Hi Michael, Ted
This change fixes the stackwalker to cache the NOT_FOUND results from a symbol
supplier. Since the StackWalker is instantiated once per minidump from
MinidumpProcessor, it should cut down on some latency when symbols aren't
available for a given module.
Thanks for the review - checked in after addressing all comments. http://breakpad.appspot.com/64001/diff/1/4 File src/google_breakpad/processor/stackwalker.h (right): ...
Thanks for the review - checked in after addressing all comments.
http://breakpad.appspot.com/64001/diff/1/4
File src/google_breakpad/processor/stackwalker.h (right):
http://breakpad.appspot.com/64001/diff/1/4#newcode147
Line 147: set<const CodeModule*> no_symbol_modules_;
On 2010/02/25 13:36:19, Ted Mielczarek wrote:
> It seems a little weird to be keying on a pointer value here. If someone made
a
> copy of a CodeModule (using Copy()), then the new module wouldn't be found
here.
> Why not key on a string and use module->code_file() instead?
Done.
http://breakpad.appspot.com/64001/diff/1/3
File src/processor/minidump_processor_unittest.cc (right):
http://breakpad.appspot.com/64001/diff/1/3#newcode165
Line 165: class CountingSymbolSupplier : public SymbolSupplier {
On 2010/02/25 13:36:19, Ted Mielczarek wrote:
> This feels like it'd be cleaner with a Mock class and some expectations.
Done.
http://breakpad.appspot.com/64001/diff/1/3#newcode248
Line 248: printf("%s: %d\n", it->first.c_str(), it->second);
On 2010/02/25 13:36:19, Ted Mielczarek wrote:
> Seems like unnecessary test spew.
removed
Issue 64001: Fix to cache NOT_FOUND results from symbol supplier on a per-minidump basis
(Closed)
Created 14 years, 6 months ago by mochalatte
Modified 14 years, 5 months ago
Reviewers: brdevmn, Ted Mielczarek
Base URL: http://google-breakpad.googlecode.com/svn/trunk/
Comments: 6