platform/upstream/llvm.git
9 years agoFix several test failures on Linux/FreeBSD caused by compiler configuration and inval...
Oleksiy Vyalov [Mon, 1 Dec 2014 23:21:18 +0000 (23:21 +0000)]
Fix several test failures on Linux/FreeBSD caused by compiler configuration and invalid environment - make lldbtest.registerSharedLibrariesWithTarget to support multiple platforms.
http://reviews.llvm.org/D6392

llvm-svn: 223091

9 years agoPerform correct lookup when '__super' is used in class with dependent base.
Nikola Smiljanic [Mon, 1 Dec 2014 23:15:01 +0000 (23:15 +0000)]
Perform correct lookup when '__super' is used in class with dependent base.

llvm-svn: 223090

9 years agoEnable enabling and disabling breakpoints on Windows.
Zachary Turner [Mon, 1 Dec 2014 23:13:52 +0000 (23:13 +0000)]
Enable enabling and disabling breakpoints on Windows.

llvm-svn: 223089

9 years agoLink test executables with LLD on Windows.
Zachary Turner [Mon, 1 Dec 2014 23:13:41 +0000 (23:13 +0000)]
Link test executables with LLD on Windows.

Clang generates DWARF debug info with section names longer
than 8 characters.  This is only supported by an extension
to PE/COFF which the MSVC linker doesn't implement.  So
trying to link DWARF-enabled object files with MSVC will
lead to corrupt debug information in the executable.

llvm-svn: 223088

9 years agoOnly normalize FileSpec paths *after* resolving them.
Zachary Turner [Mon, 1 Dec 2014 23:13:32 +0000 (23:13 +0000)]
Only normalize FileSpec paths *after* resolving them.

Normalizing paths before resolving them can cause the path to
become denormalized after resolution.

llvm-svn: 223087

9 years agoMake -fuse-ld=lld work properly on Windows.
Zachary Turner [Mon, 1 Dec 2014 23:06:47 +0000 (23:06 +0000)]
Make -fuse-ld=lld work properly on Windows.

Using lld on Windows requires calling link-lld.exe instead of
lld.exe.  This patch puts this knowledge into clang so that when
using the GCC style clang driver, it can properly delegate to
lld.

Differential Revision: http://reviews.llvm.org/D6428
Reviewed by: Reid Kleckner, Rui Ueyama

llvm-svn: 223086

9 years ago[Statepoints 2/4] Statepoint infrastructure for garbage collection: MI & x86-64 Backend
Philip Reames [Mon, 1 Dec 2014 22:52:56 +0000 (22:52 +0000)]
[Statepoints 2/4] Statepoint infrastructure for garbage collection: MI & x86-64 Backend

This is the second patch in a small series.  This patch contains the MachineInstruction and x86-64 backend pieces required to lower Statepoints.  It does not include the code to actually generate the STATEPOINT machine instruction and as a result, the entire patch is currently dead code.  I will be submitting the SelectionDAG parts within the next 24-48 hours.  Since those pieces are by far the most complicated, I wanted to minimize the size of that patch.  That patch will include the tests which exercise the functionality in this patch.  The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.

The STATEPOINT psuedo node is generated after all gc values are explicitly spilled to stack slots.  The purpose of this node is to wrap an actual call instruction while recording the spill locations of the meta arguments used for garbage collection and other purposes.  The STATEPOINT is modeled as modifing all of those locations to prevent backend optimizations from forwarding the value from before the STATEPOINT to after the STATEPOINT.  (Doing so would break relocation semantics for collectors which wish to relocate roots.)

The implementation of STATEPOINT is closely modeled on PATCHPOINT.  Eventually, much of the code in this patch will be removed.  The long term plan is to merge the functionality provided by statepoints and patchpoints.  Merging their implementations in the backend is likely to be a good starting point.

Reviewed by: atrick, ributzka

llvm-svn: 223085

9 years agoThe register keyword is deprecated in C++11, so clang complains strongly about swig...
Enrico Granata [Mon, 1 Dec 2014 22:51:03 +0000 (22:51 +0000)]
The register keyword is deprecated in C++11, so clang complains strongly about swig generating register declarations. Abuse the preprocessor to define the keyword away

llvm-svn: 223084

9 years agolldb can deadlock when launched with an non-existing executable:
Greg Clayton [Mon, 1 Dec 2014 22:41:27 +0000 (22:41 +0000)]
lldb can deadlock when launched with an non-existing executable:

% lldb /bin/nonono
(lldb) target create "/bin/nonono"
error: unable to find executable for '/usr/bin/nonono'
<deadlock>

The problem was the initial commands 'target create "/bin/nonono"' were put into a pipe and the command interpreter was being run with:

void
CommandInterpreter::RunCommandInterpreter(bool auto_handle_events,
                                          bool spawn_thread,
                                          CommandInterpreterRunOptions &options)
{
    // Always re-create the command intepreter when we run it in case
    // any file handles have changed.
    bool force_create = true;
    m_debugger.PushIOHandler(GetIOHandler(force_create, &options));
    m_stopped_for_crash = false;

    if (auto_handle_events)
        m_debugger.StartEventHandlerThread();

    if (spawn_thread)
    {
        m_debugger.StartIOHandlerThread();
    }
    else
    {
        m_debugger.ExecuteIOHanders();

        if (auto_handle_events)
            m_debugger.StopEventHandlerThread();
    }

}

If "auto_handle_events" was set to true and "spawn_thread" was false, we would execute:

m_debugger.StartEventHandlerThread();
m_debugger.ExecuteIOHanders();
m_debugger.StopEventHandlerThread();

The problem was there was no synchonization in Debugger::StartEventHandlerThread() to ensure the event handler was listening to events and the the call to "m_debugger.StopEventHandlerThread()" would do:

void
Debugger::StopEventHandlerThread()
{
    if (m_event_handler_thread.IsJoinable())
    {
        GetCommandInterpreter().BroadcastEvent(CommandInterpreter::eBroadcastBitQuitCommandReceived);
        m_event_handler_thread.Join(nullptr);
    }
}

The problem was that the event thread might not be listening for the CommandInterpreter::eBroadcastBitQuitCommandReceived event yet.

The solution is to make sure the Debugger::DefaultEventHandler() is listening to events before we return from Debugger::StartEventHandlerThread(). Once we have this synchonization we remove the race condition.

This fixes radar:

<rdar://problem/19041192>

llvm-svn: 223083

9 years agoAdded a new regular expression to the "_regexp-break" command ("b" by default):
Greg Clayton [Mon, 1 Dec 2014 22:34:03 +0000 (22:34 +0000)]
Added a new regular expression to the "_regexp-break" command ("b" by default):

(lldb) b /break here/

This will set a source level regular expression breakpoint on any text between the first '/' and the last '/'. The equivalent command will be:

(lldb) breakpoint set --source-pattern-regexp 'break here'

llvm-svn: 223082

9 years agoAdded StringExtractor::DecodeHexU8 && GetHexBytesAvail
Vince Harron [Mon, 1 Dec 2014 22:19:33 +0000 (22:19 +0000)]
Added StringExtractor::DecodeHexU8 && GetHexBytesAvail

DecodeHexU8 returns a decoded hex character pair, returns -1 if a
valid hex pair is not available.

GetHexBytesAvail decodes all available hex pairs.

llvm-svn: 223081

9 years agoStringExtractor unit tests
Vince Harron [Mon, 1 Dec 2014 22:10:15 +0000 (22:10 +0000)]
StringExtractor unit tests

Unit tests to cover StringExtractor in advance of minor new
functionality & minor refactor

llvm-svn: 223080

9 years agoUse nullptr to silence -Wsentinel when self-hosting on Windows
Reid Kleckner [Mon, 1 Dec 2014 22:02:27 +0000 (22:02 +0000)]
Use nullptr to silence -Wsentinel when self-hosting on Windows

Richard rejected my Sema change to interpret an integer literal zero in
a varargs context as a null pointer, so -Wsentinel sees an integer
literal zero and fires off a warning. Only CodeGen currently knows that
it promotes integer literal zeroes in this context to pointer size on
Windows.  I didn't want to teach -Wsentinel about that compatibility
hack. Therefore, I'm migrating to C++11 nullptr.

llvm-svn: 223079

9 years ago[Statepoints 1/4] Statepoint infrastructure for garbage collection: IR Intrinsics
Philip Reames [Mon, 1 Dec 2014 21:18:12 +0000 (21:18 +0000)]
[Statepoints 1/4] Statepoint infrastructure for garbage collection: IR Intrinsics

The statepoint intrinsics are intended to enable precise root tracking through the compiler as to support garbage collectors of all types. The addition of the statepoint intrinsics to LLVM should have no impact on the compilation of any program which does not contain them. There are no side tables created, no extra metadata, and no inhibited optimizations.

A statepoint works by transforming a call site (or safepoint poll site) into an explicit relocation operation. It is the frontend's responsibility (or eventually the safepoint insertion pass we've developed, but that's not part of this patch series) to ensure that any live pointer to a GC object is correctly added to the statepoint and explicitly relocated. The relocated value is just a normal SSA value (as seen by the optimizer), so merges of relocated and unrelocated values are just normal phis. The explicit relocation operation, the fact the statepoint is assumed to clobber all memory, and the optimizers standard semantics ensure that the relocations flow through IR optimizations correctly.

This is the first patch in a small series.  This patch contains only the IR parts; the documentation and backend support will be following separately.  The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.

Reviewed by: atrick, ributzka

llvm-svn: 223078

9 years ago[NVPTX] Do not emit .weak symbols for NVPTX
Jingyue Wu [Mon, 1 Dec 2014 21:16:17 +0000 (21:16 +0000)]
[NVPTX] Do not emit .weak symbols for NVPTX

Summary:
".weak" symbols cannot be consumed by ptxas (PR21685). This patch makes the
weak directive in MCAsmPrinter customizable, and disables emitting ".weak"
symbols for NVPTX.

Test Plan: weak-linkage.ll

Reviewers: jholewinski

Reviewed By: jholewinski

Subscribers: majnemer, jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D6455

llvm-svn: 223077

9 years agoParse 'ghccc' in .ll files as the GHC convention (cc 10)
Reid Kleckner [Mon, 1 Dec 2014 21:04:44 +0000 (21:04 +0000)]
Parse 'ghccc' in .ll files as the GHC convention (cc 10)

Previously we just used "cc 10" in the .ll files, but that isn't very
human readable.

llvm-svn: 223076

9 years ago[AArch64] Don't combine "select (setcc i1 LHS, RHS), vL, vR".
Ahmed Bougacha [Mon, 1 Dec 2014 20:59:00 +0000 (20:59 +0000)]
[AArch64] Don't combine "select (setcc i1 LHS, RHS), vL, vR".

r208210 introduced an optimization that improves the vector select
codegen by doing the setcc on vectors directly.
This is a problem they the setcc operands are i1s, because the
optimization would create vectors of i1, which aren't legal.

Part of PR21549.

Differential Revision: http://reviews.llvm.org/D6308

llvm-svn: 223075

9 years ago[AArch64] Fix v2i8->i16 bitcast legalization.
Ahmed Bougacha [Mon, 1 Dec 2014 20:52:32 +0000 (20:52 +0000)]
[AArch64] Fix v2i8->i16 bitcast legalization.

r213378 improved f16 bitcasts, so that they go directly through subregs,
instead of through the stack.  That code now causes an assertion failure
for bitcasts from other 16-bits types (most importantly v2i8).

Correct that by doing the custom lowering for i16 bitcasts only when the
input is an f16.

Part of PR21549.

Differential Revision: http://reviews.llvm.org/D6307

llvm-svn: 223074

9 years ago[Core] Remove function to not override RoundTripPasses.
Shankar Easwaran [Mon, 1 Dec 2014 20:28:54 +0000 (20:28 +0000)]
[Core] Remove function to not override RoundTripPasses.

RoundTripPasses should always be called in DEBUG mode if the environment
variable "LLD_RUN_ROUNDTRIP_TEST" is set.

Flavors should not be able to override this behavior.

llvm-svn: 223073

9 years ago[OCaml] Move Llvm.clone_module to its own Llvm_transform_utils module.
Peter Zotov [Mon, 1 Dec 2014 19:50:39 +0000 (19:50 +0000)]
[OCaml] Move Llvm.clone_module to its own Llvm_transform_utils module.

This way most code won't link this (substantially large) library,
if compiled statically with LLVM.

llvm-svn: 223072

9 years ago[OCaml] [cmake] Add CMake buildsystem for OCaml.
Peter Zotov [Mon, 1 Dec 2014 19:50:23 +0000 (19:50 +0000)]
[OCaml] [cmake] Add CMake buildsystem for OCaml.

Closes PR15325.

llvm-svn: 223071

9 years agolibc++: add NaCl and PNaCl support for std::random_device
JF Bastien [Mon, 1 Dec 2014 19:19:55 +0000 (19:19 +0000)]
libc++: add NaCl and PNaCl support for std::random_device

Summary:
The NaCl sandbox doesn't allow opening files under /dev, but it offers an API which provides the same capabilities. This is the same random device emulation that nacl_io performs for POSIX support, but nacl_io is an optional library so libc++ can't assume that device emulation will be performed. Note that NaCl only supports /dev/urandom, not /dev/random.

This patch also cleans up some of the preprocessor #endif, and fixes the test for Win32 (it accepts any token, and would therefore never throw regardless of the token provided).

Test Plan: ninja check-libcxx

Reviewers: dschuff, mclow.lists, danalbert

Subscribers: jfb, cfe-commits

Differential Revision: http://reviews.llvm.org/D6442

llvm-svn: 223068

9 years agoUse a continue to reduce indentation and clang-format. NFC.
Rafael Espindola [Mon, 1 Dec 2014 19:17:46 +0000 (19:17 +0000)]
Use a continue to reduce indentation and clang-format. NFC.

llvm-svn: 223067

9 years agoUse a range loop. NFC.
Rafael Espindola [Mon, 1 Dec 2014 19:08:07 +0000 (19:08 +0000)]
Use a range loop. NFC.

llvm-svn: 223066

9 years agoAdd missing 'break's, found by inspection. No functionality change; the
Richard Smith [Mon, 1 Dec 2014 18:59:10 +0000 (18:59 +0000)]
Add missing 'break's, found by inspection. No functionality change; the
fallthrough happened to do the right thing in both cases.

llvm-svn: 223064

9 years agoUse C++ typed enums instead of 'unsigned char' for MCInst Kind. NFC.
Pete Cooper [Mon, 1 Dec 2014 18:46:43 +0000 (18:46 +0000)]
Use C++ typed enums instead of 'unsigned char' for MCInst Kind.  NFC.

This makes it much easier to see the value of operands in the debugger.

llvm-svn: 223060

9 years ago[MachineVerifier] Accept a MBB with a single landing pad successor.
Ahmed Bougacha [Mon, 1 Dec 2014 18:43:53 +0000 (18:43 +0000)]
[MachineVerifier] Accept a MBB with a single landing pad successor.

The MachineVerifier used to check that there was always exactly one
unconditional branch to a non-landingpad (normal) successor.
If that normal successor to an invoke BB is unreachable, it seems
reasonable to only have one successor, the landing pad.
On targets other than AArch64 (and on AArch64 with a different testcase),
the branch folder turns the branch to the landing pad into a fallthrough.
The MachineVerifier, which relies on AnalyzeBranch, is unable to check
the condition, and doesn't complain. However, it does in this specific
testcase, where the branch to the landing pad remained.
Make the MachineVerifier accept it.

llvm-svn: 223059

9 years agoDrop SrcStructTypesSet. It is redundant.
Rafael Espindola [Mon, 1 Dec 2014 18:42:18 +0000 (18:42 +0000)]
Drop SrcStructTypesSet. It is redundant.

At the only point in the code it is used, we haven't added any of the src types
to DstStructTypesSet yet.

llvm-svn: 223057

9 years agoAdd a test for devirtualization of virtual operator calls.
Nico Weber [Mon, 1 Dec 2014 17:48:04 +0000 (17:48 +0000)]
Add a test for devirtualization of virtual operator calls.

There was no test coverage for this before: Modifiying
EmitCXXOperatorMemberCallee() to not call CanDevirtualizeMemberFunctionCall()
didn't make any test fail.

llvm-svn: 223056

9 years agoARM: lower tail calls correctly when using GHC calling convention.
Tim Northover [Mon, 1 Dec 2014 17:46:39 +0000 (17:46 +0000)]
ARM: lower tail calls correctly when using GHC calling convention.

Patch by Ben Gamari.

llvm-svn: 223055

9 years agoRevert r223049, r223050 and r223051 while investigating test failures.
Hans Wennborg [Mon, 1 Dec 2014 17:36:43 +0000 (17:36 +0000)]
Revert r223049, r223050 and r223051 while investigating test failures.

I didn't foresee affecting the Clang test suite :/

llvm-svn: 223054

9 years agoSimplifyCFG: Omit range checks for switch lookup tables when default is unreachable
Hans Wennborg [Mon, 1 Dec 2014 17:08:38 +0000 (17:08 +0000)]
SimplifyCFG: Omit range checks for switch lookup tables when default is unreachable

They would get optimized away later, but we might as well not emit them.

llvm-svn: 223051

9 years agoSimplifyCFG: don't remove unreachable default switch destinations
Hans Wennborg [Mon, 1 Dec 2014 17:08:35 +0000 (17:08 +0000)]
SimplifyCFG: don't remove unreachable default switch destinations

An unreachable default destination can be exploited by other optimizations, and
SDag lowering is now prepared to handle them efficiently.

For example, branches to the unreachable destination will be optimized away,
such as in the case of range checks for switch lookup tables.

On 64-bit Linux, this reduces the size of a clang bootstrap by 80 kB (and
Chromium by 30 kB).

llvm-svn: 223050

9 years agoSelectionDAG switch lowering: Replace unreachable default with most popular case.
Hans Wennborg [Mon, 1 Dec 2014 17:08:32 +0000 (17:08 +0000)]
SelectionDAG switch lowering: Replace unreachable default with most popular case.

This can significantly reduce the size of the switch, allowing for more
efficient lowering.

I also worked with the idea of exploiting unreachable defaults by
omitting the range check for jump tables, but always ended up with a
non-neglible binary size increase. It might be worth looking into some more.

llvm-svn: 223049

9 years agoUpdate R600 address space map to include generic
Matt Arsenault [Mon, 1 Dec 2014 16:46:03 +0000 (16:46 +0000)]
Update R600 address space map to include generic

llvm-svn: 223046

9 years agoPartial revert of r222986.
Rafael Espindola [Mon, 1 Dec 2014 16:32:20 +0000 (16:32 +0000)]
Partial revert of r222986.

The explicit set of destination types is not fully redundant when lazy loading
since the TypeFinder will not find types used only in function bodies.

This keeps the logic to drop the name of mapped types since it still helps
with avoiding further renaming.

llvm-svn: 223043

9 years agoR600/SI: Various instruction format bit test cleanups
Matt Arsenault [Mon, 1 Dec 2014 15:52:46 +0000 (15:52 +0000)]
R600/SI: Various instruction format bit test cleanups

- Fix missing SALU format bits
- Remove unused isSALUInstr
- Add isVALU
- Switch isDS to use a bit like the others
- Move SIInstrInfo::is* functions to header
- Reorder so they are approximately sorted by type (SALU, VALU, memory)

llvm-svn: 223038

9 years agoMake the function pointer a template argument instead of a runtime value.
Samuel Benzaquen [Mon, 1 Dec 2014 14:46:14 +0000 (14:46 +0000)]
Make the function pointer a template argument instead of a runtime value.

Summary:
Speed up the variadic matchers by removing one indirect call.
Making the function pointer a template arguments allows the compiler to
inline the call instead of doing an runtime call by pointer.
Also, optimize the allOf() case to avoid redundant kind checks.
This speeds up our clang-tidy benchmark by ~2%

Reviewers: klimek

Subscribers: klimek, cfe-commits

Differential Revision: http://reviews.llvm.org/D6424

llvm-svn: 223029

9 years ago[OPENMP] Formating and code improvement for codegen of 'omp critical' directive.
Alexey Bataev [Mon, 1 Dec 2014 11:32:38 +0000 (11:32 +0000)]
[OPENMP] Formating and code improvement for codegen of 'omp critical' directive.
No functional changes, only code improvements.

llvm-svn: 223010

9 years agoThe andi16, addiusp and jraddiusp micromips instructions were missing dedicated decod...
Vladimir Medic [Mon, 1 Dec 2014 11:12:04 +0000 (11:12 +0000)]
The andi16, addiusp and jraddiusp micromips instructions were missing dedicated decoder methods in MipsDisassembler.cpp to properly decode immediate operands. These methods are added together with corresponding tests.

llvm-svn: 223006

9 years ago[msan] Add compile-time checks for missing origins.
Evgeniy Stepanov [Mon, 1 Dec 2014 09:53:51 +0000 (09:53 +0000)]
[msan] Add compile-time checks for missing origins.

This change makes MemorySanitizer instrumentation a bit more strict
about instructions that have no origin id assigned to them.

This would have caught the bug that was fixed in r222918.

No functional change.

llvm-svn: 222997

9 years ago[PowerPC] Fix unwind info with dynamic stack realignment
Jay Foad [Mon, 1 Dec 2014 09:42:32 +0000 (09:42 +0000)]
[PowerPC] Fix unwind info with dynamic stack realignment

Summary:
PowerPC DWARF unwind info defined CFA as SP + offset even in a function
where the stack had been dynamically realigned. This clearly doesn't
work because the offset from SP to CFA is not a constant. Fix it by
defining CFA as BP instead.

This was causing the AddressSanitizer null_deref test to fail 50% of
the time, depending on whether SP happened to be 32-byte aligned on
entry to a particular function or not.

Reviewers: willschm, uweigand, hfinkel

Reviewed By: hfinkel

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D6410

llvm-svn: 222996

9 years agoRevert "Remove threshold for lifetime marker insertion of named temporaries"
Arnaud A. de Grandmaison [Mon, 1 Dec 2014 09:30:16 +0000 (09:30 +0000)]
Revert "Remove threshold for lifetime marker insertion of named temporaries"

Revert r222993 while I investigate some MemorySanitizer failures.

llvm-svn: 222995

9 years agoRemoved extra whitespace. (Testing commit access). NFC.
Sonam Kumari [Mon, 1 Dec 2014 09:27:46 +0000 (09:27 +0000)]
Removed extra whitespace. (Testing commit access). NFC.

llvm-svn: 222994

9 years agoRemove threshold for lifetime marker insertion of named temporaries
Arnaud A. de Grandmaison [Mon, 1 Dec 2014 09:13:54 +0000 (09:13 +0000)]
Remove threshold for lifetime marker insertion of named temporaries

Now that TailRecursionElimination has been fixed with r222354, the
threshold on size for lifetime marker insertion can be removed. This
only affects named temporary though, as the patch for unnamed temporaries
is still in progress.

llvm-svn: 222993

9 years agoAdd post-decode checking of HVC instruction.
Charlie Turner [Mon, 1 Dec 2014 08:50:27 +0000 (08:50 +0000)]
Add post-decode checking of HVC instruction.

Add checkDecodedInstruction for post-decode checking of instructions, to catch
the corner cases like HVC that don't fit into the general pattern. Needed to
check for an invalid condition field in instruction encoding despite HVC not
taking a predicate.

Patch by Matthew Wahab.

Change-Id: I48e28de981d7a9e43569594da3c45fb478b4f795
llvm-svn: 222992

9 years ago[asan] Change dynamic alloca instrumentation to only consider allocas that are domina...
Yury Gribov [Mon, 1 Dec 2014 08:47:58 +0000 (08:47 +0000)]
[asan] Change dynamic alloca instrumentation to only consider allocas that are dominating all exits from function.

Reviewed in http://reviews.llvm.org/D6412

llvm-svn: 222991

9 years agoAdd Thumb HVC and ERET virtualisation extension instructions.
Charlie Turner [Mon, 1 Dec 2014 08:39:19 +0000 (08:39 +0000)]
Add Thumb HVC and ERET virtualisation extension instructions.

Patch by Matthew Wahab.

Change-Id: I131f71c1150d5fa797066a18e09d526c19bf9016
llvm-svn: 222990

9 years agoAdd ARM ERET and HVC virtualisation extension instructions.
Charlie Turner [Mon, 1 Dec 2014 08:33:28 +0000 (08:33 +0000)]
Add ARM ERET and HVC virtualisation extension instructions.

Patch by Matthew Wahab.

Change-Id: Iad75f078fbaa4ecc7d7a4820ad9b3930679cbbbb
llvm-svn: 222989

9 years agoFix capitalization. NFC.
Akira Hatanaka [Mon, 1 Dec 2014 06:14:52 +0000 (06:14 +0000)]
Fix capitalization. NFC.

llvm-svn: 222988

9 years ago[stack protector] Set edge weights for newly created basic blocks.
Akira Hatanaka [Mon, 1 Dec 2014 04:27:03 +0000 (04:27 +0000)]
[stack protector] Set edge weights for newly created basic blocks.

This commit fixes a bug in stack protector pass where edge weights were not set
when new basic blocks were added to lists of successor basic blocks.

Differential Revision: http://reviews.llvm.org/D5766

llvm-svn: 222987

9 years agoChange how we keep track of which types are in the dest module.
Rafael Espindola [Mon, 1 Dec 2014 04:15:59 +0000 (04:15 +0000)]
Change how we keep track of which types are in the dest module.

Instead of keeping an explicit set, just drop the names of types we choose
to map to some other type.

This has the advantage that the name of the unused will not cause the context
to rename types on module read.

llvm-svn: 222986

9 years agoAdd a test showing what the linker IdentifiedStructTypes is for.
Rafael Espindola [Mon, 1 Dec 2014 03:20:57 +0000 (03:20 +0000)]
Add a test showing what the linker IdentifiedStructTypes is for.

Without this it could just be deleted and all tests would pass.

llvm-svn: 222985

9 years agoRelax an assert a bit to avoid a crash on unreachable code.
Rafael Espindola [Mon, 1 Dec 2014 02:55:24 +0000 (02:55 +0000)]
Relax an assert a bit to avoid a crash on unreachable code.

Patch by Duncan Exon Smith with a small tweak by me.

llvm-svn: 222984

9 years ago[Core] Add flag to check if RoundTripPasses need to be run.
Shankar Easwaran [Mon, 1 Dec 2014 01:04:11 +0000 (01:04 +0000)]
[Core] Add flag to check if RoundTripPasses need to be run.

This would allow other flavor specific contexts to override the default value,
if they want to optionally run the round trip passes.

There is some information lost like the original file owner of the atom with
RoundTripPasses. The Gnu flavor needs this information inorder to implement
LinkerScript matching and for other diagnostic outputs such as Map files.

The flag also can be used to record information in the Atom if the information
to the Writer needs to be conveyed through References too.

llvm-svn: 222983

9 years ago[ELF] Fix layout of output sections.
Shankar Easwaran [Sun, 30 Nov 2014 22:36:29 +0000 (22:36 +0000)]
[ELF] Fix layout of output sections.

The AtomSections were improperly merging sections from various input files. This
patch fixes the problem, with an updated test that was provided by Simon.

Thanks to Simon Atanasyan for catching this issue.

llvm-svn: 222982

9 years ago[clang-tidy] google-explicit-constructor: don't match in template instantiations
Alexander Kornienko [Sun, 30 Nov 2014 19:41:41 +0000 (19:41 +0000)]
[clang-tidy] google-explicit-constructor: don't match in template instantiations

This helps avoiding false positives related to the recently added
std::initializer_list<> handling.

llvm-svn: 222981

9 years agoDrop unused enum value
Tobias Grosser [Sun, 30 Nov 2014 15:57:07 +0000 (15:57 +0000)]
Drop unused enum value

llvm-svn: 222980

9 years agoRemove Polly's IndVarSimplify pass
Tobias Grosser [Sun, 30 Nov 2014 14:33:41 +0000 (14:33 +0000)]
Remove Polly's IndVarSimplify pass

Polly had a copy of this pass to create the canonical induction variables
necessary for the non-scev-based code generation. As we now always use SCEV
based code generation, canonical induction variables are not needed any more.

llvm-svn: 222979

9 years agoRemove -polly-codegen-scev option and related code
Tobias Grosser [Sun, 30 Nov 2014 14:33:31 +0000 (14:33 +0000)]
Remove -polly-codegen-scev option and related code

SCEV based code generation has been the default for two weeks after having
been tested for a long time. We now drop the support the non-scev-based code
generation.

llvm-svn: 222978

9 years agocmake: Add polly libraries to LLVMExports
Tobias Grosser [Sun, 30 Nov 2014 12:45:44 +0000 (12:45 +0000)]
cmake: Add polly libraries to LLVMExports

This fixes LINK_POLLY_INTO_TOOLS=ON builds, which previously failed with:

CMake Error: install(EXPORT "LLVMExports" ...) includes target "opt" which
requires target "Polly" that is not in the export set.

CMake Error: install(EXPORT "LLVMExports" ...) includes target "bugpoint" which
requires target "Polly" that is not in the export set.

llvm-svn: 222977

9 years ago[PowerPC] Add asm support for cache-inhibited ld/st instructions
Hal Finkel [Sun, 30 Nov 2014 10:15:56 +0000 (10:15 +0000)]
[PowerPC] Add asm support for cache-inhibited ld/st instructions

Add assembler support for the fixed-point cache-inhibited load/store
instructions. These are hypervisor-level only, so don't get too excited ;)

Fixes PR21650.

llvm-svn: 222976

9 years ago[ELF] Rename align2 to alignment.
Shankar Easwaran [Sun, 30 Nov 2014 04:17:06 +0000 (04:17 +0000)]
[ELF] Rename align2 to alignment.

No change in functionality.

llvm-svn: 222975

9 years ago[ELF] Fix creation of segments.
Shankar Easwaran [Sun, 30 Nov 2014 04:01:14 +0000 (04:01 +0000)]
[ELF] Fix creation of segments.

Linker was creating a separate output segment in some cases if input sections
had huge alignments. This patch fixes the issue.

llvm-svn: 222974

9 years ago[ELF] Create input and output section names
Shankar Easwaran [Sun, 30 Nov 2014 03:43:38 +0000 (03:43 +0000)]
[ELF] Create input and output section names

No change in functionality.

llvm-svn: 222973

9 years ago[ELF] Rename MergedSection to OutputSection.
Shankar Easwaran [Sun, 30 Nov 2014 03:18:08 +0000 (03:18 +0000)]
[ELF] Rename MergedSection to OutputSection.

No change in functionality.

llvm-svn: 222972

9 years agoRevert r222957 "Replace std::map<K, V*> with std::map<K, V> to handle ownership and...
Craig Topper [Sun, 30 Nov 2014 01:20:17 +0000 (01:20 +0000)]
Revert r222957 "Replace std::map<K, V*> with std::map<K, V> to handle ownership and deletion of the values."

Upon further review I think the MultiClass is being copied into the map instead of being moved due to the copy constructor on the nested Record type. This ultimately got exposed when the vector in DefPrototype vector was changed to hold unique_ptrs in another commit. This caused gcc 4.7 to fail due to the use of the copy constructor on unique_ptr with the error pointing back to one of the insert calls from this commit. Not sure why clang was able to build.

This reverts commit 710cdf729f84b428bf41aa8d32dbdb35fff79fde.

llvm-svn: 222971

9 years agoRevert r222965 "Make MultiClass::DefPrototypes own their Records to fix memory leaks."
Hans Wennborg [Sun, 30 Nov 2014 00:43:50 +0000 (00:43 +0000)]
Revert r222965 "Make MultiClass::DefPrototypes own their Records to fix memory leaks."

The bots started failing with the error below. I suspect this revision was the cause.

FAILED: /home/bb/bin/g++47   -DGTEST_HAS_RTTI=0 -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -fPIC -fvisibility-inlines-hidden -Wall -W -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wno-missing-field-initializers -pedantic -Wno-long-long -Wno-maybe-uninitialized -Wno-comment -std=c++11 -ffunction-sections -fdata-sections -O3 -DNDEBUG -Ilib/TableGen -I/home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen -Iinclude -I/home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include    -fno-exceptions -fno-rtti -MMD -MT lib/TableGen/CMakeFiles/LLVMTableGen.dir/TGParser.cpp.o -MF "lib/TableGen/CMakeFiles/LLVMTableGen.dir/TGParser.cpp.o.d" -o lib/TableGen/CMakeFiles/LLVMTableGen.dir/TGParser.cpp.o -c /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.cpp
In file included from /usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_tempbuf.h:62:0,
                 from /usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_algo.h:64,
                 from /usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/algorithm:63,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include/llvm/ADT/StringRef.h:13,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGLexer.h:17,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.h:17,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.cpp:14:
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_construct.h: In instantiation of 'void std::_Construct(_T1*, _Args&& ...) [with _T1 = std::unique_ptr<llvm::Record>; _Args = {const std::unique_ptr<llvm::Record, std::default_delete<llvm::Record> >&}]':
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_uninitialized.h:77:3:   required from 'static _ForwardIterator std::__uninitialized_copy<_TrivialValueTypes>::__uninit_copy(_InputIterator, _InputIterator, _ForwardIterator) [with _InputIterator = __gnu_cxx::__normal_iterator<const std::unique_ptr<llvm::Record>*, std::vector<std::unique_ptr<llvm::Record> > >; _ForwardIterator = std::unique_ptr<llvm::Record>*; bool _TrivialValueTypes = false]'
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_uninitialized.h:119:41:   required from '_ForwardIterator std::uninitialized_copy(_InputIterator, _InputIterator, _ForwardIterator) [with _InputIterator = __gnu_cxx::__normal_iterator<const std::unique_ptr<llvm::Record>*, std::vector<std::unique_ptr<llvm::Record> > >; _ForwardIterator = std::unique_ptr<llvm::Record>*]'
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_uninitialized.h:260:63:   required from '_ForwardIterator std::__uninitialized_copy_a(_InputIterator, _InputIterator, _ForwardIterator, std::allocator<_Tp>&) [with _InputIterator = __gnu_cxx::__normal_iterator<const std::unique_ptr<llvm::Record>*, std::vector<std::unique_ptr<llvm::Record> > >; _ForwardIterator = std::unique_ptr<llvm::Record>*; _Tp = std::unique_ptr<llvm::Record>]'
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_vector.h:310:9:   required from 'std::vector<_Tp, _Alloc>::vector(const std::vector<_Tp, _Alloc>&) [with _Tp = std::unique_ptr<llvm::Record>; _Alloc = std::allocator<std::unique_ptr<llvm::Record> >]'
/home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include/llvm/TableGen/Record.h:1664:8:   required from 'constexpr std::pair<_T1, _T2>::pair(_U1&&, _U2&&) [with _U1 = std::basic_string<char>&; _U2 = llvm::MultiClass; <template-parameter-2-3> = void; _T1 = std::basic_string<char>; _T2 = llvm::MultiClass]'
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_pair.h:273:72:   required from 'constexpr std::pair<typename std::__decay_and_strip<_T1>::__type, typename std::__decay_and_strip<_T2>::__type> std::make_pair(_T1&&, _T2&&) [with _T1 = std::basic_string<char>&; _T2 = llvm::MultiClass; typename std::__decay_and_strip<_T2>::__type = llvm::MultiClass; typename std::__decay_and_strip<_T1>::__type = std::basic_string<char>]'
/home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.cpp:2295:78:   required from here
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/stl_construct.h:77:7: error: use of deleted function 'std::unique_ptr<_Tp, _Dp>::unique_ptr(const std::unique_ptr<_Tp, _Dp>&) [with _Tp = llvm::Record; _Dp = std::default_delete<llvm::Record>]'
In file included from /usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/memory:86:0,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include/llvm/ADT/STLExtras.h:25,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include/llvm/ADT/ArrayRef.h:14,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/include/llvm/Support/SourceMgr.h:19,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.h:19,
                 from /home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/lib/TableGen/TGParser.cpp:14:
/usr/lib/gcc/x86_64-redhat-linux/4.7.2/../../../../include/c++/4.7.2/bits/unique_ptr.h:262:7: error: declared here
ninja: build stopped: subcommand failed.

llvm-svn: 222970

9 years agoclang/test/CodeGenOpenCL/opencl_types.cl: Appease i686-msvc.
NAKAMURA Takumi [Sun, 30 Nov 2014 00:32:02 +0000 (00:32 +0000)]
clang/test/CodeGenOpenCL/opencl_types.cl: Appease i686-msvc.

llvm-svn: 222969

9 years agoQualify one more make_unique call.
Hans Wennborg [Sun, 30 Nov 2014 00:31:49 +0000 (00:31 +0000)]
Qualify one more make_unique call.

The previous patch had effect, but missed this one. It seems MSVC
gets ADL-confused by the calls where the first argument is a function call?

llvm-svn: 222968

9 years agoSpeculatively qualify some llvm::make_unique calls trying to please MSVC
Hans Wennborg [Sun, 30 Nov 2014 00:24:43 +0000 (00:24 +0000)]
Speculatively qualify some llvm::make_unique calls trying to please MSVC

It was failing with this kind of error:

C:\b\build\slave\CrWinClang\build\src\third_party\llvm\lib\TableGen\TGParser.cpp(1243) : error C2668: 'llvm::make_unique' : ambiguous call to overloaded function

        C:\b\build\slave\CrWinClang\build\src\third_party\llvm\include\llvm/ADT/STLExtras.h(408): could be 'std::unique_ptr<llvm::Record,std::default_delete<_Ty>> llvm::make_unique<llvm::Record,std::string,llvm::SMLoc&,llvm::RecordKeeper&,bool>(std::string &&,llvm::SMLoc &,llvm::RecordKeeper &,bool &&)'
        with
        [
            _Ty=llvm::Record
        ]
        C:\b\depot_tools\win_toolchain\vs2013_files\win8sdk\bin\..\..\VC\include\memory(1637): or       'std::unique_ptr<llvm::Record,std::default_delete<_Ty>> std::make_unique<llvm::Record,std::string,llvm::SMLoc&,llvm::RecordKeeper&,bool>(std::string &&,llvm::SMLoc &,llvm::RecordKeeper &,bool &&)' [found using argument-dependent lookup]
        with
        [
            _Ty=llvm::Record
        ]
        while trying to match the argument list '(std::string, llvm::SMLoc, llvm::RecordKeeper, bool)'

llvm-svn: 222967

9 years agoUse an unsigned type because there seems to be no reason for it to be signed.
Craig Topper [Sun, 30 Nov 2014 00:24:32 +0000 (00:24 +0000)]
Use an unsigned type because there seems to be no reason for it to be signed.

llvm-svn: 222966

9 years agoMake MultiClass::DefPrototypes own their Records to fix memory leaks.
Craig Topper [Sun, 30 Nov 2014 00:19:28 +0000 (00:19 +0000)]
Make MultiClass::DefPrototypes own their Records to fix memory leaks.

llvm-svn: 222965

9 years agoFix typo.
Nico Weber [Sat, 29 Nov 2014 23:57:35 +0000 (23:57 +0000)]
Fix typo.

llvm-svn: 222964

9 years agoFix grammar error in err_bad_reinterpret_cast_reference error message.
Nathan Sidwell [Sat, 29 Nov 2014 22:51:47 +0000 (22:51 +0000)]
Fix grammar error in err_bad_reinterpret_cast_reference error message.

llvm-svn: 222963

9 years agoSwitch lowering: reformat some for loops etc. NFC
Hans Wennborg [Sat, 29 Nov 2014 21:24:12 +0000 (21:24 +0000)]
Switch lowering: reformat some for loops etc. NFC

llvm-svn: 222962

9 years agoSwitch lowering: Fix broken 'Figure out which block is next' code
Hans Wennborg [Sat, 29 Nov 2014 21:17:05 +0000 (21:17 +0000)]
Switch lowering: Fix broken 'Figure out which block is next' code

This doesn't seem to have worked in a long time, but other optimizations
would clean it up.

llvm-svn: 222961

9 years agoTarget triple OS detection tidyup. NFC
Simon Pilgrim [Sat, 29 Nov 2014 19:18:21 +0000 (19:18 +0000)]
Target triple OS detection tidyup. NFC

Use Triple::isOS*() helpers where possible.

llvm-svn: 222960

9 years agoRemove some unnecessary vector::reserve/assign calls.
David Blaikie [Sat, 29 Nov 2014 18:13:42 +0000 (18:13 +0000)]
Remove some unnecessary vector::reserve/assign calls.

llvm-svn: 222959

9 years agoRemove indirection of vector<T*> in favor of deque<T>
David Blaikie [Sat, 29 Nov 2014 18:13:39 +0000 (18:13 +0000)]
Remove indirection of vector<T*> in favor of deque<T>

llvm-svn: 222958

9 years agoReplace std::map<K, V*> with std::map<K, V> to handle ownership and deletion of the...
Craig Topper [Sat, 29 Nov 2014 18:12:59 +0000 (18:12 +0000)]
Replace std::map<K, V*> with std::map<K, V> to handle ownership and deletion of the values.

llvm-svn: 222957

9 years agoTweak clang/test/CodeGenOpenCL/opencl_types.cl to appease msvc since r222941.
NAKAMURA Takumi [Sat, 29 Nov 2014 17:27:07 +0000 (17:27 +0000)]
Tweak clang/test/CodeGenOpenCL/opencl_types.cl to appease msvc since r222941.

llvm-svn: 222956

9 years agoRemove 'else' after 'return'. Fix formatting of a 'switch' statement.
Craig Topper [Sat, 29 Nov 2014 16:05:27 +0000 (16:05 +0000)]
Remove 'else' after 'return'. Fix formatting of a 'switch' statement.

llvm-svn: 222955

9 years agoAdd a HasSuppressionType method into SuppressionContext
Kuba Brecka [Sat, 29 Nov 2014 14:18:05 +0000 (14:18 +0000)]
Add a HasSuppressionType method into SuppressionContext

Extending SuppressionContext to add a HasSuppressionType method that tells whether a certain suppression type is currently used or not. It's a step to implement issue suppressions for ASan, see http://reviews.llvm.org/D6280.

Reviewed at http://reviews.llvm.org/D6443

llvm-svn: 222954

9 years ago[mips][microMIPS] Implement NOP aliases
Jozef Kolek [Sat, 29 Nov 2014 13:29:24 +0000 (13:29 +0000)]
[mips][microMIPS] Implement NOP aliases

This patch implements microMIPS 16-bit (MOVE16 $0, $0) and
32-bit (SLL $0, $0, 0) NOP aliases.

http://reviews.llvm.org/D6440

llvm-svn: 222953

9 years agoAdd support for 32-bit i386 binaries.
Jason Molenda [Sat, 29 Nov 2014 09:17:40 +0000 (09:17 +0000)]
Add support for 32-bit i386 binaries.

llvm-svn: 222952

9 years agoThe finishing touches on getting the compact unwind dumping completed
Jason Molenda [Sat, 29 Nov 2014 07:05:07 +0000 (07:05 +0000)]
The finishing touches on getting the compact unwind dumping completed
for x86_64.  i386, arm, arm64 aren't handled yet but those are minor
variations on this format.

This commit also adds code to read the symbol table out of the
binary and read the LC_FUNCTION_STARTS to augment the symbol table
with the start addresses of all the functions - and print the
function names when showing the unwind info.

llvm-svn: 222951

9 years agoUse deque<T> rather than vector<T*> since it provides the same invalidation semantics...
David Blaikie [Sat, 29 Nov 2014 07:04:51 +0000 (07:04 +0000)]
Use deque<T> rather than vector<T*> since it provides the same invalidation semantics (at least when removal is not needed) without the extra indirection/ownership complexity

Order matters for this container, it seems (using a forward_list and
replacing the original push_backs with emplace_fronts caused test
failures). I didn't look too deeply into why.

(& in retrospect, I might go back & change some of the forward_lists I
introduced to deques anyway - since most don't require removal, deque is
a more memory-friendly data structure (moderate locality while not
invalidating pointers))

llvm-svn: 222950

9 years agoConstify some things in preparation for CodeGenSubRegIndex to be stored by value...
David Blaikie [Sat, 29 Nov 2014 07:04:49 +0000 (07:04 +0000)]
Constify some things in preparation for CodeGenSubRegIndex to be stored by value in their container, removing the indirection

llvm-svn: 222949

9 years agoMake RecordKeeper::addClass/addDef take unique_ptrs instead of creating one internally.
Craig Topper [Sat, 29 Nov 2014 05:52:51 +0000 (05:52 +0000)]
Make RecordKeeper::addClass/addDef take unique_ptrs instead of creating one internally.

llvm-svn: 222948

9 years agoUse unique_ptr to remove some explicit deletes on some error case returns. At least...
Craig Topper [Sat, 29 Nov 2014 05:31:10 +0000 (05:31 +0000)]
Use unique_ptr to remove some explicit deletes on some error case returns. At least one spot of weird ownership passing that needs some future cleanup.

llvm-svn: 222947

9 years agoDebugIR: Remove empty directories from SVN
Duncan P. N. Exon Smith [Sat, 29 Nov 2014 03:17:43 +0000 (03:17 +0000)]
DebugIR: Remove empty directories from SVN

llvm-svn: 222946

9 years agoDebugIR: Delete -debug-ir
Duncan P. N. Exon Smith [Sat, 29 Nov 2014 03:15:47 +0000 (03:15 +0000)]
DebugIR: Delete -debug-ir

llvm-svn: 222945

9 years agoSupport: remove some variable names
Saleem Abdulrasool [Sat, 29 Nov 2014 00:10:26 +0000 (00:10 +0000)]
Support: remove some variable names

If built with -Wunused-variable, clang objects to the declarations due to the
unused variable; drop the names.  NFC.

llvm-svn: 222944

9 years agoRevert "Simplify some more ownership using forward_list<T> rather than vector<unique_...
Duncan P. N. Exon Smith [Fri, 28 Nov 2014 23:00:22 +0000 (23:00 +0000)]
Revert "Simplify some more ownership using forward_list<T> rather than vector<unique_ptr<T>>"

This reverts commit r222935 and its follow-up r222938 ("Push unique_ptr
a bit further through some APIs and simplify some cleanup"), since it
causes bot failures (at least on Darwin):

http://lab.llvm.org:8080/green/job/clang-stage1-cmake-RA-incremental/1391/

llvm-svn: 222943

9 years agoR600/SI: Fix assertion on sign extend of 3 vectors
Matt Arsenault [Fri, 28 Nov 2014 22:51:38 +0000 (22:51 +0000)]
R600/SI: Fix assertion on sign extend of 3 vectors

This was trying to create an MVT with 3x vectors which
created an invalid EVT

llvm-svn: 222942

9 years agoAST: Consider pseudo-struct builtin types as substitutable
David Majnemer [Fri, 28 Nov 2014 22:22:46 +0000 (22:22 +0000)]
AST: Consider pseudo-struct builtin types as substitutable

We didn't consider types like ObjCSel as a substitution candidate.

This fixes PR21688.

llvm-svn: 222941

9 years agoReapply "Use std::map<K, V> rather than std::map<K, std::unique_ptr<V>>""
David Blaikie [Fri, 28 Nov 2014 22:15:06 +0000 (22:15 +0000)]
Reapply "Use std::map<K, V> rather than std::map<K, std::unique_ptr<V>>""

Just avoid using std::map::emplace since it's not implemented in
libstdc++ 4.7.

Reapplies r222937, reverted in r222939.

llvm-svn: 222940

9 years agoRevert "Use std::map<K, V> rather than std::map<K, std::unique_ptr<V>>"
David Blaikie [Fri, 28 Nov 2014 22:01:06 +0000 (22:01 +0000)]
Revert "Use std::map<K, V> rather than std::map<K, std::unique_ptr<V>>"

Seems libstdc++ on some buildbots is lacking std::map::emplace, which is
weird... reverting while I look into it.

This reverts commit r222937.

llvm-svn: 222939

9 years agoPush unique_ptr a bit further through some APIs and simplify some cleanup
David Blaikie [Fri, 28 Nov 2014 21:59:58 +0000 (21:59 +0000)]
Push unique_ptr a bit further through some APIs and simplify some cleanup

llvm-svn: 222938