 Oct 2014 23:54:26 +0000 (23:54 +0000)]
Appease the buildbots with the special case for non-set variables

llvm-svn: 219994

10 years agoTeach SanitizerBlacklist to blacklist by SourceLocation. NFC.
Alexey Samsonov [Thu, 16 Oct 2014 23:50:26 +0000 (23:50 +0000)]
Teach SanitizerBlacklist to blacklist by SourceLocation. NFC.

llvm-svn: 219993

10 years agoAlways compile debuginfo-tests for the host triple
Filipe Cabecinhas [Thu, 16 Oct 2014 23:43:34 +0000 (23:43 +0000)]
Always compile debuginfo-tests for the host triple

Since these tests expect a working debugger, always compile them for the
host triple, assuming a working debugger is present.

This enables us to compile and run them, even when clang is, by default,
a cross-compiler (but can still target the host).

llvm-svn: 219992

10 years agoAdd our own copy of the find_executable function to cope with installations
Peter Collingbourne [Thu, 16 Oct 2014 23:43:20 +0000 (23:43 +0000)]
Add our own copy of the find_executable function to cope with installations
that do not have the distutils.spawn package. Should hopefully fix the
aarch64 buildbot.

llvm-svn: 219991

10 years ago[mach-o] Fix crash when -all_load used with dylibs
Nick Kledzik [Thu, 16 Oct 2014 23:42:42 +0000 (23:42 +0000)]
[mach-o] Fix crash when -all_load used with dylibs

-all_load tells the darwin linker to immediately load all members of all
archives.  The code do that used reinterpret_cast<> instead of dyn_cast<>.
If the file was a dylib, the reinterpret_cast<> turned a pointer to a dylib
into a pointer to an archive...boom.

Added test case to reproduce the crash, simplified the code and used dyn_cast<>.

llvm-svn: 219990

10 years agoAdded %itanium_abi_host_triple to fix debuginfo-tests when clang is a cross-compiler...
Filipe Cabecinhas [Thu, 16 Oct 2014 23:41:40 +0000 (23:41 +0000)]
Added %itanium_abi_host_triple to fix debuginfo-tests when clang is a cross-compiler by default

llvm-svn: 219989

10 years agoR600/SI: Remove another VALU pattern
Matt Arsenault [Thu, 16 Oct 2014 23:33:37 +0000 (23:33 +0000)]
R600/SI: Remove another VALU pattern

llvm-svn: 219988

10 years ago[libcxx] Add support for LLVM_USE_SANITIZER=Undefined
Eric Fiselier [Thu, 16 Oct 2014 23:21:59 +0000 (23:21 +0000)]
[libcxx] Add support for LLVM_USE_SANITIZER=Undefined

LLVM_USE_SANITIZER=Undefined support was added to the LLVM CMake configuration.
Update libc++'s handling of LLVM_USE_SANITIZER to support this as well.

llvm-svn: 219987

10 years agoAttach a bug number to these failures
Enrico Granata [Thu, 16 Oct 2014 23:17:46 +0000 (23:17 +0000)]
Attach a bug number to these failures

llvm-svn: 219986

10 years agoSplit this test case to handle each literal kind uniquely
Enrico Granata [Thu, 16 Oct 2014 23:16:13 +0000 (23:16 +0000)]
Split this test case to handle each literal kind uniquely

llvm-svn: 219985

10 years agoAdded a new kind of test case: the "inline" test
Sean Callanan [Thu, 16 Oct 2014 23:15:22 +0000 (23:15 +0000)]
Added a new kind of test case: the "inline" test
case.  This test case style attempts to shed all
of the boilerplate that is required for test
cases, and let 80% of test cases use a much terser
syntax.

Inline testcases have much simplified python files
(the corresponding .py file should contain two
lines of code) and require no Makefile, because the
Makefile is generated automatically.  Breakpoints
are set automatically and the indicated breakpoint
actions (specified after a magic //% comment) are
executed when the breakpoint is hit.

All other testcases are unaffected.

One thing I'm not really happy with yet is the way
multiple actions for the same line are specified.
I'm going to use lang/c/struct_types as a guinea
pig to develop this further.

llvm-svn: 219984

10 years agoChange a use of mktemp() to mkstemp() for better security.
Jason Molenda [Thu, 16 Oct 2014 23:10:03 +0000 (23:10 +0000)]
Change a use of mktemp() to mkstemp() for better security.
We have two more uses of mktemp still in the source base
but they'll take a little more consideration.
clang static analyzer fixit.

llvm-svn: 219983

10 years agoThe number '5' triggers a bug unrelated to LLDB, and is not instrumental to this...
Enrico Granata [Thu, 16 Oct 2014 23:06:40 +0000 (23:06 +0000)]
The number '5' triggers a bug unrelated to LLDB, and is not instrumental to this test in any way. Use another, randomly chosen, number to make the test pass again and provide useful actionable feedback about things that truly matter

llvm-svn: 219982

10 years agoSome more test marking
Enrico Granata [Thu, 16 Oct 2014 23:03:06 +0000 (23:03 +0000)]
Some more test marking

llvm-svn: 219981

10 years agoThis test needs the SB headers to make sense
Enrico Granata [Thu, 16 Oct 2014 23:02:45 +0000 (23:02 +0000)]
This test needs the SB headers to make sense

llvm-svn: 219980

10 years agoAdd a test for the -b (batch mode) option to the lldb driver.
Jim Ingham [Thu, 16 Oct 2014 23:02:14 +0000 (23:02 +0000)]
Add a test for the -b (batch mode) option to the lldb driver.

llvm-svn: 219979

10 years agoFix this test case to actually work - it was relying on a certain 'po' output which...
Enrico Granata [Thu, 16 Oct 2014 23:02:03 +0000 (23:02 +0000)]
Fix this test case to actually work - it was relying on a certain 'po' output which wasn't occurring

llvm-svn: 219978

10 years agoRe-commit r217995 and follow-up patches (r217997, r218011, r218053). These were
Richard Smith [Thu, 16 Oct 2014 23:00:46 +0000 (23:00 +0000)]
Re-commit r217995 and follow-up patches (r217997, r218011, r218053). These were
reverted in r218058 because they triggered a rejects-valid bug in MSVC.

Original commit message from r217995:

Instantiate exception specifications when instantiating function types (other
than the type of a function declaration). We previously didn't instantiate
these at all! This also covers the pathological case where the only mention of
a parameter pack is within the exception specification; this gives us a second
way (other than alias templates) to reach the horrible state where a type
contains an unexpanded pack, but its canonical type does not.

llvm-svn: 219977

10 years agoInitial version of Go bindings.
Peter Collingbourne [Thu, 16 Oct 2014 22:48:02 +0000 (22:48 +0000)]
Initial version of Go bindings.

This code is based on the existing LLVM Go bindings project hosted at:
https://github.com/go-llvm/llvm

Note that all contributors to the gollvm project have agreed to relicense
their changes under the LLVM license and submit them to the LLVM project.

Differential Revision: http://reviews.llvm.org/D5684

llvm-svn: 219976

10 years agoIntroduce LLVMParseCommandLineOptions C API function.
Peter Collingbourne [Thu, 16 Oct 2014 22:47:52 +0000 (22:47 +0000)]
Introduce LLVMParseCommandLineOptions C API function.

llvm-svn: 219975

10 years agoSema: handle additional case of qualified types
Saleem Abdulrasool [Thu, 16 Oct 2014 22:42:53 +0000 (22:42 +0000)]
Sema: handle additional case of qualified types

A second instance of attributed types escaped the previous change, identified
thanks to Richard Smith!  When deducing the void case, we would also assume that
the type would not be attributed.  Furthermore, properly handle multiple
attributes being applied to a single TypeLoc.

Properly handle this case and future-proof a bit by ignoring parenthesis
further.  The test cases do use the additional parenthesis to ensure that this
case remains properly handled.

Addresses post-commit review comments from Richard Smith to SVN r219851.

llvm-svn: 219974

10 years agoSome more failure to bug tracking
Enrico Granata [Thu, 16 Oct 2014 22:27:17 +0000 (22:27 +0000)]
Some more failure to bug tracking

llvm-svn: 219973

10 years agoAssociate a bug tracking ID with this test case
Enrico Granata [Thu, 16 Oct 2014 22:06:26 +0000 (22:06 +0000)]
Associate a bug tracking ID with this test case

llvm-svn: 219972

10 years agoThis test actually works alright - we were just checking for the wrong string
Enrico Granata [Thu, 16 Oct 2014 22:04:05 +0000 (22:04 +0000)]
This test actually works alright - we were just checking for the wrong string

llvm-svn: 219971

10 years agoI see this test case crash - skip for now
Enrico Granata [Thu, 16 Oct 2014 21:42:34 +0000 (21:42 +0000)]
I see this test case crash - skip for now

llvm-svn: 219970

10 years agotest/CodeGen/sections.c: add triple
Hans Wennborg [Thu, 16 Oct 2014 21:36:23 +0000 (21:36 +0000)]
test/CodeGen/sections.c: add triple

llvm-svn: 219969

10 years agoReduce code duplication between patchpoint and non-patchpoint lowering. NFC.
Juergen Ributzka [Thu, 16 Oct 2014 21:26:35 +0000 (21:26 +0000)]
Reduce code duplication between patchpoint and non-patchpoint lowering. NFC.

This is in preparation for another patch that makes patchpoints invokable.

Reviewers: atrick, ributzka
Reviewed By: ributzka
Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5657

llvm-svn: 219967

10 years agoOnly call RemovePersistentVariable on expr_result if that shared
Jason Molenda [Thu, 16 Oct 2014 21:25:43 +0000 (21:25 +0000)]
Only call RemovePersistentVariable on expr_result if that shared
pointer contains something.

llvm-svn: 219966

10 years agotrying to fix the new test on hexagon-build
Kostya Serebryany [Thu, 16 Oct 2014 21:22:40 +0000 (21:22 +0000)]
trying to fix the new test on hexagon-build

llvm-svn: 219965

10 years agoRework this code so that it does not trigger a compiler warning. NFC
Enrico Granata [Thu, 16 Oct 2014 21:18:58 +0000 (21:18 +0000)]
Rework this code so that it does not trigger a compiler warning. NFC

llvm-svn: 219964

10 years ago[SROA] Switch the common variable name for the 'AllocaSlices' class to
Chandler Carruth [Thu, 16 Oct 2014 21:11:55 +0000 (21:11 +0000)]
[SROA] Switch the common variable name for the 'AllocaSlices' class to
'AS'.

Using 'S' as this was a terrible idea. Arguably, 'AS' is not much
better, but it at least follows the idea of using initialisms and
removes active confusion about the AllocaSlices variable and a Slice
variable.

llvm-svn: 219963

10 years ago[SROA] More range-based cleanups to SROA, these brought to you by
Chandler Carruth [Thu, 16 Oct 2014 21:05:14 +0000 (21:05 +0000)]
[SROA] More range-based cleanups to SROA, these brought to you by
clang-modernize.

I did have to clean up the variable types and whitespace a bit because
the use of auto made the code much less readable here.

llvm-svn: 219962

10 years agoInsert poisoned paddings between fields in C++ classes so that AddressSanitizer can...
Kostya Serebryany [Thu, 16 Oct 2014 20:54:52 +0000 (20:54 +0000)]
Insert poisoned paddings between fields in C++ classes so that AddressSanitizer can find intra-object-overflow bugs

Summary:
The general approach is to add extra paddings after every field
in AST/RecordLayoutBuilder.cpp, then add code to CTORs/DTORs that poisons the paddings
(CodeGen/CGClass.cpp).

Everything is done under the flag -fsanitize-address-field-padding.
The blacklist file (-fsanitize-blacklist) allows to avoid the transformation
for given classes or source files.

See also https://code.google.com/p/address-sanitizer/wiki/IntraObjectOverflow

Test Plan: run SPEC2006 and some of the Chromium tests with  -fsanitize-address-field-padding

Reviewers: samsonov, rnk, rsmith

Reviewed By: rsmith

Subscribers: majnemer, cfe-commits

Differential Revision: http://reviews.llvm.org/D5687

llvm-svn: 219961

10 years agoMS Compat: mark globals emitted in read-only sections const
Hans Wennborg [Thu, 16 Oct 2014 20:52:46 +0000 (20:52 +0000)]
MS Compat: mark globals emitted in read-only sections const

They cannot be written to, so marking them const makes sense and may improve
optimisation.

As a side-effect, SectionInfos has to be moved from Sema to ASTContext.

It also fixes this problem, that occurs when compiling ATL:

  warning LNK4254: section 'ATL' (C0000040) merged into '.rdata' (40000040) with different attributes

The ATL headers are putting variables in a special section that's marked
read-only. However, Clang currently can't model that read-onlyness in the IR.
But, by making the variables const, the section does become read-only, and
the linker warning is avoided.

Differential Revision: http://reviews.llvm.org/D5812

llvm-svn: 219960

10 years ago[mach-o] update __eh_frame handling for Nick's suggestions
Tim Northover [Thu, 16 Oct 2014 20:52:18 +0000 (20:52 +0000)]
[mach-o] update __eh_frame handling for Nick's suggestions

First, add a comment to support more variation in FDE formats. Second, refactor
fde -> function handling into a separate function living in the ArchHandler.

llvm-svn: 219959

10 years ago[SROA] Switch a couple of overly complex iterator accessors to just be
Chandler Carruth [Thu, 16 Oct 2014 20:42:08 +0000 (20:42 +0000)]
[SROA] Switch a couple of overly complex iterator accessors to just be
ArrayRef accessors.

I think this even came up in review that this was over-engineered, and
indeed it was. Time to un-build it.

llvm-svn: 219958

10 years agoErase fence insertion from SelectionDAGBuilder.cpp (NFC)
Robin Morisset [Thu, 16 Oct 2014 20:34:57 +0000 (20:34 +0000)]
Erase fence insertion from SelectionDAGBuilder.cpp (NFC)

Summary:
Backends can use setInsertFencesForAtomic to signal to the middle-end that
montonic is the only memory ordering they can accept for
stores/loads/rmws/cmpxchg. The code lowering those accesses with a stronger
ordering to fences + monotonic accesses is currently living in
SelectionDAGBuilder.cpp. In this patch I propose moving this logic out of it
for several reasons:
- There is lots of redundancy to avoid: extremely similar logic already
  exists in AtomicExpand.
- The current code in SelectionDAGBuilder does not use any target-hooks, it
  does the same transformation for every backend that requires it
- As a result it is plain *unsound*, as it was apparently designed for ARM.
  It happens to mostly work for the other targets because they are extremely
  conservative, but Power for example had to switch to AtomicExpand to be
  able to use lwsync safely (see r218331).
- Because it produces IR-level fences, it cannot be made sound ! This is noted
  in the C++11 standard (section 29.3, page 1140):
```
Fences cannot, in general, be used to restore sequential consistency for atomic
operations with weaker ordering semantics.
```
It can also be seen by the following example (called IRIW in the litterature):
```
atomic<int> x = y = 0;
int r1, r2, r3, r4;
Thread 0:
  x.store(1);
Thread 1:
  y.store(1);
Thread 2:
  r1 = x.load();
  r2 = y.load();
Thread 3:
  r3 = y.load();
  r4 = x.load();
```
r1 = r3 = 1 and r2 = r4 = 0 is impossible as long as the accesses are all seq_cst.
But if they are lowered to monotonic accesses, no amount of fences can prevent it..

This patch does three things (I could cut it into parts, but then some of them
would not be tested/testable, please tell me if you would prefer that):
- it provides a default implementation for emitLeadingFence/emitTrailingFence in
terms of IR-level fences, that mimic the original logic of SelectionDAGBuilder.
As we saw above, this is unsound, but the best that can be done without knowing
the targets well (and there is a comment warning about this risk).
- it then switches Mips/Sparc/XCore to use AtomicExpand, relying on this default
implementation (that exactly replicates the logic of SelectionDAGBuilder, so no
functional change)
- it finally erase this logic from SelectionDAGBuilder as it is dead-code.

Ideally, each target would define its own override for emitLeading/TrailingFence
using target-specific fences, but I do not know the Sparc/Mips/XCore memory model
well enough to do this, and they appear to be dealing fine with the ARM-inspired
default expansion for now (probably because they are overly conservative, as
Power was). If anyone wants to compile fences more agressively on these
platforms, the long comment should make it clear why he should first override
emitLeading/TrailingFence.

Test Plan: make check-all, no functional change

Reviewers: jfb, t.p.northover

Subscribers: aemerson, llvm-commits

Differential Revision: http://reviews.llvm.org/D5474

llvm-svn: 219957

10 years agoR600/SI: Remove unnecessary VALU patterns
Matt Arsenault [Thu, 16 Oct 2014 20:31:50 +0000 (20:31 +0000)]
R600/SI: Remove unnecessary VALU patterns

These haven't been necessary since allowing
selecting SALU instructions in non-entry blocks
was enabled.

llvm-svn: 219956

10 years ago[SROA] Start more deeply moving SROA to use ranges rather than just
Chandler Carruth [Thu, 16 Oct 2014 20:24:07 +0000 (20:24 +0000)]
[SROA] Start more deeply moving SROA to use ranges rather than just
iterators.

There are a ton of places where it essentially wants ranges
rather than just iterators. This is just the first step that adds the
core slice range typedefs and uses them in a couple of places. I still
have to explicitly construct them because they've not been punched
throughout the entire set of code. More range-based cleanups incoming.

llvm-svn: 219955

10 years agoNo longer emit diagnostics about unused results (comparisons, etc) from unevaluated...
Aaron Ballman [Thu, 16 Oct 2014 20:13:28 +0000 (20:13 +0000)]
No longer emit diagnostics about unused results (comparisons, etc) from unevaluated contexts. Fixes PR18571.

llvm-svn: 219954

10 years agoR600: Fix nonsensical implementation of computeKnownBits for BFE
Matt Arsenault [Thu, 16 Oct 2014 20:07:40 +0000 (20:07 +0000)]
R600: Fix nonsensical implementation of computeKnownBits for BFE

This was resulting in invalid simplifications of sdiv

llvm-svn: 219953

10 years agoUpdate for llvm change.
Rafael Espindola [Thu, 16 Oct 2014 20:00:22 +0000 (20:00 +0000)]
Update for llvm change.

llvm-svn: 219952

10 years agoDelete -std-compile-opts.
Rafael Espindola [Thu, 16 Oct 2014 20:00:02 +0000 (20:00 +0000)]
Delete -std-compile-opts.

These days -std-compile-opts was just a silly alias for -O3.

llvm-svn: 219951

10 years agoAllow call-slop optzn for destinations with a suitable dereferenceable attribute
Bjorn Steinbrink [Thu, 16 Oct 2014 19:43:08 +0000 (19:43 +0000)]
Allow call-slop optzn for destinations with a suitable dereferenceable attribute

Summary:
Currently, call slot optimization requires that if the destination is an
argument, the argument has the sret attribute. This is to ensure that
the memory access won't trap. In addition to sret, we can also allow the
optimization to happen for arguments that have the new dereferenceable
attribute, which gives the same guarantee.

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5832

llvm-svn: 219950

10 years ago[mach-o] Add support for upward linking
Nick Kledzik [Thu, 16 Oct 2014 19:31:28 +0000 (19:31 +0000)]
[mach-o] Add support for upward linking

To deal with cycles in shared library dependencies, the darwin linker supports
marking specific link dependencies as "upward".  An upward link is when a
lower level library links against a higher level library.

llvm-svn: 219949

10 years ago[PECOFF] Support delay-load import table for x86
Rui Ueyama [Thu, 16 Oct 2014 19:30:44 +0000 (19:30 +0000)]
[PECOFF] Support delay-load import table for x86

This patch creates the import address table and sets its
address to the delay-load import table. This also creates
wrapper functions for __delayLoadHelper2.

x86 only for now.

llvm-svn: 219948

10 years agoFix lang-ref doc bug: s/icmp lt/icmp slt/
Jonathan Roelofs [Thu, 16 Oct 2014 19:28:10 +0000 (19:28 +0000)]
Fix lang-ref doc bug: s/icmp lt/icmp slt/

llvm-svn: 219947

10 years ago[asan] make sure coverage is dumped even if leaks are reported
Kostya Serebryany [Thu, 16 Oct 2014 18:59:07 +0000 (18:59 +0000)]
[asan] make sure coverage is dumped even if leaks are reported

llvm-svn: 219946

10 years ago[llvm-objdump] Fix -private-headers for mach-o to print all LC_*_DYLIB variants
Nick Kledzik [Thu, 16 Oct 2014 18:58:20 +0000 (18:58 +0000)]
[llvm-objdump] Fix -private-headers for mach-o to print all LC_*_DYLIB variants

llvm-svn: 219945

10 years agofold: sqrt(x * x * y) -> fabs(x) * sqrt(y)
Sanjay Patel [Thu, 16 Oct 2014 18:48:17 +0000 (18:48 +0000)]
fold: sqrt(x * x * y) -> fabs(x) * sqrt(y)

If a square root call has an FP multiplication argument that can be reassociated,
then we can hoist a repeated factor out of the square root call and into a fabs().

In the simplest case, this:

   y = sqrt(x * x);

becomes this:

   y = fabs(x);

This patch relies on an earlier optimization in instcombine or reassociate to put the
multiplication tree into a canonical form, so we don't have to search over
every permutation of the multiplication tree.

Because there are no IR-level FastMathFlags for intrinsics (PR21290), we have to
use function-level attributes to do this optimization. This needs to be fixed
for both the intrinsics and in the backend.

Differential Revision: http://reviews.llvm.org/D5787

llvm-svn: 219944

10 years agoDebugInfo: Follow up to r219736, also test/demonstrate that we emit the constant...
David Blaikie [Thu, 16 Oct 2014 18:38:36 +0000 (18:38 +0000)]
DebugInfo: Follow up to r219736, also test/demonstrate that we emit the constant value in this case as well.

llvm-svn: 219943

10 years agoFixing comment grammar; NFC.
Aaron Ballman [Thu, 16 Oct 2014 18:09:29 +0000 (18:09 +0000)]
Fixing comment grammar; NFC.

llvm-svn: 219942

10 years agoSpeculatively fix GCC 4.7 build after r219938
Hans Wennborg [Thu, 16 Oct 2014 17:57:41 +0000 (17:57 +0000)]
Speculatively fix GCC 4.7 build after r219938

llvm-svn: 219941

10 years agoSwitching to range-based for loops; NFC.
Aaron Ballman [Thu, 16 Oct 2014 17:53:07 +0000 (17:53 +0000)]
Switching to range-based for loops; NFC.

llvm-svn: 219940

10 years agoFix code to follow the "Don’t use else after a return" rule.
Samuel Benzaquen [Thu, 16 Oct 2014 17:50:19 +0000 (17:50 +0000)]
Fix code to follow the "Don’t use else after a return" rule.

Summary:
Fix code to follow the "Don’t use else after a return" rule.
This is a followup from rL219792.

Reviewers: alexfh

Subscribers: klimek, cfe-commits

Differential Revision: http://reviews.llvm.org/D5826

llvm-svn: 219939

10 years agoUse iterators and algorithms to possibly make this code a bit tidier
David Blaikie [Thu, 16 Oct 2014 17:23:58 +0000 (17:23 +0000)]
Use iterators and algorithms to possibly make this code a bit tidier

(also, the code executed once the element was found was split half
inside the loop and half after it - now put it all together after the
find operation)

I'm a bit concerned that this code is rather untested (commenting out
this whole function and running check-clang doesn't fail any tests)...

And I wish I had polymorphic lambdas.

llvm-svn: 219938

10 years agoRemove one of SanitizerBlacklist::isIn() overloads. NFC.
Alexey Samsonov [Thu, 16 Oct 2014 17:10:38 +0000 (17:10 +0000)]
Remove one of SanitizerBlacklist::isIn() overloads. NFC.

The final goal is to get rid of all the rest overloads that
accept LLVM objects (llvm::Function and llvm::GlobalVariable),
and pass in source-level entities instead.

llvm-svn: 219937

10 years agoFix MemoryHistory plugin to check whether the plugin
Jason Molenda [Thu, 16 Oct 2014 16:59:23 +0000 (16:59 +0000)]
Fix MemoryHistory plugin to check whether the plugin
was able to create itself before returning the shared
pointer to it.
clang warning.

llvm-svn: 219936

10 years agoUse array_lengthof; NFC.
Hans Wennborg [Thu, 16 Oct 2014 16:54:36 +0000 (16:54 +0000)]
Use array_lengthof; NFC.

llvm-svn: 219935

10 years ago[AArch64] Fix miscompile of sdiv-by-power-of-2.
Juergen Ributzka [Thu, 16 Oct 2014 16:41:15 +0000 (16:41 +0000)]
[AArch64] Fix miscompile of sdiv-by-power-of-2.

When the constant divisor was larger than 32bits, then the optimized code
generated for the AArch64 backend would emit the wrong code, because the shift
was defined as a shift of a 32bit constant '(1<<Lg2(divisor))' and we would
loose the upper 32bits.

This fixes rdar://problem/18678801.

llvm-svn: 219934

10 years ago[AArch64] Enable A53 erratum workaround (835769) by default for Android targets
Bradley Smith [Thu, 16 Oct 2014 16:35:14 +0000 (16:35 +0000)]
[AArch64] Enable A53 erratum workaround (835769) by default for Android targets

llvm-svn: 219933

10 years agotests: move test to more appropriate location
Saleem Abdulrasool [Thu, 16 Oct 2014 16:12:41 +0000 (16:12 +0000)]
tests: move test to more appropriate location

The test is a C++ semantic analysis test, move it to SemaCXX from Sema.  NFC.

llvm-svn: 219932

10 years ago[mips] Account for endianess when expanding BuildPairF64/ExtractElementF64 nodes.
Vasileios Kalintiris [Thu, 16 Oct 2014 15:41:51 +0000 (15:41 +0000)]
[mips] Account for endianess when expanding BuildPairF64/ExtractElementF64 nodes.

Summary:
In order to support big endian targets for the BuildPairF64 nodes we
just need to swap the low/high pair registers. Additionally, for the
ExtractElementF64 nodes we have to calculate the correct stack offset
with respect to the node's register/operand that we want to extract.

Reviewers: dsanders

Reviewed By: dsanders

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5753

llvm-svn: 219931

10 years agoOpenCL: Emit global variables in the constant addr space as constant globals
Tom Stellard [Thu, 16 Oct 2014 15:29:19 +0000 (15:29 +0000)]
OpenCL: Emit global variables in the constant addr space as constant globals

llvm-svn: 219929

10 years agoOpenCL: Add -ffake-address-space-map to a test
Tom Stellard [Thu, 16 Oct 2014 15:29:17 +0000 (15:29 +0000)]
OpenCL: Add -ffake-address-space-map to a test

The ensures there is an explicit address space id in the output.

llvm-svn: 219928

10 years ago[mips] Marked the DI/EI instruction aliases as MIPS32r2
Vasileios Kalintiris [Thu, 16 Oct 2014 15:23:52 +0000 (15:23 +0000)]
[mips] Marked the DI/EI instruction aliases as MIPS32r2

Reviewers: dsanders

Reviewed By: dsanders

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5751

llvm-svn: 219927

10 years ago[clang-tidy] Minor fixes for the NamespaceCommentCheck.
Alexander Kornienko [Thu, 16 Oct 2014 15:11:54 +0000 (15:11 +0000)]
[clang-tidy] Minor fixes for the NamespaceCommentCheck.

  * Make SmallVector size enough for all groups.
  * Allow trailing period in the comment.
  * Fix "// anonymous namespace qqq".

llvm-svn: 219926

10 years agoTest commit access: remove extra new line at the end of file
Vasileios Kalintiris [Thu, 16 Oct 2014 14:37:00 +0000 (14:37 +0000)]
Test commit access: remove extra new line at the end of file

llvm-svn: 219925

10 years agoUBSAN stable runtime more generic with arm* targets
Renato Golin [Thu, 16 Oct 2014 12:43:10 +0000 (12:43 +0000)]
UBSAN stable runtime more generic with arm* targets

llvm-svn: 219924

10 years ago[clang-tidy] Default options in modules.
Alexander Kornienko [Thu, 16 Oct 2014 11:27:57 +0000 (11:27 +0000)]
[clang-tidy] Default options in modules.

Summary:
This patch allows modules to specify default options for the checks
defined in them. This way a sufficiently configurable check can be registered in
multiple modules with different default options. E.g. the SpacesBeforeComments
option may be set to 1 for the "llvm-namespace-comments" check and to 2 for the
"google-readability-namespace-comment" check without modifying or extending the
check code.

This patch also registers the google-readability-braces-around-statements check
with suitable defaults.

Reviewers: djasper

Reviewed By: djasper

Subscribers: curdeius, cfe-commits

Differential Revision: http://reviews.llvm.org/D5798

llvm-svn: 219923

10 years agoAdd missing header guard.
Benjamin Kramer [Thu, 16 Oct 2014 10:10:07 +0000 (10:10 +0000)]
Add missing header guard.

llvm-svn: 219922

10 years agoclang-format: Fix behavior with comments before conditional expressions
Daniel Jasper [Thu, 16 Oct 2014 09:10:11 +0000 (09:10 +0000)]
clang-format: Fix behavior with comments before conditional expressions

Before:
  SomeFunction(aaaaaaaaaaaaaaaaa,
               // comment.
                   ccccccccccccccccc ? aaaaaaaaaaaaaaaaaaaa
                                     : bbbbbbbbbbbbbbbbbbbb);

After:
  SomeFunction(aaaaaaaaaaaaaaaaa,
               // comment.
               ccccccccccccccccc ? aaaaaaaaaaaaaaaaaaaa : bbbbbbbbbbbbbbbbbbbb);

llvm-svn: 219921

10 years agoMost of this function checks to see if m_process is non-null before
Jason Molenda [Thu, 16 Oct 2014 08:43:27 +0000 (08:43 +0000)]
Most of this function checks to see if m_process is non-null before
dereferencing it, except for this one section of code.  Add a null
check around it.
clang static analyzer fix.

llvm-svn: 219920

10 years agoclang-format: [ObjC] Fix method expression detection.
Daniel Jasper [Thu, 16 Oct 2014 08:38:51 +0000 (08:38 +0000)]
clang-format: [ObjC] Fix method expression detection.

Before:
  return (a)[foo bar : baz];

After:
  return (a)[foo bar:baz];

llvm-svn: 219919

10 years agoAnother logical-or vrs. bitwise-or mixup in ClangUserExpression.
Jason Molenda [Thu, 16 Oct 2014 08:27:27 +0000 (08:27 +0000)]
Another logical-or vrs. bitwise-or mixup in ClangUserExpression.
clang unreachable code warning.

llvm-svn: 219918

10 years agoEnable warnings in the debugserver project file..
Jason Molenda [Thu, 16 Oct 2014 08:16:29 +0000 (08:16 +0000)]
Enable warnings in the debugserver project file..

llvm-svn: 219917

10 years agoA series of bit-flag values should be bitwise-or'ed not logical-or'ed.
Jason Molenda [Thu, 16 Oct 2014 08:15:11 +0000 (08:15 +0000)]
A series of bit-flag values should be bitwise-or'ed not logical-or'ed.
clang unreachable code warning.

llvm-svn: 219916

10 years agoRemove unreachable code.
Jason Molenda [Thu, 16 Oct 2014 08:08:13 +0000 (08:08 +0000)]
Remove unreachable code.

llvm-svn: 219915

10 years agoRemove unreachable code.
Jason Molenda [Thu, 16 Oct 2014 08:07:54 +0000 (08:07 +0000)]
Remove unreachable code.

llvm-svn: 219914

10 years agoAdd /* DISABLES CODE */ annotation before if (0) to mark it as intentional.
Jason Molenda [Thu, 16 Oct 2014 08:07:20 +0000 (08:07 +0000)]
Add /* DISABLES CODE */ annotation before if (0) to mark it as intentional.

llvm-svn: 219913

10 years agoRemove unreachable code.
Jason Molenda [Thu, 16 Oct 2014 08:05:49 +0000 (08:05 +0000)]
Remove unreachable code.

llvm-svn: 219912

10 years agoRemove unreachable code.
Jason Molenda [Thu, 16 Oct 2014 08:05:22 +0000 (08:05 +0000)]
Remove unreachable code.

llvm-svn: 219911

10 years agoXcode recommended these changes to the project file. It recommended
Jason Molenda [Thu, 16 Oct 2014 08:04:28 +0000 (08:04 +0000)]
Xcode recommended these changes to the project file.  It recommended
a number of warnings to be enabled.  The one making the most noise
across the code base right now is CLANG_WARN_UNREACHABLE_CODE = YES.

llvm-svn: 219910

10 years agoRemove unused initialization.
Jason Molenda [Thu, 16 Oct 2014 07:53:46 +0000 (07:53 +0000)]
Remove unused initialization.
clang static analyzer fixit.

llvm-svn: 219909

10 years agoRemove dead store.
Jason Molenda [Thu, 16 Oct 2014 07:52:17 +0000 (07:52 +0000)]
Remove dead store.
clang static analyzer fixit.

llvm-svn: 219908

10 years agoRemove dead store.
Jason Molenda [Thu, 16 Oct 2014 07:49:27 +0000 (07:49 +0000)]
Remove dead store.
clang static analyzer fixit.

llvm-svn: 219907

10 years agoAdds stable-runtime to ubsan to avoid broken ARM tests with asan
Renato Golin [Thu, 16 Oct 2014 07:48:27 +0000 (07:48 +0000)]
Adds stable-runtime to ubsan to avoid broken ARM tests with asan

llvm-svn: 219906