design limitation in how we handle Objective-C class extensions.  This was causing the CursorVisitor
to essentially visit an @property twice (once in the @interface, the other in the class extension).
Fixes <rdar://problem/7410145>.

llvm-svn: 104055

there it may work everywhere...

llvm-svn: 104053

llvm-svn: 104052

Continuously refine the register class of REG_SEQUENCE def with all the source registers and sub-register indices.

llvm-svn: 104051

Fix PR7162: Use source register classes and sub-indices to determine the correct register class of the definitions of REG_SEQUENCE.

llvm-svn: 104050

llvm-svn: 104043

llvm-svn: 104041

CurPtr[0] == '\n' when testing the character after a "0b" when looking
to see if it part of a something like "jmp 0b".

llvm-svn: 104039

correctly.  The Lexer was incorrectly eating the newline casusing it to branch
to address 0.  Updated the test case to use a "0:" label and a branch to "0b".

llvm-svn: 104038

llvm-svn: 104037

llvm-svn: 104035

The old approach was wrong. It had an off-by-one error.

llvm-svn: 104034

Also, create symbol data for LHS of assignment, to match 'as' symbol ordering better.

llvm-svn: 104033

llvm-svn: 104032

llvm-svn: 104031

  X86::ADC32ri $0, %eax
to
  X86::ADC32i32 $0

llvm-svn: 104030

llvm-svn: 104029

llvm-svn: 104027

llvm-svn: 104026

Correctly initialize bases with member pointers. This should fix PR6441 but that test case is a bit weird and I'd like to investigate further before closing that bug.

llvm-svn: 104025

non-function-local declarations with names similar to what the user
typed. For example, this allows us to correct 'supper' to 'super' in
an Objective-C message send, even though the C function 'isupper' has
the same edit distance.

llvm-svn: 104023

consider "super" as a candidate whenever we're parsing an expression
within an Objective-C method in an interface that has a superclass. At
some point, we'd like to give "super" a little edge over non-local
names; that will come later.

llvm-svn: 104022

Update the comment.

llvm-svn: 104021

- Allocate MCLabels in the context so they don't leak.
- Avoid duplicated densemap lookup.

llvm-svn: 104020

llvm-svn: 104019

Also rename ABSTRACT to ABSTRACT_STMT

llvm-svn: 104018

Also rename ABSTRACT to ABSTRACT_STMT, in keeping with the other .def
files

llvm-svn: 104017

for Visual Studio 2010.  It also adds a registry search
for the Express edition,", from Steven Watanabe!

llvm-svn: 104015

Revert much of the implementation of C++98/03 [temp.friend]p5 in
r103943 and its follow-ons r103948 and r103952. While our
implementation was technically correct, other compilers don't seem to
implement this paragraph (which forces the instantiation of friend
functions defined in a class template when a class template
specialization is instantiated), and doing so broke a bunch of Boost
libraries. 

Since this behavior has changed in C++0x (which instantiates the
friend function definitions when they are used), we're going to skip
the nowhere-implemented C++98/03 semantics and go straight to the
C++0x semantics.

This commit is a band-aid to get Boost up and running again. It
doesn't really fix PR6952 (which this commit un-fixes), but it does
deal with the way Boost.Units abuses this particular paragraph.

llvm-svn: 104014

llvm-svn: 104013

Start laying out bases as individual fields. We still use ugly i8 arrays but this is a step in the right direction.

llvm-svn: 104012

llvm-svn: 104011

If a switch condition is constant, warn if there's no case for it.

Constant switch conditions do come up in reasonable template code.

llvm-svn: 104010

llvm-svn: 104008

number of times the analyzer will go through a loop.

llvm-svn: 104007

llvm-svn: 104006

llvm-svn: 104005

FIX PR7158. SimplifyVBinOp was asserting when it fails to constant fold (op (build_vector), (build_vector)). 

llvm-svn: 104004

llvm-svn: 104001

a temporary with a noreturn destructor has been created. Fixes PR6884 for now.

llvm-svn: 104000