Thanks to Duncan for noticing this

llvm-svn: 68518

Tweak Sema::ActOnInstanceMessage() to look for a class method when dealing with qualified id's. This change is motivated by our desire to not support the "Class<foo>" idiom. Note that the change makes perfect sense (since all ObjC classes are also id/instances).

This allow us to document a simple migration path...change "Class <foo>" to "id <foo>".

This effects: 
- <rdar://problem/6761939> TASK: File source change radars for "qualified Class" errors
- <rdar://problem/6761864> Protocol qualified Class is unsupported

llvm-svn: 68517

llvm-svn: 68516

llvm-svn: 68515

This fixes <rdar://problem/6757102> clang type for @"xxx" is "NSConstantString *" (GCC type is "NSString *").

llvm-svn: 68514

exception with this attribute we don't need to emit a weak definition
for the exception type information.

llvm-svn: 68513

llvm-svn: 68512

GCC ignores macro definitions after \n's or that real code depends
on this.

llvm-svn: 68511

 - Changed method names to match gcc (categories names still aren't
   mangled in).

 - Expose correct name for class and metadata symbols (although
   -fvisibility=hidden isn't yet correct).

 - Remove several things from llvm.used that didn't need to be there
   (I suspect this can still be trimmed).

 - Don't use asm-prefix extension for _objc_empty_{cache,vtable} (not
   needed).

 - Hide EH type class info with -fvisibility=hidden

 - Change setGlobal[Option]Visibility to not change the visibility of
   functions with internal linkage.

llvm-svn: 68510

llvm-svn: 68505

are considered 'live'. This hack isn't needed anymore because we have a
separation in the path-sensitive analyzer between variable names and bindings;
the analyzer can continue to reason about the storage of a variable after its
name is no longer directly referenced.  Now the live variables analysis literally means "is this name live".

Along this line, update the dead stores checker to explicitly look for variables
whose values have escaped.

llvm-svn: 68504

Fixes rdar://6759546, some code that actually used __va_copy because they
thought it was "more portable".

llvm-svn: 68503

llvm-svn: 68502

llvm-svn: 68500

llvm-svn: 68497

llvm-svn: 68496

llvm-svn: 68495

and are even set in C mode.  As such, move them to Targets.cpp.

__OBJC_GC__ is also darwin specific, but seems reasonable to always
define it when in objc-gc mode.

This fixes rdar://6761450

llvm-svn: 68494

had too many false positives in the analyzer.

llvm-svn: 68492

 - Ignore some more -W options and -[fm] options which we can somewhat
   safely ignore.

 - Recognize that -W is an alias for -Wextra

W: -Wdeclaration-after-statement -Werror-implicit-function-declaration -Wfour-char-constants -Winit-self -Wmissing-format-attribute -Wno-#warnings -Wno-comment -Wno-long-long -Wno-variadic-macros -Wold-style-definition -Wstrict-prototypes -Wunused-parameter

f: -fconstant-cfstrings -fdollars-in-identifiers -finline -finline-functions -fno-inline -fno-keep-inline-functions -fno-strict-aliasing -fobjc-atdefs -fobjc-call-cxx-cdtors -fobjc-new-property -fstack-protector

m: -mconstant-cfstrings -mfix-and-continue
llvm-svn: 68487

them by reference, instead of packing each byte into a
smallvector.

llvm-svn: 68486

llvm-svn: 68485

-mtriple).

llvm-svn: 68484

pieces between block entrance and block end unless necessary.

llvm-svn: 68483

__sync_bool_compare_and_swap.
 - <rdar://problem/6762223> [sema/irgen] support
   __sync_bool_compare_and_swap and __sync_add_and_fetch

llvm-svn: 68482

llvm-svn: 68481

llvm-svn: 68479

Use the new getFunctionInfo that takes a BlockPointerType parameter, and get rid of getBlockFunctionType from CGBlocks.cpp

llvm-svn: 68478

ExplodedGraph backwards. That may inadvertently result in reverse control-flow
edges in the PathDiagostic.

llvm-svn: 68477

llvm-svn: 68476

Be kind to so many projects which are doing this (and be
like gcc).

llvm-svn: 68474

contexts".  This allows us to use a stack of contexts to keep track of what control-flow pieces to include when exiting blocks like 'if', 'for', etc.

llvm-svn: 68473

llvm-svn: 68472

llvm-svn: 68471

llvm-svn: 68470

without converting each byte to a uint64_t to stick in a SmallVector.

llvm-svn: 68469

This will simplify clang adoption, and is probably better "etiquette" (since gcc has always accepted this idiom without warning). Once we are over the adoption hurdle, we can turn this into an error.

llvm-svn: 68468

llvm-svn: 68466

elements in a form that is efficient for the reader to just get a
pointer in memory and start reading.  APIs to do efficient reading
and writing are still todo.

llvm-svn: 68465

Patch by Richard Pennington.

llvm-svn: 68464