PointerUnion == PointerUnion does not do what I thought it did. Also, fix a thinko in a PointerUnion::get call.

llvm-svn: 90719

http://llvm.org/viewvc/llvm-project?view=rev&revision=81871
http://llvm.org/viewvc/llvm-project?view=rev&revision=81936
http://llvm.org/viewvc/llvm-project?view=rev&revision=81945

llvm-svn: 90718

to fix it ATM.

llvm-svn: 90717

llvm-svn: 90716

llvm::DenseMap, for a 20% performance improvement in the
Cocoa-big-list performance benchmark.

llvm-svn: 90715

llvm-svn: 90714

intended.  On the first testcase in the bug, we now produce:

cxx-decl.cpp:12:2: error: unexpected ':' in nested name specifier
y:a a2;
 ^
 ::

instead of:

t.cc:8:1: error: C++ requires a type specifier for all declarations
x:a a2;
^
t.cc:8:2: error: invalid token after top level declarator
x:a a2;
 ^
 ;
t.cc:9:11: error: use of undeclared identifier 'a2'
x::a a3 = a2;
          ^

llvm-svn: 90713

llvm-svn: 90712

llvm-svn: 90711

llvm-svn: 90710

llvm-svn: 90709

llvm-svn: 90706

and I can't figure out how to write it more portably.

llvm-svn: 90705

Document that CompilerInvocation::createDiagnostics keeps a reference to the DiagnosticOptions, and update callers to make sure they don't pass in a temporary.

llvm-svn: 90704

llvm-svn: 90703

Fixes PR5695.

llvm-svn: 90702

Also, update SourceManager.h doxyments for getBuffer() to reflect reality.

llvm-svn: 90701

Change Preprocessor::EnterSourceFile to make ErrorStr non-optional, clients should be forced to deal with error conditions.

llvm-svn: 90700

files.
 - The issue is that PCH uses a stat cache, which may reference files which have
   been deleted or moved. In such cases ContentCache::getBuffer was returning 0
   but most clients are incapable of dealing with this (i.e., they don't).

   For the time being, resolve this issue by just making up some invalid file
   contents and. Eventually we should detect that we are in an inconsistent
   situation and error out with a nice message that the PCH is out of date.

llvm-svn: 90699

http://llvm.org/viewvc/llvm-project?view=rev&revision=86026

Note: The 'improved debugging' changes weren't integrated (since they were later reverted, since they didn't improve debugging).
llvm-svn: 90693

http://llvm.org/viewvc/llvm-project?view=rev&revision=82174

llvm-svn: 90692

http://llvm.org/viewvc/llvm-project?view=rev&revision=72893

llvm-svn: 90690

llvm-svn: 90689

http://llvm.org/viewvc/llvm-project?view=rev&revision=71473

llvm-svn: 90688

llvm-svn: 90687

Pass the desired vtable linkage to GenerateVtable directly. Only call MaybeMarkVirtualImplicitMembersReferenced for non-inline functions.

llvm-svn: 90686

llvm-svn: 90684

it gets called during template instantiation.

llvm-svn: 90682

Only emit the vtable definition if the class has a key function and we're emitting it, or if the class doesn't have a key function and we are emitting the complete constructor.

llvm-svn: 90681

llvm-svn: 90680

llvm-svn: 90679

llvm-svn: 90678

http://llvm.org/viewvc/llvm-project?view=rev&revision=71225
http://llvm.org/viewvc/llvm-project?view=rev&revision=73207
http://llvm.org/viewvc/llvm-project?view=rev&revision=73414

llvm-svn: 90677

CodeGenModule::GenerateVtable now returns a pointer directly to the vtable and not to the address point.

llvm-svn: 90676

llvm-svn: 90675

llvm-svn: 90674

llvm-svn: 90673

Let the VTT builder pretend that getVtable returns a pointer to the vtable and not to the vtable address point.

llvm-svn: 90672

llvm-svn: 90671

llvm-svn: 90670