platform/upstream/llvm.git
11 years agoThe diagnostic is now a warning instead of an error. Also don't check lines beginning...
Bill Wendling [Wed, 23 Jan 2013 23:04:29 +0000 (23:04 +0000)]
The diagnostic is now a warning instead of an error. Also don't check lines beginning with '#', since they could contain a path with the unexpected word in them.

llvm-svn: 173305

11 years agoPush down the conversion of the alignment from the bit mask to a real number into...
Bill Wendling [Wed, 23 Jan 2013 23:00:05 +0000 (23:00 +0000)]
Push down the conversion of the alignment from the bit mask to a real number into the attribute implementation class.

llvm-svn: 173304

11 years agoFix for case-sensitive file systems. Ugh
Douglas Gregor [Wed, 23 Jan 2013 22:45:24 +0000 (22:45 +0000)]
Fix for case-sensitive file systems. Ugh

llvm-svn: 173303

11 years agoRemove dead methods.
Bill Wendling [Wed, 23 Jan 2013 22:38:33 +0000 (22:38 +0000)]
Remove dead methods.

llvm-svn: 173302

11 years agoImplement the writer side of the global module index.
Douglas Gregor [Wed, 23 Jan 2013 22:38:11 +0000 (22:38 +0000)]
Implement the writer side of the global module index.

The global module index is a "global" index for all of the module
files within a particular subdirectory in the module cache, which
keeps track of all of the "interesting" identifiers and selectors
known in each of the module files. One can perform a fast lookup in
the index to determine which module files will have more information
about entities with a particular name/selector. This information can
help eliminate redundant lookups into module files (a serious
performance problem) and help with creating auto-import/auto-include
Fix-Its.

The global module index is created or updated at the end of a
translation unit that has triggered a (re)build of a module by
scraping all of the .pcm files out of the module cache subdirectory,
so it catches everything. As with module rebuilds, we use the file
system's atomicity to synchronize.

llvm-svn: 173301

11 years agoAdd SectionPosition and OrderPass
Nick Kledzik [Wed, 23 Jan 2013 22:32:56 +0000 (22:32 +0000)]
Add SectionPosition and OrderPass

llvm-svn: 173300

11 years ago[ELF] Read IFUNC symbols correctly as typeResolver.
Michael J. Spencer [Wed, 23 Jan 2013 22:08:46 +0000 (22:08 +0000)]
[ELF] Read IFUNC symbols correctly as typeResolver.

llvm-svn: 173299

11 years agoFix small typo
Eli Bendersky [Wed, 23 Jan 2013 22:05:19 +0000 (22:05 +0000)]
Fix small typo

llvm-svn: 173298

11 years agoR600: Add a llvm.R600.store.swizzle intrinsics
Tom Stellard [Wed, 23 Jan 2013 21:39:49 +0000 (21:39 +0000)]
R600: Add a llvm.R600.store.swizzle intrinsics

This intrinsic is translated to ALLOC_EXPORT_WORD1_SWIZ, hence its
name. It is used to store vs/fs outputs

Patch by: Vincent Lejeune

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 173297

11 years agoR600: Simplify stream outputs intrinsic
Tom Stellard [Wed, 23 Jan 2013 21:39:47 +0000 (21:39 +0000)]
R600: Simplify stream outputs intrinsic

Patch by: Vincent Lejeune

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 173296

11 years agoAdd a test case for 'analyzer_noreturn' on category methods.
Ted Kremenek [Wed, 23 Jan 2013 21:29:13 +0000 (21:29 +0000)]
Add a test case for 'analyzer_noreturn' on category methods.

llvm-svn: 173295

11 years agoAlways process an index.html file if we have HTML output. Patch by Justin Bogner.
Ted Kremenek [Wed, 23 Jan 2013 21:22:14 +0000 (21:22 +0000)]
Always process an index.html file if we have HTML output.  Patch by Justin Bogner.

llvm-svn: 173294

11 years agoConstantFolding: Tweak r173289, it should evaluate in the intptr type, not the index...
Benjamin Kramer [Wed, 23 Jan 2013 21:21:24 +0000 (21:21 +0000)]
ConstantFolding: Tweak r173289, it should evaluate in the intptr type, not the index type.

llvm-svn: 173293

11 years agoAdd missing null check. Not sure why my tests passed before.
Ted Kremenek [Wed, 23 Jan 2013 21:12:49 +0000 (21:12 +0000)]
Add missing null check.  Not sure why my tests passed before.

llvm-svn: 173292

11 years agoHonor attribute 'analyzer_noreturn' on Objective-C methods.
Ted Kremenek [Wed, 23 Jan 2013 21:00:27 +0000 (21:00 +0000)]
Honor attribute 'analyzer_noreturn' on Objective-C methods.

This isn't likely a full solution, but it catches the common cases
and can be refined over time.

Fixes <rdar://problem/11634353>.

llvm-svn: 173291

11 years agoAdd extra indent for nested calls inside if's.
Daniel Jasper [Wed, 23 Jan 2013 20:41:06 +0000 (20:41 +0000)]
Add extra indent for nested calls inside if's.

Before:
if (aaaaaaaaaa(
    aaaaaaaaaa)) {}

After:
if (aaaaaaaaaa(
        aaaaaaaaaa)) {}

llvm-svn: 173290

11 years agoConstantFolding: Evaluate GEP indices in the index type.
Benjamin Kramer [Wed, 23 Jan 2013 20:41:05 +0000 (20:41 +0000)]
ConstantFolding: Evaluate GEP indices in the index type.

This fixes some edge cases that we would get wrong with uint64_ts.
PR14986.

llvm-svn: 173289

11 years agoAdd instruction encodings / disassembly support for l6r instructions.
Richard Osborne [Wed, 23 Jan 2013 20:08:11 +0000 (20:08 +0000)]
Add instruction encodings / disassembly support for l6r instructions.

llvm-svn: 173288

11 years agoAdd PassManager.
Michael J. Spencer [Wed, 23 Jan 2013 20:03:10 +0000 (20:03 +0000)]
Add PassManager.

It owns and manages passes.

llvm-svn: 173287

11 years agoImplement -Wvla correctly
Dmitri Gribenko [Wed, 23 Jan 2013 20:02:51 +0000 (20:02 +0000)]
Implement -Wvla correctly

GCC implements -Wvla as "warn on every VLA" (this is useful to find every VLA,
for example, if they are forbidden by coding guidelines).  Currently Clang
implements -Wvla as "warn on VLA when it is an extension".

The attached patch makes our behavior match GCC.  The existing vla extwarn is
moved under -Wvla-extension and is still included into -Wgnu.

This fixes PR5953.

llvm-svn: 173286

11 years agoInitialize SSPBufferSize. PR14999. Patch by Vinson Lee.
Chad Rosier [Wed, 23 Jan 2013 19:32:37 +0000 (19:32 +0000)]
Initialize SSPBufferSize. PR14999. Patch by Vinson Lee.

llvm-svn: 173285

11 years agoRemove unused methods and ivars.
Bill Wendling [Wed, 23 Jan 2013 19:06:01 +0000 (19:06 +0000)]
Remove unused methods and ivars.

llvm-svn: 173284

11 years agoFactor the trait for lookup into the on-based hash table of
Douglas Gregor [Wed, 23 Jan 2013 18:53:14 +0000 (18:53 +0000)]
Factor the trait for lookup into the on-based hash table of
identifiers into two parts: the part that involves dealing with the
key (which can be re-used) and the ASTReader-specific part that
creates the IdentifierInfos. While I'm at it, StringRef'ify this code,
which was using pair<const char*, unsigned>. No functionality change.

llvm-svn: 173283

11 years agodocs: clear up ambiguity
Sean Silva [Wed, 23 Jan 2013 18:28:48 +0000 (18:28 +0000)]
docs: clear up ambiguity

Anybody using clang is a "developer".

llvm-svn: 173282

11 years ago[PCH] Temporarily disable the "ambiguous macro" warning that is currently bogus with...
Argyrios Kyrtzidis [Wed, 23 Jan 2013 18:21:56 +0000 (18:21 +0000)]
[PCH] Temporarily disable the "ambiguous macro" warning that is currently bogus with a PCH
that redefined a macro without undef'ing it first.

Proper reconstruction of the macro info history from modules will properly fix this in subsequent commits.

rdar://13016031

llvm-svn: 173281

11 years agoDocumentation: add a note that -cc1 options are not guaranteed to be stable
Dmitri Gribenko [Wed, 23 Jan 2013 18:02:28 +0000 (18:02 +0000)]
Documentation: add a note that -cc1 options are not guaranteed to be stable

llvm-svn: 173280

11 years agoRevert "InstCombine: Clean up weird code that talks about a modulus that's long gone."
Benjamin Kramer [Wed, 23 Jan 2013 17:52:29 +0000 (17:52 +0000)]
Revert "InstCombine: Clean up weird code that talks about a modulus that's long gone."

This causes crashes during the build of compiler-rt during selfhost. Add a
testcase for coverage.

llvm-svn: 173279

11 years agolibclang: change return type of getCursorDecl() to 'const Decl *'
Dmitri Gribenko [Wed, 23 Jan 2013 17:25:27 +0000 (17:25 +0000)]
libclang: change return type of getCursorDecl() to 'const Decl *'

llvm-svn: 173278

11 years agoUse 'const Decl *' throughout code completion in Sema
Dmitri Gribenko [Wed, 23 Jan 2013 17:21:11 +0000 (17:21 +0000)]
Use 'const Decl *' throughout code completion in Sema

llvm-svn: 173277

11 years agoInstCombine: Clean up weird code that talks about a modulus that's long gone.
Benjamin Kramer [Wed, 23 Jan 2013 17:16:22 +0000 (17:16 +0000)]
InstCombine: Clean up weird code that talks about a modulus that's long gone.

This does the right thing unless the multiplication overflows, but the old code
didn't handle that case either.

llvm-svn: 173276

11 years agoFix powerpc test failure - forgot to initialize stack slot size for PPCLinuxMCAsmInfo
Eli Bendersky [Wed, 23 Jan 2013 17:12:15 +0000 (17:12 +0000)]
Fix powerpc test failure - forgot to initialize stack slot size for PPCLinuxMCAsmInfo

llvm-svn: 173275

11 years agoAdd constness for NestedNameSpecifier::Create parameter
Dmitri Gribenko [Wed, 23 Jan 2013 17:06:56 +0000 (17:06 +0000)]
Add constness for NestedNameSpecifier::Create parameter

llvm-svn: 173274

11 years agoDon't try to align builder-type continuations on assignments.
Daniel Jasper [Wed, 23 Jan 2013 16:58:21 +0000 (16:58 +0000)]
Don't try to align builder-type continuations on assignments.

Before:
int aaaa = aaaaa().aaaaa() // force break
           .aaaaa();
After:
int aaaa = aaaaa().aaaaa() // force break
    .aaaaa();

The other indent is just wrong and confusing.

llvm-svn: 173273

11 years agoConstify some getters in RedeclarableTemplateDecl
Dmitri Gribenko [Wed, 23 Jan 2013 16:52:57 +0000 (16:52 +0000)]
Constify some getters in RedeclarableTemplateDecl

llvm-svn: 173272

11 years agoClean up assignment of CalleeSaveStackSlotSize: get rid of the default and explicitly...
Eli Bendersky [Wed, 23 Jan 2013 16:22:04 +0000 (16:22 +0000)]
Clean up assignment of CalleeSaveStackSlotSize: get rid of the default and explicitly set this in every target that needs to change it from the default.

llvm-svn: 173270

11 years agoRemove uneeded casts
Dmitri Gribenko [Wed, 23 Jan 2013 15:56:07 +0000 (15:56 +0000)]
Remove uneeded casts

llvm-svn: 173269

11 years agoDon't try to do a hanging ident after assignments.
Daniel Jasper [Wed, 23 Jan 2013 15:55:19 +0000 (15:55 +0000)]
Don't try to do a hanging ident after assignments.

Before:
bool aaaa = aaaaaaaaaaa(
                aaaaaaaaaaaaaaaaa);

After:
bool aaaa = aaaaaaaaaaa(
    aaaaaaaaaaaaaaaaa);

The other indentation was a nice attempt but doesn't work in many cases.
Not sure what the right long term solution is as the "After: " is still
not nice. We either need to figure out what to do in the cases where it
"doesn't work" or come up with a third solution, e.g. falling back to:

bool aaaa =
    aaaaaaaaaaa(
        aaaaaaaaaaaaaaaaa);

which should always work and nicely highlight the structure.

llvm-svn: 173268

11 years agoNVPTX: Stop leaking memory by using a managed constant instead of a new Argument.
Benjamin Kramer [Wed, 23 Jan 2013 15:21:44 +0000 (15:21 +0000)]
NVPTX: Stop leaking memory by using a managed constant instead of a new Argument.

This is still an egregious hack since we don't have a nice interface for this
kind of thing but should help the valgrind leak check buildbot to become green.

llvm-svn: 173267

11 years agochange from using 2 bits to 4 bits, as 2 bits arent enough to hold the enumeration...
Shankar Easwaran [Wed, 23 Jan 2013 15:11:22 +0000 (15:11 +0000)]
change from using 2 bits to 4 bits, as 2 bits arent enough to hold the enumeration OutputKind

llvm-svn: 173266

11 years agoMake sure metarenamer won't rename special stuff (intrinsics and explicitly renamed...
Anton Korobeynikov [Wed, 23 Jan 2013 15:03:08 +0000 (15:03 +0000)]
Make sure metarenamer won't rename special stuff (intrinsics and explicitly renamed stuff).
Otherwise this might hide the problems.

llvm-svn: 173265

11 years agoFix handling of macro definitions.
Manuel Klimek [Wed, 23 Jan 2013 14:37:36 +0000 (14:37 +0000)]
Fix handling of macro definitions.

Now correctly formats:
 #define A (1)

llvm-svn: 173264

11 years ago[msan] Allow waitpid while in __msan_init().
Evgeniy Stepanov [Wed, 23 Jan 2013 14:22:05 +0000 (14:22 +0000)]
[msan] Allow waitpid while in __msan_init().

This is used in symbolizer init.

llvm-svn: 173263

11 years agoFixes layouting regression and invalid-read.
Manuel Klimek [Wed, 23 Jan 2013 14:08:21 +0000 (14:08 +0000)]
Fixes layouting regression and invalid-read.

Layouting would prevent breaking before + in
a[b + c] = d;
Regression detected by code review.

Also fixes an invalid-read found by the valgrind bot.

llvm-svn: 173262

11 years ago[sanitizer] increase the minimal mmap size in allocator to 2^16; fix the asan Mac...
Kostya Serebryany [Wed, 23 Jan 2013 14:07:17 +0000 (14:07 +0000)]
[sanitizer] increase the minimal mmap size in allocator to 2^16; fix the asan Mac build

llvm-svn: 173261

11 years ago[asan] initialize kHighMemEnd at startup (instead of at compile time) to simplify...
Kostya Serebryany [Wed, 23 Jan 2013 13:27:43 +0000 (13:27 +0000)]
[asan] initialize kHighMemEnd at startup (instead of at compile time) to simplify further changes for various address space layouts. Fix asan_allocator2 for PowerPC (tested on 44-bit address space)

llvm-svn: 173260

11 years agoFix the formatting of pointer/reference types in range-based for loops.
Daniel Jasper [Wed, 23 Jan 2013 12:58:14 +0000 (12:58 +0000)]
Fix the formatting of pointer/reference types in range-based for loops.

Before: for (int & a : Values) {}
After:  for (int &a : Values) {}
llvm-svn: 173259

11 years ago[asan] use ADD instead of OR when applying shadow offset of PowerPC. See http://gcc...
Kostya Serebryany [Wed, 23 Jan 2013 12:54:55 +0000 (12:54 +0000)]
[asan] use ADD instead of OR when applying shadow offset of PowerPC. See gcc.gnu.org/bugzilla/show_bug.cgi?id=55975 for details

llvm-svn: 173258

11 years agoRemoving the penalty for breaking after "=".
Daniel Jasper [Wed, 23 Jan 2013 12:27:43 +0000 (12:27 +0000)]
Removing the penalty for breaking after "=".

Having seen more cases, this actually was not a good thing to do in the
first place. We can still improve on what we do now, but breaking after
the "=" is good in many cases.

Before:
aaaaaaaaaaaaa = aa->aaaaaaaaaaaaaaaaaaaa(
                    aaaaaaaaaaaaaaaaaaaaaaaa(aaaaaaaaaaaaaa));

After:
aaaaaaaaaaaaa =
    aa->aaaaaaaaaaaaaaaaaaaa(aaaaaaaaaaaaaaaaaaaaaaaa(aaaaaaaaaaaaaa));

llvm-svn: 173257

11 years agoFix another regression for pointer types.
Daniel Jasper [Wed, 23 Jan 2013 12:10:53 +0000 (12:10 +0000)]
Fix another regression for pointer types.

Before: if (int * a = &b) ...
After:  if (int *a = &b) ...

Also changed all the existing tests to test the expressions in question
both in a declaration and in an expression context.

llvm-svn: 173256

11 years agotsan: add stubs for malloc introspection (similar to what we have in asan)
Dmitry Vyukov [Wed, 23 Jan 2013 12:08:03 +0000 (12:08 +0000)]
tsan: add stubs for malloc introspection (similar to what we have in asan)

llvm-svn: 173255

11 years agoAdd a new LangOpt NativeHalfType. This option allows for native half/fp16
Joey Gouly [Wed, 23 Jan 2013 11:56:20 +0000 (11:56 +0000)]
Add a new LangOpt NativeHalfType. This option allows for native half/fp16
operations (as opposed to storage only half/fp16).

Also add some semantic checks for OpenCL half types.

llvm-svn: 173254

11 years agotsan: add missing interceptors
Dmitry Vyukov [Wed, 23 Jan 2013 11:52:19 +0000 (11:52 +0000)]
tsan: add missing interceptors

llvm-svn: 173253

11 years agoFix regression in formatting pointer types.
Daniel Jasper [Wed, 23 Jan 2013 11:15:14 +0000 (11:15 +0000)]
Fix regression in formatting pointer types.

We will need a more principled solution, but we should not leave this
unfixed until we come up with one.

Before: void f() { int * a; }
After:  void f() { int *a; }
llvm-svn: 173252

11 years ago[asan] simplify the code that poisons global redzones, add some more tests
Kostya Serebryany [Wed, 23 Jan 2013 11:14:21 +0000 (11:14 +0000)]
[asan] simplify the code that poisons global redzones, add some more tests

llvm-svn: 173251

11 years agoFixes incorrect handling of the declaration context stack.
Manuel Klimek [Wed, 23 Jan 2013 11:03:04 +0000 (11:03 +0000)]
Fixes incorrect handling of the declaration context stack.

llvm-svn: 173250

11 years ago[msan] gethostbyname interceptor.
Evgeniy Stepanov [Wed, 23 Jan 2013 10:43:38 +0000 (10:43 +0000)]
[msan] gethostbyname interceptor.

llvm-svn: 173249

11 years agoFix segfaults in the formatter.
Manuel Klimek [Wed, 23 Jan 2013 10:09:28 +0000 (10:09 +0000)]
Fix segfaults in the formatter.

Also: expletive deleted.
llvm-svn: 173247

11 years agoAdd option to allow putting all parameters onto the next line.
Daniel Jasper [Wed, 23 Jan 2013 10:08:28 +0000 (10:08 +0000)]
Add option to allow putting all parameters onto the next line.

This only affects styles where BinPackParameters is false.

With AllowAllParametersOnNextLine:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(
    aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa);

Without it:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(aaaaaaaaaa,
                                         aaaaaaaaaa,
                                         aaaaaaaaaa,
                                         aaaaaaaaaaa,
                                         aaaaaaaaaaa);

llvm-svn: 173246

11 years agoAllow us to better guess the context of an unwrapped line.
Manuel Klimek [Wed, 23 Jan 2013 09:32:48 +0000 (09:32 +0000)]
Allow us to better guess the context of an unwrapped line.

This gives us the ability to guess better defaults for whether a *
between identifiers is a pointer dereference or binary operator.

Now correctly formats:
void f(a *b);
void f() { f(a * b); }

llvm-svn: 173243

11 years agoInitialize the components of this class. Otherwise GCC thinks that Array may be
Duncan Sands [Wed, 23 Jan 2013 09:09:50 +0000 (09:09 +0000)]
Initialize the components of this class.  Otherwise GCC thinks that Array may be
used uninitialized, since it fails to understand that Array is only used when
SingleValue is not, and outputs a warning.  It also seems generally safer given
that the constructor is non-trivial and has plenty of early exits.

llvm-svn: 173242

11 years agoExplicitly cast away the const-ness instead of doing it implicitly.
Bill Wendling [Wed, 23 Jan 2013 08:58:23 +0000 (08:58 +0000)]
Explicitly cast away the const-ness instead of doing it implicitly.

llvm-svn: 173241

11 years agollvm/unittests: Use OwningPtr to fix --vg-leak.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:33:13 +0000 (08:33 +0000)]
llvm/unittests: Use OwningPtr to fix --vg-leak.

llvm-svn: 173240

11 years agoIRTests/WaymarkTest.cpp: Fix in --vg-leak.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:33:05 +0000 (08:33 +0000)]
IRTests/WaymarkTest.cpp: Fix in --vg-leak.

llvm-svn: 173239

11 years agoIRTests/IRBuilderTest.cpp: GetIntTy: Delete DL at yourself since it is not linked.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:31:28 +0000 (08:31 +0000)]
IRTests/IRBuilderTest.cpp: GetIntTy: Delete DL at yourself since it is not linked.

llvm-svn: 173238

11 years agoIRTests/IRBuilderTest.cpp: Let GV added to the module.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:30:39 +0000 (08:30 +0000)]
IRTests/IRBuilderTest.cpp: Let GV added to the module.

It fixes --vg-leak.

llvm-svn: 173237

11 years agoIRTests/ConstantsTest.cpp: AsInstructionsTest: Delete each instruction immediately...
NAKAMURA Takumi [Wed, 23 Jan 2013 08:30:26 +0000 (08:30 +0000)]
IRTests/ConstantsTest.cpp: AsInstructionsTest: Delete each instruction immediately if it is not linked.

llvm-svn: 173236

11 years agoUntabify.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:30:21 +0000 (08:30 +0000)]
Untabify.

llvm-svn: 173235

11 years agoWhitespace.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:30:15 +0000 (08:30 +0000)]
Whitespace.

llvm-svn: 173234

11 years agoDominatorTreeTest.cpp: Add the file header.
NAKAMURA Takumi [Wed, 23 Jan 2013 08:30:10 +0000 (08:30 +0000)]
DominatorTreeTest.cpp: Add the file header.

llvm-svn: 173233

11 years agoExplicitly cast away the const-ness instead of doing it implicitly.
Bill Wendling [Wed, 23 Jan 2013 08:25:41 +0000 (08:25 +0000)]
Explicitly cast away the const-ness instead of doing it implicitly.

llvm-svn: 173232

11 years agoAdd the heuristic to differentiate SSPStrong from SSPRequired.
Bill Wendling [Wed, 23 Jan 2013 06:43:53 +0000 (06:43 +0000)]
Add the heuristic to differentiate SSPStrong from SSPRequired.

The requirements of the strong heuristic are:

* A Protector is required for functions which contain an array, regardless of
  type or length.

* A Protector is required for functions which contain a structure/union which
  contains an array, regardless of type or length.  Note, there is no limit to
  the depth of nesting.

* A protector is required when the address of a local variable (i.e., stack
  based variable) is exposed. (E.g., such as through a local whose address is
  taken as part of the RHS of an assignment or a local whose address is taken as
  part of a function argument.)

llvm-svn: 173231

11 years agoAdd the IR attribute 'sspstrong'.
Bill Wendling [Wed, 23 Jan 2013 06:41:41 +0000 (06:41 +0000)]
Add the IR attribute 'sspstrong'.

SSPStrong applies a heuristic to insert stack protectors in these situations:

* A Protector is required for functions which contain an array, regardless of
  type or length.

* A Protector is required for functions which contain a structure/union which
  contains an array, regardless of type or length.  Note, there is no limit to
  the depth of nesting.

* A protector is required when the address of a local variable (i.e., stack
  based variable) is exposed. (E.g., such as through a local whose address is
  taken as part of the RHS of an assignment or a local whose address is taken as
  part of a function argument.)

This patch implements the SSPString attribute to be equivalent to
SSPRequired. This will change in a subsequent patch.

llvm-svn: 173230

11 years agoRemove the last of uses that use the Attribute object as a collection of attributes.
Bill Wendling [Wed, 23 Jan 2013 06:15:10 +0000 (06:15 +0000)]
Remove the last of uses that use the Attribute object as a collection of attributes.

Collections of attributes are handled via the AttributeSet class now. This
finally frees us up to make significant changes to how attributes are structured.

llvm-svn: 173229

11 years agoRemove the last of uses that use the Attribute object as a collection of attributes.
Bill Wendling [Wed, 23 Jan 2013 06:14:59 +0000 (06:14 +0000)]
Remove the last of uses that use the Attribute object as a collection of attributes.

Collections of attributes are handled via the AttributeSet class now. This
finally frees us up to make significant changes to how attributes are structured.

llvm-svn: 173228

11 years agoMake __attribute__((nonnull)) use the general expression evaluator to search for
Nick Lewycky [Wed, 23 Jan 2013 05:08:29 +0000 (05:08 +0000)]
Make __attribute__((nonnull)) use the general expression evaluator to search for
nulls instead of limiting itself to the language-defined "null pointer
constant".

llvm-svn: 173227

11 years agoRemove a compile time warning in RNBRemote::HandlePacket_qProcessInfo
Jason Molenda [Wed, 23 Jan 2013 04:39:43 +0000 (04:39 +0000)]
Remove a compile time warning in RNBRemote::HandlePacket_qProcessInfo
for non-x86 builds.

llvm-svn: 173226

11 years agoChange the container-regs kv pair in the qRegsiterInfo
Jason Molenda [Wed, 23 Jan 2013 04:38:32 +0000 (04:38 +0000)]
Change the container-regs kv pair in the qRegsiterInfo
reply to be hex encoded, not decimal.

Fix the whitespace in the container-regs/invalidate-regs
documentation, fix one ambiguous hex/decimal number in an
example.

llvm-svn: 173225

11 years agodocs: Update title of external tutorial.
Sean Silva [Wed, 23 Jan 2013 03:21:41 +0000 (03:21 +0000)]
docs: Update title of external tutorial.

llvm-svn: 173224

11 years agoR600: rework handling of the constants
Tom Stellard [Wed, 23 Jan 2013 02:09:06 +0000 (02:09 +0000)]
R600: rework handling of the constants

Remove Cxxx registers, add new special register - "ALU_CONST" and new
operand for each alu src - "sel". ALU_CONST is used to designate that the
new operand contains the value to override src.sel, src.kc_bank, src.chan
for constants in the driver.

Patch by: Vadim Girlin

Vincent Lejeune:
  - Use pointers for constants
  - Fold CONST_ADDRESS when possible

Tom Stellard:
  - Give CONSTANT_BUFFER_0 its own address space
  - Use integer types for constant loads

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 173222

11 years agoR600: Add a CONST_ADDRESS node to model constant buf read
Tom Stellard [Wed, 23 Jan 2013 02:09:03 +0000 (02:09 +0000)]
R600: Add a CONST_ADDRESS node to model constant buf read

Patch by: Vincent Lejeune

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 173221

11 years agoR600: Factorise VTX_WORD0 and VTX_WORD1 in tblgen def
Tom Stellard [Wed, 23 Jan 2013 02:09:01 +0000 (02:09 +0000)]
R600: Factorise VTX_WORD0 and VTX_WORD1 in tblgen def

Patch by: Vincent Lejeune

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 173220

11 years agoAdd support for reverse pointer induction variables. These are loops that contain...
Nadav Rotem [Wed, 23 Jan 2013 01:35:00 +0000 (01:35 +0000)]
Add support for reverse pointer induction variables. These are loops that contain pointers that count backwards.
For example, this is the hot loop in BZIP:

  do {
    m = *--p;
    *p = ( ... );
  } while (--n);

llvm-svn: 173219

11 years agoclang/test/Driver/output-file-is-dir.c: This requires shell due to 'cd'.
NAKAMURA Takumi [Wed, 23 Jan 2013 01:25:23 +0000 (01:25 +0000)]
clang/test/Driver/output-file-is-dir.c: This requires shell due to 'cd'.

llvm-svn: 173218

11 years agoMove everything over to TargetInfo.
Michael J. Spencer [Wed, 23 Jan 2013 01:18:43 +0000 (01:18 +0000)]
Move everything over to TargetInfo.

I really would have liked to split this patch up, but it would greatly
complicate the lld-core and lld drivers having to deal with both
{Reader,Writer}Option and TargetInfo.

llvm-svn: 173217

11 years ago<rdar://problem/12711206>
Enrico Granata [Wed, 23 Jan 2013 01:17:27 +0000 (01:17 +0000)]
<rdar://problem/12711206>

Extending ValueObjectDynamicValue so that it stores a TypeAndOrName instead of a TypeSP.
This change allows us to reflect the notion that a ValueObject can have a dynamic type for which we have no debug information.
Previously, we would coalesce that to the static type of the object, potentially losing relevant information or even getting it wrong.
This fix ensures we can correctly report the class name for Cocoa objects whose types are hidden classes that we know nothing about (e.g. __NSArrayI for immutable arrays).
As a side effect, our --show-types argument to frame variable no longer needs to append custom dynamic type information.

llvm-svn: 173216

11 years ago[Driver] Don't remove non-regular files that were outputs.
Daniel Dunbar [Wed, 23 Jan 2013 01:08:23 +0000 (01:08 +0000)]
[Driver] Don't remove non-regular files that were outputs.

llvm-svn: 173215

11 years agoUse the AttributeSet when removing multiple attributes. Use Attribute::AttrKind
Bill Wendling [Wed, 23 Jan 2013 00:45:55 +0000 (00:45 +0000)]
Use the AttributeSet when removing multiple attributes. Use Attribute::AttrKind
when removing one attribute. This further encapsulates the use of the attributes.

llvm-svn: 173214

11 years agoDocumentation cleanup: fixing documentation for FrontendAction.
James Dennett [Wed, 23 Jan 2013 00:45:44 +0000 (00:45 +0000)]
Documentation cleanup: fixing documentation for FrontendAction.
* Fix a typo, s/BeginSourceAction/BeginSourceFile/, so that the documentation
  for FrontendAction::BeginSourceFileAction links correctly to BeginSourceFile;
* Add some basic \file documentation for FrontendAction.h;
* More use of "\brief" instead of repeating the name of the entity being
  documented;
* Stop using Doxygen-style "///" comments in FrontendAction.cpp, as they were
  polluting the documentation for BeginSourceFile;
* Drop incorrect "\see" markup that broke Doxygen's formatting;
* Other minor documentation fixes.

llvm-svn: 173213

11 years ago[Support][ErrorOr] Don't use nullptr :(
Michael J. Spencer [Wed, 23 Jan 2013 00:22:30 +0000 (00:22 +0000)]
[Support][ErrorOr] Don't use nullptr :(

llvm-svn: 173212

11 years agoUse the AttributeSet when adding multiple attributes and an Attribute::AttrKind
Bill Wendling [Wed, 23 Jan 2013 00:21:06 +0000 (00:21 +0000)]
Use the AttributeSet when adding multiple attributes and an Attribute::AttrKind
when adding a single attribute to the function.

llvm-svn: 173211

11 years agoUse the AttributeSet when adding multiple attributes and an Attribute::AttrKind
Bill Wendling [Wed, 23 Jan 2013 00:20:53 +0000 (00:20 +0000)]
Use the AttributeSet when adding multiple attributes and an Attribute::AttrKind
when adding a single attribute to the function.

llvm-svn: 173210

11 years ago[Support][ErrorOr] Add optimized specialization of ErrorOr<void>.
Michael J. Spencer [Wed, 23 Jan 2013 00:18:31 +0000 (00:18 +0000)]
[Support][ErrorOr] Add optimized specialization of ErrorOr<void>.

ErrorOr<void> represents an operation that returns nothing, but can still fail.
It should be used in cases where you need the aditional user data that ErrorOr
provides over error_code.

llvm-svn: 173209

11 years agoFix compilation on Linux, which defines PATH_MAX in a weird place,
Douglas Gregor [Tue, 22 Jan 2013 23:49:45 +0000 (23:49 +0000)]
Fix compilation on Linux, which defines PATH_MAX in a weird place,
from Saleem Abdulrasool!

llvm-svn: 173208

11 years agoAdd instruction encodings / disassembly support for u10 / lu10 instructions.
Richard Osborne [Tue, 22 Jan 2013 22:55:04 +0000 (22:55 +0000)]
Add instruction encodings / disassembly support for u10 / lu10 instructions.

llvm-svn: 173204

11 years ago[Support][ErrorOr] Make old gcc happy.
Michael J. Spencer [Tue, 22 Jan 2013 22:48:46 +0000 (22:48 +0000)]
[Support][ErrorOr] Make old gcc happy.

Apparently this is how C++98 worked pre-DR. (Thanks Richard).

llvm-svn: 173203

11 years agoFixed typo.
Michael Gottesman [Tue, 22 Jan 2013 21:53:43 +0000 (21:53 +0000)]
Fixed typo.

llvm-svn: 173202

11 years ago[ObjCARC] Refactored out the inner most 2-loops from PerformCodePlacement into the...
Michael Gottesman [Tue, 22 Jan 2013 21:49:00 +0000 (21:49 +0000)]
[ObjCARC] Refactored out the inner most 2-loops from PerformCodePlacement into the method ConnectTDBUTraversals.

The method PerformCodePlacement was doing too much (i.e. 3x loops, lots of
different checking). This refactoring separates the analysis section of the
method into a separate function while leaving the actual code placement and
analysis preparation in PerformCodePlacement.

*NOTE* Really this part of ObjCARC should be refactored out of the main pass
class into its own seperate class/struct. But, it is not time to make that
change yet though (don't want to make such an invasive change without fixing all
of the bugs first).

llvm-svn: 173201

11 years agoFix an issue of pseudo atomic instruction DAG schedule
Michael Liao [Tue, 22 Jan 2013 21:47:38 +0000 (21:47 +0000)]
Fix an issue of pseudo atomic instruction DAG schedule

- Add list of physical registers clobbered in pseudo atomic insts
  Physical registers are clobbered when pseudo atomic instructions are
  expanded. Add them in clobber list to prevent DAG scheduler to
  mis-schedule them after these insns are declared side-effect free.
- Add test case from Michael Kuperstein <michael.m.kuperstein@intel.com>

llvm-svn: 173200

11 years agoAdd a warning when there is a macro defintion that has named parameters but
Kevin Enderby [Tue, 22 Jan 2013 21:44:53 +0000 (21:44 +0000)]
Add a warning when there is a macro defintion that has named parameters but
the body does not use them and it appears the body has positional parameters.

This can cause unexpected results as in the added test case.  As the darwin
version of gas(1) which only supported positional parameters, happened to
ignore the named parameters.  Now that we want to support both styles of
macros we issue a warning in this specific case.

rdar://12861644

llvm-svn: 173199

11 years agoAdd a triple, per Ben's suggestion.
Chad Rosier [Tue, 22 Jan 2013 21:39:58 +0000 (21:39 +0000)]
Add a triple, per Ben's suggestion.

llvm-svn: 173198