]> granicus.if.org Git - clang/log
clang
11 years agoObjective-C: Another case of issuing warning about misusing
Fariborz Jahanian [Fri, 7 Jun 2013 20:26:51 +0000 (20:26 +0000)]
Objective-C: Another case of issuing warning about misusing
property auto-synthesis before knowingit it is to be
auto-synthesized. // rdar://14094682

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183556 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPR16273: Weaken a bogus assertion.
Richard Smith [Fri, 7 Jun 2013 20:03:01 +0000 (20:03 +0000)]
PR16273: Weaken a bogus assertion.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183550 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDefine id to the correct type.
Rafael Espindola [Fri, 7 Jun 2013 18:41:01 +0000 (18:41 +0000)]
Define id to the correct type.

Found this while working on pr16247. The fix for that bug will cause clang to
reject this.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183545 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoObjective-C: Removes a bogus warning about auto-synthesis
Fariborz Jahanian [Fri, 7 Jun 2013 18:32:55 +0000 (18:32 +0000)]
Objective-C: Removes a bogus warning about auto-synthesis
of properties. Fixes // rdar://14085456

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183542 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoThread the 'Invalid' out parameter through SourceManager::getSLocEntry() and callees...
Argyrios Kyrtzidis [Fri, 7 Jun 2013 17:57:59 +0000 (17:57 +0000)]
Thread the 'Invalid' out parameter through SourceManager::getSLocEntry() and callees of SourceManager::getSLocEntryByID().

Also add an 'Invalid' check in SourceManager::computeMacroArgsCache().

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183538 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImproved handling of escaped newlines at the token start.
Alexander Kornienko [Fri, 7 Jun 2013 17:45:07 +0000 (17:45 +0000)]
Improved handling of escaped newlines at the token start.

Summary: Remove them from the TokenText as well.

Reviewers: klimek

Reviewed By: klimek

CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D935

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183536 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoHandle Unicode characters in fix-it replacement strings.
Jordan Rose [Fri, 7 Jun 2013 17:16:01 +0000 (17:16 +0000)]
Handle Unicode characters in fix-it replacement strings.

Patch by Sukolsak Sakshuwong!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183535 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agofix up recogtion of block pointer type in my last patch.
Fariborz Jahanian [Fri, 7 Jun 2013 16:07:38 +0000 (16:07 +0000)]
fix up recogtion of block pointer type in my last patch.
// rdar://14085217.
e-This line, and those below, will be ignored--

M    lib/Sema/SemaExpr.cpp

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183531 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFixed calculation of penalty when breaking tokens.
Alexander Kornienko [Fri, 7 Jun 2013 16:02:52 +0000 (16:02 +0000)]
Fixed calculation of penalty when breaking tokens.

Summary:
Introduced two new style parameters: PenaltyBreakComment and
PenaltyBreakString. Add penalty for each character of a breakable token beyond
the column limit (this relates mainly to comments, as they are broken only on
whitespace). Tuned PenaltyBreakComment to prefer comment breaking over breaking
inside most binary expressions.
Fixed a bug that prevented *, & and && from being considered TT_BinaryOperator
in the presense of adjacent comments.

Reviewers: klimek, djasper

Reviewed By: klimek

CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D933

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183530 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdds a test that verifies angle bracket locations.
Manuel Klimek [Fri, 7 Jun 2013 11:27:57 +0000 (11:27 +0000)]
Adds a test that verifies angle bracket locations.

Regression test to make sure TemplateSpecializationTypeLocs have the
correct locations for angle brackets.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183514 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoReverts r183466: "Perform dynamic alignment computations..."
Manuel Klimek [Fri, 7 Jun 2013 11:27:53 +0000 (11:27 +0000)]
Reverts r183466: "Perform dynamic alignment computations..."

This introduces bugs in TemplateSpecializationTypeLoc's angle bracket
locations.

Regression test follows in a subsequent commit.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183513 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoReverting test commit
Faisal Vali [Fri, 7 Jun 2013 05:33:21 +0000 (05:33 +0000)]
Reverting test commit

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183487 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agotest commit
Faisal Vali [Fri, 7 Jun 2013 05:19:33 +0000 (05:19 +0000)]
test commit

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183486 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPR16243: Use CXXThisOverride during template instantiation, and fix up the
Richard Smith [Fri, 7 Jun 2013 02:33:37 +0000 (02:33 +0000)]
PR16243: Use CXXThisOverride during template instantiation, and fix up the
places which weren't setting it up properly. This allows us to get the right
cv-qualifiers for 'this' when it appears outside a method body in a class
template.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183483 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoWhen we're synthesizing copy/move-assignment, we can't form a reference to an
Eli Friedman [Fri, 7 Jun 2013 01:48:56 +0000 (01:48 +0000)]
When we're synthesizing copy/move-assignment, we can't form a reference to an
invalid field; make sure we don't try.  Fixes <rdar://problem/14084171>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183479 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agodocumentation: remove confusing reference to properties.
Adrian Prantl [Fri, 7 Jun 2013 01:10:48 +0000 (01:10 +0000)]
documentation: remove confusing reference to properties.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183475 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoObjC Debug Info: Emit the names of accessors whenever they diverge from
Adrian Prantl [Fri, 7 Jun 2013 01:10:45 +0000 (01:10 +0000)]
ObjC Debug Info: Emit the names of accessors whenever they diverge from
the default names, not just when the isImplicit flag is set.

rdar://problem/14035789

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183474 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImprove documentation.
Adrian Prantl [Fri, 7 Jun 2013 01:10:41 +0000 (01:10 +0000)]
Improve documentation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183473 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoblocks: fixes an ast bug when block pointer variable
Fariborz Jahanian [Fri, 7 Jun 2013 00:48:14 +0000 (00:48 +0000)]
blocks: fixes an ast bug when block pointer variable
is evaluated in a condition expression and then
dereferenced to envoke the block. This is
pr15663 and I applied a slight variation of the
patch with a test case. (patch is from
Arthur O'Dwyer). Also // rdar://14085217

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183471 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDiagnose malformed x86 inline asm using 'y' constraint.
Tim Northover [Fri, 7 Jun 2013 00:04:50 +0000 (00:04 +0000)]
Diagnose malformed x86 inline asm using 'y' constraint.

X86's 'y' inline assembly constraint represents an MMX register, this change
prevents Clang from hitting an assertion when passed an incompatible type to
deal with.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183467 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPerform dynamic alignment computations so that the data in TypeLocs is
Eli Friedman [Fri, 7 Jun 2013 00:04:31 +0000 (00:04 +0000)]
Perform dynamic alignment computations so that the data in TypeLocs is
correctly aligned.  Not performing such computations led to misaligned loads,
which crash on some platforms and are generally bad on other platforms.

The implementation of TypeLocBuilder::pushImpl is rather messy; code using
TypeLocBuilder accidentally assumes that partial TypeLocs are
laid out like a complete TypeLoc.  As a followup, I intend to work on
fixing the TypeLocBuilder API to avoid exposing partial TypeLocs; this should
substantially simplify the implemementation.

Fixes PR16144.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183466 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImplement DR7
David Majnemer [Thu, 6 Jun 2013 23:43:20 +0000 (23:43 +0000)]
Implement DR7

Disallowing deriving from classes that have private virtual base classes
except in instances where the deriving class would be able to cast
itself to the private virtual base via a different derivation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183462 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Address Jordan’s code review for r183451
Anna Zaks [Thu, 6 Jun 2013 22:32:11 +0000 (22:32 +0000)]
[analyzer] Address Jordan’s code review for r183451

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183455 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Ensure that pieces with invalid locations always get removed from the...
Anna Zaks [Thu, 6 Jun 2013 22:02:58 +0000 (22:02 +0000)]
[analyzer] Ensure that pieces with invalid locations always get removed from the BugReport

The function in which we were doing it used to be conditionalized. Add a new unconditional
cleanup step.

This fixes PR16227 (radar://14073870) - a crash when generating html output for one of the test files.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183451 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] fixup the comment
Anna Zaks [Thu, 6 Jun 2013 22:02:55 +0000 (22:02 +0000)]
[analyzer] fixup the comment

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183450 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Simplify edges in a C++11 for-range loop.
Jordan Rose [Thu, 6 Jun 2013 21:53:45 +0000 (21:53 +0000)]
[analyzer; new edges] Simplify edges in a C++11 for-range loop.

Previously our edges were completely broken here; now, the final result
is a very simple set of edges in most cases: one up to the "for" keyword
for context, and one into the body of the loop. This matches the behavior
for ObjC for-in loops.

In the AST, however, CXXForRangeStmts are handled very differently from
ObjCForCollectionStmts. Since they are specified in terms of equivalent
statements in the C++ standard, we actually have implicit AST nodes for
all of the semantic statements. This makes evaluation very easy, but
diagnostic locations a bit trickier. Fortunately, the problem can be
generally defined away by marking all of the implicit statements as
part of the top-level for-range statement.

One of the implicit statements in a for-range statement is the declaration
of implicit iterators __begin and __end. The CFG synthesizes two
separate DeclStmts to match each of these decls, but until now these
synthetic DeclStmts weren't in the function's ParentMap. Now, the CFG
keeps track of its synthetic statements, and the AnalysisDeclContext will
make sure to add them to the ParentMap.

<rdar://problem/14038483>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183449 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[libclang] CMake: add an option to control building libclang as a static library...
Argyrios Kyrtzidis [Thu, 6 Jun 2013 19:51:37 +0000 (19:51 +0000)]
[libclang] CMake: add an option to control building libclang as a static library (off by default)

This avoids building libclang twice by default.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183437 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix incorrect line breaking before trailing block comments.
Daniel Jasper [Thu, 6 Jun 2013 16:08:57 +0000 (16:08 +0000)]
Fix incorrect line breaking before trailing block comments.

Before, clang-format would happily move a trailing block comment to a
new line, which normally changes the perceived binding of that comment.

E.g., it would move:
void f() { /* comment */
  ...
}
to:
void f() {
  /* comment */
  ...
}

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183420 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix a crash with -Wassign-enum, where we didn't adjust the APInt type of the
Joey Gouly [Thu, 6 Jun 2013 13:48:00 +0000 (13:48 +0000)]
Fix a crash with -Wassign-enum, where we didn't adjust the APInt type of the
constant. Also fix some spelling mistakes and formatting issues.

Reviewed by Richard Smith over IRC.

Fixes PR15069.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183409 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoclang-check: add missing space in Makefile
Hans Wennborg [Thu, 6 Jun 2013 13:24:44 +0000 (13:24 +0000)]
clang-check: add missing space in Makefile

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183406 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoclang-check: Link the static analyzer in the Makefile build too.
Benjamin Kramer [Thu, 6 Jun 2013 13:16:41 +0000 (13:16 +0000)]
clang-check: Link the static analyzer in the Makefile build too.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183404 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdd support for static analysis to clang-check
Pavel Labath [Thu, 6 Jun 2013 12:35:43 +0000 (12:35 +0000)]
Add support for static analysis to clang-check

Summary:
This adds a command line argument '-analyze' to clang-check which runs the
clang static analyzer on the source files.

Reviewers: klimek

CC: cfe-commits, revane
Differential Revision: http://llvm-reviews.chandlerc.com/D926

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183399 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoClangTool: strip -o from the command line
Pavel Labath [Thu, 6 Jun 2013 11:52:19 +0000 (11:52 +0000)]
ClangTool: strip -o from the command line

Summary:
This patch creates a new ArgumentsAdjuster, which removes all -o parameters from
the command line. This adjuster is inserted by default into the ClangTool pipeline.

Reviewers: klimek

CC: cfe-commits, revane
Differential Revision: http://llvm-reviews.chandlerc.com/D925

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183398 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImplement DR1270: braces can be elided in all aggregate initialization, not
Richard Smith [Thu, 6 Jun 2013 11:41:05 +0000 (11:41 +0000)]
Implement DR1270: braces can be elided in all aggregate initialization, not
just copy-list-initialization in a variable declaration. This effectively
reverts r142147.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183397 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDisallow reinterpret_cast from pointer to bool on Windows
Hans Wennborg [Thu, 6 Jun 2013 09:16:36 +0000 (09:16 +0000)]
Disallow reinterpret_cast from pointer to bool on Windows

This became allowed by accident in r131201, but triggers an assert.
That patch added an exception to allow conversion from pointers to
narrow integral types for MSVC compatibility. However, a pointer can
already be converted to bool in a civilized manner; allowing conversion
via reinterpret_cast is a bad idea.

Fixes PR16222.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183394 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix clang-format's expression parser for leading }s.
Daniel Jasper [Thu, 6 Jun 2013 09:11:58 +0000 (09:11 +0000)]
Fix clang-format's expression parser for leading }s.

The leading "}" in the construct "} else if (..) {" was confusing the
expression parser. Thus, no fake parentheses were generated and the
indentation was broken in some cases.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183393 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImprove c-style cast detection.
Daniel Jasper [Thu, 6 Jun 2013 08:20:20 +0000 (08:20 +0000)]
Improve c-style cast detection.

Before:
return (my_int) aaaa;
template <> void f<int>(int i)SOME_ANNOTATION;
f("aaaa" SOME_MACRO(aaaa)"aaaa");

After:
return (my_int)aaaa;
template <> void f<int>(int i) SOME_ANNOTATION;
f("aaaa" SOME_MACRO(aaaa) "aaaa");

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183389 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoWhen a static storage duration temporary appears in a constant expression, it
Richard Smith [Thu, 6 Jun 2013 08:19:16 +0000 (08:19 +0000)]
When a static storage duration temporary appears in a constant expression, it
must be initialized by a constant expression (not just a core constant
expression), because we're going to emit it as a global. Core issue for this is
pending.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183388 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDon't create a StringRef from a temporary string.
Richard Trieu [Thu, 6 Jun 2013 02:22:29 +0000 (02:22 +0000)]
Don't create a StringRef from a temporary string.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183372 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Look through ExprWithCleanups to see if an expr's consumed.
Jordan Rose [Thu, 6 Jun 2013 01:57:24 +0000 (01:57 +0000)]
[analyzer] Look through ExprWithCleanups to see if an expr's consumed.

We based decisions during analysis and during path generation on whether
or not an expression is consumed, so if a top-level expression has
cleanups it's important for us to look through that.

<rdar://problem/14076125>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183368 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Improve debug output for PathDiagnosticPieces.
Jordan Rose [Thu, 6 Jun 2013 01:57:19 +0000 (01:57 +0000)]
[analyzer] Improve debug output for PathDiagnosticPieces.

You can now dump a single PathDiagnosticPiece or PathDiagnosticLocation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183367 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoclang/unittests/Format/FormatTest.cpp: Suppress utf8 literals with _MSC_VER. MS cl...
NAKAMURA Takumi [Thu, 6 Jun 2013 01:14:58 +0000 (01:14 +0000)]
clang/unittests/Format/FormatTest.cpp: Suppress utf8 literals with _MSC_VER. MS cl.exe is unaware of BOM-less utf8 source files.

FIXME: Encode Cyrillic and CJK characters below to appease MS compilers.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183366 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Fix a crash that occurs when processing an rvalue array.
Anna Zaks [Thu, 6 Jun 2013 00:19:36 +0000 (00:19 +0000)]
[analyzer] Fix a crash that occurs when processing an rvalue array.

When processing ArrayToPointerDecay, we expect the array to be a location, not a LazyCompoundVal.
Special case the rvalue arrays by using a location to represent them. This case is handled similarly
elsewhere in the code.

Fixes PR16206.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183359 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Don't crash if the top-level entry edge is missing.
Jordan Rose [Thu, 6 Jun 2013 00:12:41 +0000 (00:12 +0000)]
[analyzer; new edges] Don't crash if the top-level entry edge is missing.

We previously asserted that there was a top-level function entry edge, but
if the function decl's location is invalid (or within a macro) this edge
might not exist. Change the assertion to an actual check, and don't drop
the first path piece if it doesn't match.

<rdar://problem/14070304>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183358 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Ignore self-edges, not all edges with the same location.
Jordan Rose [Thu, 6 Jun 2013 00:12:37 +0000 (00:12 +0000)]
[analyzer; new edges] Ignore self-edges, not all edges with the same location.

The edge optimizer needs to see edges for, say, implicit casts (which have
the same source location as their operand) to uniformly simplify the
entire path. However, we still don't want to produce edges from a statement
to /itself/, which could occur when two nodes in a row have the same
statement location.

This necessitated moving the check for redundant notes to after edge
optimization, since the check relies on notes being adjacent in the path.

<rdar://problem/14061675>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183357 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoIf we're splitting the dwarf for fission and don't want to use the
Eric Christopher [Wed, 5 Jun 2013 23:58:15 +0000 (23:58 +0000)]
If we're splitting the dwarf for fission and don't want to use the
integrated assembler then go ahead and still split the dwarf anyhow.

Add two tests, one to exercise existing behavior of not splitting
when we're just emitting assembly files and the other to test
that we split when we're not in integrated as mode.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183355 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoObjective-C: Provide fixit with suggested spelling correction
Fariborz Jahanian [Wed, 5 Jun 2013 18:46:14 +0000 (18:46 +0000)]
Objective-C: Provide fixit with suggested spelling correction
for -Wundeclared-selector warnings. // rdar://14039037

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183331 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPR14763: Debug info for non-trivial record parameters
David Blaikie [Wed, 5 Jun 2013 18:30:31 +0000 (18:30 +0000)]
PR14763: Debug info for non-trivial record parameters

There seems to have been some erroneous code attempting to describe the
ABI of parameters (non-trivial record parameters are passed by
reference). This would break the type of the function (especially when
it caused a mismatch between the type of a declaration & a definition)
causing PR14763 and PR14645.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183329 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoWhen the template specialization header is missing, set a valid source location for...
Argyrios Kyrtzidis [Wed, 5 Jun 2013 17:52:24 +0000 (17:52 +0000)]
When the template specialization header is missing, set a valid source location for the template keyword when recovering.

Otherwise ClassTemplateSpecializationDecl::getSourceRange() will mistakenly consider itself as an implicit partial specialization
and lead to a crash.

Fixes rdar://14063074

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183325 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[ms-cxxabi] Thread GlobalDecls through to CodeGenModule::getFunctionLinkage.
Peter Collingbourne [Wed, 5 Jun 2013 17:49:37 +0000 (17:49 +0000)]
[ms-cxxabi] Thread GlobalDecls through to CodeGenModule::getFunctionLinkage.

This is so that we can give destructor variants different linkage later.

Differential Revision: http://llvm-reviews.chandlerc.com/D819

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183324 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoclang-check: Enable specification of additional compiler arguments
Pavel Labath [Wed, 5 Jun 2013 16:23:30 +0000 (16:23 +0000)]
clang-check: Enable specification of additional compiler arguments

Summary:
This adds two command-line parameters: -extra-arg and -extra-arg-before, which
enable the user to pass additional parameters to the compiler command.

Reviewers: klimek

CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D919

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183320 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[ms-cxxabi] Fix vbptr offsets in memptrs when the vbptr is in an nvbase
Reid Kleckner [Wed, 5 Jun 2013 15:58:29 +0000 (15:58 +0000)]
[ms-cxxabi] Fix vbptr offsets in memptrs when the vbptr is in an nvbase

Also addresses a review comment from John from on r180985 by removing
the "== -1" check, since it's now reusing the correct code which has the
comment.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183318 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoSilence GCC warning.
Benjamin Kramer [Wed, 5 Jun 2013 15:37:50 +0000 (15:37 +0000)]
Silence GCC warning.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183317 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdded test with characters using four bytes in UTF-8 representation.
Alexander Kornienko [Wed, 5 Jun 2013 15:08:20 +0000 (15:08 +0000)]
Added test with characters using four bytes in UTF-8 representation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183315 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoUTF-8 support for clang-format.
Alexander Kornienko [Wed, 5 Jun 2013 14:09:10 +0000 (14:09 +0000)]
UTF-8 support for clang-format.

Summary:
Detect if the file is valid UTF-8, and if this is the case, count code
points instead of just using number of bytes in all (hopefully) places, where
number of columns is needed. In particular, use the new
FormatToken.CodePointCount instead of TokenLength where appropriate.
Changed BreakableToken implementations to respect utf-8 character boundaries
when in utf-8 mode.

Reviewers: klimek, djasper

Reviewed By: djasper

CC: cfe-commits, rsmith, gribozavr
Differential Revision: http://llvm-reviews.chandlerc.com/D918

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183312 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoRecommit r183298+r183300 'Get rid of VTableContext::ComputeMethodVTableIndices()...
Timur Iskhodzhanov [Wed, 5 Jun 2013 14:05:50 +0000 (14:05 +0000)]
Recommit r183298+r183300 'Get rid of VTableContext::ComputeMethodVTableIndices() and VTableContext::getNumVirtualFunctionPointers()'

In r183298, I've used llvm::SmallPtrSet<..., 8> instead of llvm::SmallVector<..., 8> for NewVirtualFunctionsTy by mistake.
This only manifested when a class had more than 8 virtual functions, which wasn't covered by unit-tests

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183310 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago... and actually run it
Timur Iskhodzhanov [Wed, 5 Jun 2013 13:50:24 +0000 (13:50 +0000)]
... and actually run it

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183309 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdd a test for the breakage from r183298
Timur Iskhodzhanov [Wed, 5 Jun 2013 13:49:11 +0000 (13:49 +0000)]
Add a test for the breakage from r183298

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183308 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoUnrevert the tests from r183298 as they pass with both old and new code
Timur Iskhodzhanov [Wed, 5 Jun 2013 13:12:36 +0000 (13:12 +0000)]
Unrevert the tests from r183298 as they pass with both old and new code

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183306 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoRevert r183298 and r183300 as the former broke the virtual function lookup in libcxx...
Timur Iskhodzhanov [Wed, 5 Jun 2013 12:24:46 +0000 (12:24 +0000)]
Revert r183298 and r183300 as the former broke the virtual function lookup in libcxx __locale

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183305 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoMake clang tools ignore -fcolor-diagnostics and -fdiagnostics-color retrieved from...
Alexander Kornienko [Wed, 5 Jun 2013 11:33:11 +0000 (11:33 +0000)]
Make clang tools ignore -fcolor-diagnostics and -fdiagnostics-color retrieved from the compilation database.

Summary:
Clang tools' diagnostic output could be force colored when a command
line from the compilation database contains -fcolor-diagnostics or
-fdiagnostics-color. This is not what we want e.g. for vim integration.

Reviewers: klimek

Reviewed By: klimek

CC: cfe-commits, revane, jordan_rose
Differential Revision: http://llvm-reviews.chandlerc.com/D917

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183304 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoASTTests/DeclTest.cpp: Tweak on msvc target to add "-fno-ms-extensions".
NAKAMURA Takumi [Wed, 5 Jun 2013 06:54:06 +0000 (06:54 +0000)]
ASTTests/DeclTest.cpp: Tweak on msvc target to add "-fno-ms-extensions".

MS extensions don't prefer imaginary suffix even with -std=c++11.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183301 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix fallout from r183298, unused function (CollectPrimaryBases) causing the -Werror...
David Blaikie [Wed, 5 Jun 2013 06:50:40 +0000 (06:50 +0000)]
Fix fallout from r183298, unused function (CollectPrimaryBases) causing the -Werror build to fail

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183300 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix the DUMP_OVERRIDERS macro in the VTableBuilder.cpp
Timur Iskhodzhanov [Wed, 5 Jun 2013 06:40:07 +0000 (06:40 +0000)]
Fix the DUMP_OVERRIDERS macro in the VTableBuilder.cpp

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183299 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoGet rid of VTableContext::ComputeMethodVTableIndices() and VTableContext::getNumVirtu...
Timur Iskhodzhanov [Wed, 5 Jun 2013 06:36:37 +0000 (06:36 +0000)]
Get rid of VTableContext::ComputeMethodVTableIndices() and VTableContext::getNumVirtualFunctionPointers(); also add some tests for the VTable indices output to make sure we don't regress

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183298 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPR16214: Debug Info: -flimit-debug-info doesn't omit definitions for types used via...
David Blaikie [Wed, 5 Jun 2013 05:32:23 +0000 (05:32 +0000)]
PR16214: Debug Info: -flimit-debug-info doesn't omit definitions for types used via typedefs

In an effort to make -flimit-debug-info more consistent I over-shot the
mark & made types used via typedefs never produce definitions in the
debug info (even if the type was used in a way that would require a
definition).

The fix for this is to do exactly what I was hoping to do at some point
- plumb the declaration/definition choice through the various layers of
"CreateType" in CGDebugInfo. In this way we can produce declarations
whenever they are sufficient & definitions otherwise - including when
qualifiers are used, for example (discovered in PR14467). This may not
be complete (there may be other types/situations where we need to
propagate the "declaration/definition" choice) but it lays the basic
foundation which we can enhance in future iterations.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183296 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDefault to PIE and using stack protection on OpenBSD.
Rafael Espindola [Wed, 5 Jun 2013 04:28:55 +0000 (04:28 +0000)]
Default to PIE and using stack protection on OpenBSD.

Patch by Matthew Dempsky.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183295 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoImplement SparcV9ABIInfo::EmitVAArg.
Jakob Stoklund Olesen [Wed, 5 Jun 2013 03:00:18 +0000 (03:00 +0000)]
Implement SparcV9ABIInfo::EmitVAArg.

This could actually be implemented with the LLVM IR va_arg instruction,
but it doesn't seem to offer any advantages over accessing the va_list
pointer directly.

Using the va_list pointer directly makes it possible to perform type
coercion directly from the argument array, and the va_list updates are
exposed to the optimizers.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183292 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix CoerceIntOrPtrToIntOrPtr on big-endian targets.
Jakob Stoklund Olesen [Wed, 5 Jun 2013 03:00:13 +0000 (03:00 +0000)]
Fix CoerceIntOrPtrToIntOrPtr on big-endian targets.

Type coercion for argument passing is equivalent to storing the source
type and loading the destination type from the same pointer. On
big-endian targets, this means that the high bits of integers are
preserved.

This patch fixes the CoerceIntOrPtrToIntOrPtr() function on big-endian
targets by inserting the required shift instructions to preserve the
high bits instead of the low bits.

This is used by SparcABIInfo when passing small structs in the high bits
of registers.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183291 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoHeed ABIArgInfo::getInReg() for return values.
Jakob Stoklund Olesen [Wed, 5 Jun 2013 03:00:09 +0000 (03:00 +0000)]
Heed ABIArgInfo::getInReg() for return values.

The 'inreg' attribute can also be applied to function return values in
LLVM IR. The SPARC v9 backend is using the flag when returning structs
containing 32-bit floats.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183290 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoModel temporary lifetime-extension explicitly in the AST. Use this model to
Richard Smith [Wed, 5 Jun 2013 00:46:14 +0000 (00:46 +0000)]
Model temporary lifetime-extension explicitly in the AST. Use this model to
handle temporaries which have been lifetime-extended to static storage duration
within constant expressions. This correctly handles nested lifetime extension
(through reference members of aggregates in aggregate initializers) but
non-constant-expression emission hasn't yet been updated to do the same.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183283 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[ms-cxxabi] Factor out some loops into helpers for readability
Reid Kleckner [Tue, 4 Jun 2013 21:32:29 +0000 (21:32 +0000)]
[ms-cxxabi] Factor out some loops into helpers for readability

No functionality change, covered by the existing virtual base adjustment
tests.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183251 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[libclang] When annotating tokens, don't override a property annotation with a getter...
Argyrios Kyrtzidis [Tue, 4 Jun 2013 18:24:30 +0000 (18:24 +0000)]
[libclang] When annotating tokens, don't override a property annotation with a getter/setter method annotation.

rdar://13764549

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183242 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAnalysis: Add a CFG successor to a SwitchStmt if it is both empty and fully covered
David Majnemer [Tue, 4 Jun 2013 17:38:44 +0000 (17:38 +0000)]
Analysis: Add a CFG successor to a SwitchStmt if it is both empty and fully covered

Consider the case where a SwitchStmt satisfied isAllEnumCasesCovered()
as well as having no cases at all (i.e. the enum it covers has no
enumerators).

In this case, we should add a successor to repair the CFG.

This fixes PR16212.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183237 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdded Lexer::getBufferEnd().
Alexander Kornienko [Tue, 4 Jun 2013 16:58:03 +0000 (16:58 +0000)]
Added Lexer::getBufferEnd().

Summary:
There's Lexer::getBufferStart(), and we need getBufferEnd() to access
the whole input buffer in clang::format::reformat. We don't want to
rely on the fact that the Lexer::BufferEnd always points to '\0', as there can
be embedded '\0's as well.

Reviewers: jordan_rose

Reviewed By: jordan_rose

CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D916

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183236 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoObjective-C encoding. Fixes up encodeing for
Fariborz Jahanian [Tue, 4 Jun 2013 16:04:37 +0000 (16:04 +0000)]
Objective-C encoding. Fixes up encodeing for
arrays of empty structs. // rdar://14053082
(also pr13062).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183234 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoParser/Registry argument enhancements.
Samuel Benzaquen [Tue, 4 Jun 2013 15:46:22 +0000 (15:46 +0000)]
Parser/Registry argument enhancements.

Summary:
 Parser/Registry argument enhancements.
  - 2 argument support.
  - unsigned values support.

Reviewers: klimek

CC: cfe-commits, revane
Differential Revision: http://llvm-reviews.chandlerc.com/D915

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183231 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoEnables chaining of argument adjusters in clang tools.
Manuel Klimek [Tue, 4 Jun 2013 14:44:44 +0000 (14:44 +0000)]
Enables chaining of argument adjusters in clang tools.

This enables changing clang-check to get extra arguments.

Patch by Pavel Labath.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183227 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix a typo (chek => check)
Sylvestre Ledru [Tue, 4 Jun 2013 13:43:50 +0000 (13:43 +0000)]
Fix a typo (chek => check)

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183223 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix linkage computation for local types in template functions.
Rafael Espindola [Tue, 4 Jun 2013 13:43:35 +0000 (13:43 +0000)]
Fix linkage computation for local types in template functions.

Template functions (and member functions of class templates) present the same
problem as inline functions. They need to be uniqued, so we need to assign
VisibleNoLinkage linkage to types defined in them.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183222 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix typo in comment. Found by -Wdocumentation.
Benjamin Kramer [Tue, 4 Jun 2013 09:13:21 +0000 (09:13 +0000)]
Fix typo in comment. Found by -Wdocumentation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183217 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoReset the release notes for the 3.4 release.
Bill Wendling [Tue, 4 Jun 2013 06:17:46 +0000 (06:17 +0000)]
Reset the release notes for the 3.4 release.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183211 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoRephrase asm_mismatched_size_modifier diagnostic. <rdar://problem/14050339>
Bob Wilson [Tue, 4 Jun 2013 06:10:09 +0000 (06:10 +0000)]
Rephrase asm_mismatched_size_modifier diagnostic. <rdar://problem/14050339>

The text of this diagnostic was unnecessarily specific to the current ARM
implementation of validateConstraintModifier, and it gave a potentially
confusing suggestion for fixing the problem. The ARM-specific issue is not
a big deal since that is the only target that currently does any checking of
asm operand modifiers, but until my change in 183172 it was still wrong for
output operands in the way that it referred to the value being truncated when
put into a register, since output operands are retrieved from the registers
instead of being put into them. The bigger problem is that its suggestion to
"use a modifier" is wrong and confusing in the case where a "q" modifier is
incorrectly used with an "r" constraint. In that case, the solution might
well be to remove the modifier or perhaps change the constraint. It's better
to just leave the diagnostic message more generic.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183209 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoRemove some unreachable (and wrong) code and replace it with an assertion.
Richard Smith [Tue, 4 Jun 2013 04:45:03 +0000 (04:45 +0000)]
Remove some unreachable (and wrong) code and replace it with an assertion.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183206 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdding support for MSVC #pragma detect_mismatch functionality by emitting a FAILIFMIS...
Aaron Ballman [Tue, 4 Jun 2013 02:07:14 +0000 (02:07 +0000)]
Adding support for MSVC #pragma detect_mismatch functionality by emitting a FAILIFMISMATCH linker command into the object file.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183178 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoBound member function diagnostic - suggest no-args calls and note overload candidates
David Blaikie [Tue, 4 Jun 2013 00:28:46 +0000 (00:28 +0000)]
Bound member function diagnostic - suggest no-args calls and note overload candidates

Still missing cases for templates, but this is a step in the right
direction. Also omits suggestions that would be ambiguous (eg: void
func(int = 0); + void func(float = 0); func;)

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183173 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoDo not report -Wasm-operand-widths for ARM output operands. <rdar://14050337>
Bob Wilson [Mon, 3 Jun 2013 23:57:13 +0000 (23:57 +0000)]
Do not report -Wasm-operand-widths for ARM output operands. <rdar://14050337>

We're getting reports of this warning getting triggered in cases where it
is not adding any value. There is no asm operand modifier that you can use
to silence it, and there's really nothing wrong with having an LDRB, for
example, with a "char" output.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183172 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer] Enable the new edge algorithm by default.
Jordan Rose [Mon, 3 Jun 2013 23:00:19 +0000 (23:00 +0000)]
[analyzer] Enable the new edge algorithm by default.

...but don't yet migrate over the existing plist tests. Some of these
would be trivial to migrate; others could use a bit of inspection first.
In any case, though, the new edge algorithm seems to have proven itself,
and we'd like more coverage (and more usage) of it going forwards.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183165 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Omit subexpression back-edges that span multiple lines.
Jordan Rose [Mon, 3 Jun 2013 23:00:09 +0000 (23:00 +0000)]
[analyzer; new edges] Omit subexpression back-edges that span multiple lines.

A.1 -> A -> B
becomes
A.1 -> B

This only applies if there's an edge from a subexpression to its parent
expression, and that is immediately followed by another edge from the
parent expression to a subsequent expression. Normally this is useful for
bringing the edges back to the left side of the code, but when the
subexpression is on a different line the backedge ends up looking strange,
and may even obscure code. In these cases, it's better to just continue
to the next top-level statement.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183164 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Don't eliminate subexpr edge cycles if the line is long.
Jordan Rose [Mon, 3 Jun 2013 23:00:05 +0000 (23:00 +0000)]
[analyzer; new edges] Don't eliminate subexpr edge cycles if the line is long.

Specifically, if the line is over 80 characters, or if the top-level
statement spans mulitple lines, we should preserve sub-expression edges
even if they form a simple cycle as described in the last commit, because
it's harder to infer what's going on than it is for shorter lines.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183163 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Eliminate "cycle edges" for a single subexpression.
Jordan Rose [Mon, 3 Jun 2013 23:00:00 +0000 (23:00 +0000)]
[analyzer; new edges] Eliminate "cycle edges" for a single subexpression.

Generating context arrows can result in quite a few arrows surrounding a
relatively simple expression, often containing only a single path note.

|
1 +--2---+
v/       v
auto m = new m // 3 (the path note)
|\       |
5 +--4---+
v

Note also that 5 and 1 are two ends of the "same" arrow, i.e. they go from
event to event. 3 is not an arrow but the path note itself.

Now, if we see a pair of edges like 2 and 4---where 4 is the reverse of 2
and there is optionally a single path note between them---we will
eliminate /both/ edges. Anything more complicated will be left as is
(more edges involved, an inlined call, etc).

The next commit will refine this to preserve the arrows in a larger
expression, so that we don't lose all context.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183162 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Extra test case.
Jordan Rose [Mon, 3 Jun 2013 22:59:56 +0000 (22:59 +0000)]
[analyzer; new edges] Extra test case.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183161 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Improve enclosing contexts for logical expressions.
Jordan Rose [Mon, 3 Jun 2013 22:59:53 +0000 (22:59 +0000)]
[analyzer; new edges] Improve enclosing contexts for logical expressions.

The old edge builder didn't have a notion of nested statement contexts,
so there was no special treatment of a logical operator inside an if
(or inside another logical operator). The new edge builder always tries
to establish the full context up to the top-level statement, so it's
important to know how much context has been established already rather
than just checking the innermost context.

This restores some of the old behavior for the old edge generation:
the context of a logical operator's non-controlling expression is the
subexpression in the old edge algorithm, but the entire operator
expression in the new algorithm.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183160 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Include context for edges to sub-expressions.
Jordan Rose [Mon, 3 Jun 2013 22:59:48 +0000 (22:59 +0000)]
[analyzer; new edges] Include context for edges to sub-expressions.

The current edge-generation algorithm sometimes creates edges from a
top-level statement A to a sub-expression B.1 that's not at the start of B.
This creates a "swoosh" effect where the arrow is drawn on top of the
text at the start of B. In these cases, the results are clearer if we see
an edge from A to B, then another one from B to B.1.

Admittedly, this does create a /lot/ of arrows, some of which merely hop
into a subexpression and then out again for a single note. The next commit
will eliminate these if the subexpression is simple enough.

This updates and reuses some of the infrastructure from the old edge-
generation algorithm to find the "enclosing statement" context for a
given expression. One change in particular marks the context of the
LHS or RHS of a logical binary operator (&&, ||) as the entire operator
expression, rather than the subexpression itself. This matches our behavior
for ?:, and allows us to handle nested context information.

<rdar://problem/13902816>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183159 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[analyzer; new edges] Include a top-level function entry edge while optimizing.
Jordan Rose [Mon, 3 Jun 2013 22:59:45 +0000 (22:59 +0000)]
[analyzer; new edges] Include a top-level function entry edge while optimizing.

Although we don't want to show a function entry edge for a top-level path,
having it makes optimizing edges a little more uniform.

This does not affect any edges now, but will affect context edge generation
(next commit).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183158 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoCFG: In a DeclStmt, skip anything that's not a VarDecl.
Jordan Rose [Mon, 3 Jun 2013 22:59:41 +0000 (22:59 +0000)]
CFG: In a DeclStmt, skip anything that's not a VarDecl.

Neither the compiler nor the analyzer are doing anything with non-VarDecl
decls in the CFG, and having them there creates extra nodes in the
analyzer's path diagnostics. Simplify the CFG (and the path edges) by
simply leaving them out. We can always add interesting decls back in when
they become relevant.

Note that this only affects decls declared in a DeclStmt, and then only
those that appear within a function body.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183157 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoPacify compilers that think you can drop off a fully covered switch.
Benjamin Kramer [Mon, 3 Jun 2013 21:26:13 +0000 (21:26 +0000)]
Pacify compilers that think you can drop off a fully covered switch.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183155 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoUnresolvedMemberExpr should have an expr location of the member (not the base)
David Blaikie [Mon, 3 Jun 2013 19:46:44 +0000 (19:46 +0000)]
UnresolvedMemberExpr should have an expr location of the member (not the base)

This matches the behavior of MemberExpr and makes diagnostics such as
"reference to non-static member function must be called" more legible in
the case that the base & member are split over multiple lines (prior to
this change the diagnostic would point to the base, not the member -
making it very unclear in chained multi-line builder-style calls)

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183149 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago__uint128_t is only available on 64 bit platforms, only test it when it's around.
Benjamin Kramer [Mon, 3 Jun 2013 19:37:18 +0000 (19:37 +0000)]
__uint128_t is only available on 64 bit platforms, only test it when it's around.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183146 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoAdd support for .bind("foo") expressions on the dynamic matchers.
Samuel Benzaquen [Mon, 3 Jun 2013 19:31:08 +0000 (19:31 +0000)]
Add support for .bind("foo") expressions on the dynamic matchers.

Summary: Add support on the parser, registry, and DynTypedMatcher for binding IDs dynamically.

Reviewers: klimek

CC: cfe-commits, revane
Differential Revision: http://llvm-reviews.chandlerc.com/D911

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@183144 91177308-0d34-0410-b5e6-96231b3b80d8