Change PrintFunctions to print functions in address order and not print duplicate public functions.

Change PrintFunctions to print functions in address order and not print
duplicate public functions.

BUG=427
R=mark@chromium.org

Committed: https://code.google.com/p/google-breakpad/source/detail?r=1319

[Will Harris]

This code now produces almost identical symbols to the original code.  I've
manually tested with the crash dumps that were failing before.

There's a few bugs/features in DIA that impacted me here:

1.  Breakpad's GetSymbolFunctionName function didn't work with the output from
DIA 10.0, but worked fine with DIA 8.0 and DIA 12.0 - so I'm now linking with
DIA 12.0 (VS2013)
2.  DIA does not handle correctly the case when you are trying to get private
symbol information on an address that only contains public symbol information. 
This appears to be the underlying cause of the Win64 hangs
(https://connect.microsoft.com/VisualStudio/feedback/details/722366/).  To solve
this I store which addresses are public symbols and do not call findSymbolByRVA
with SymTagFunction on those addresses.
3.  DIA 12.0 fixes a long standing bug where SymTagData children of a
SymTagFunction were an incorrect type when the function was being called using
LocIsEnregistered parameters, this was causing the GetFunctionStackParamSize to
return the wrong stack size when the function was only using registers to pass
arguments.

[Mark Mentovai]

LGTM

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
File src/common/windows/pdb_source_line_writer.cc (right):

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:350: std::set<DWORD> rvas;
Let’s move this down so it’s not declared until closer to its first use. Right
before line 362 looks good.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:351: std::set<DWORD>
public_only_rvas;
And this one, right before line 385.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:408: CComPtr<IDiaSymbol> symbol =
NULL;
This doesn’t need to be declared until inside the loop.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:438: fprintf(stderr,
"findSymbolByRVA failed on the rva\n");
Line 425 identified the findSymbolByRVA as SymTagFunction, so this one should
identify it as SymTagPublicSymbol.

[Will Harris]

Also added tests.  These are now back to exactly matching the original test
output, except the stack size for __fastcall functions is now correctly 0.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
File src/common/windows/pdb_source_line_writer.cc (right):

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:350: std::set<DWORD> rvas;
On 2014/04/23 13:59:33, Mark Mentovai wrote:
> Let’s move this down so it’s not declared until closer to its first use. Right
> before line 362 looks good.

Done.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:351: std::set<DWORD>
public_only_rvas;
On 2014/04/23 13:59:33, Mark Mentovai wrote:
> And this one, right before line 385.

Done.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:408: CComPtr<IDiaSymbol> symbol =
NULL;
On 2014/04/23 13:59:33, Mark Mentovai wrote:
> This doesn’t need to be declared until inside the loop.

Done.

https://breakpad.appspot.com/1634002/diff/1/src/common/windows/pdb_source_lin...
src/common/windows/pdb_source_line_writer.cc:438: fprintf(stderr,
"findSymbolByRVA failed on the rva\n");
On 2014/04/23 13:59:33, Mark Mentovai wrote:
> Line 425 identified the findSymbolByRVA as SymTagFunction, so this one should
> identify it as SymTagPublicSymbol.

Done.

[Will Harris]

Committed patchset #2 manually as r1319 (presubmit successful).