Added the base exploitability module for windows.

Added the base exploitability module for windows. This only adds the very basic
exception type based analysis for now.

BUG=NONE
TEST=MinidumpProcessorTest.TestExploitilityEngine

[Cris Neckar]

Neal, this just adds the basic tests for the windows platform. If there is
someone else I can push these reviews off on let me know.

Thank you for all your help :)

[mochalatte]

http://breakpad.appspot.com/188001/diff/2001/3004
File src/processor/exploitability_win.cc (right):

http://breakpad.appspot.com/188001/diff/2001/3004#newcode41
Line 41: #include "google_breakpad/common/minidump_exception_win32.h"
alphabetical order for includes

http://breakpad.appspot.com/188001/diff/2001/3004#newcode46
Line 46: // from null.
does 4096 come from a common page size from modern OS, or some other estimate?

http://breakpad.appspot.com/188001/diff/2001/3004#newcode89
Line 89: // These exceptions will typically mean that we have jumped where we
shouldn't.
80 column limit

http://breakpad.appspot.com/188001/diff/2001/3004#newcode133
Line 133: case 8:
are there constants you can use for this and the weights?

http://breakpad.appspot.com/188001/diff/2001/3004#newcode162
Line 162: } // namespace google_breakpad
two spaces before inline comments, e.g. "}  // namespace..."

[Cris Neckar]

thanks for the review

http://breakpad.appspot.com/188001/diff/2001/3004
File src/processor/exploitability_win.cc (right):

http://breakpad.appspot.com/188001/diff/2001/3004#newcode41
Line 41: #include "google_breakpad/common/minidump_exception_win32.h"
On 2010/09/14 21:11:36, mochalatte wrote:
> alphabetical order for includes

Done.

http://breakpad.appspot.com/188001/diff/2001/3004#newcode46
Line 46: // from null.
On 2010/09/14 21:11:36, mochalatte wrote:
> does 4096 come from a common page size from modern OS, or some other estimate?

This is sort of an arbitrary number I chose as a maximum offset to check
against. The theory is that anything below (0 + this offset) is likely just a
null dereference instead of a bad address. 

The fact that I decided on a page is sort of arbitrary but sort of based on the
idea that most object etc won't be larger than a page. We could tie it directly
to the page size from the dump but I think it is likely overkill as it will be a
guess regardless.

http://breakpad.appspot.com/188001/diff/2001/3004#newcode89
Line 89: // These exceptions will typically mean that we have jumped where we
shouldn't.
On 2010/09/14 21:11:36, mochalatte wrote:
> 80 column limit

Done.

http://breakpad.appspot.com/188001/diff/2001/3004#newcode133
Line 133: case 8:
On 2010/09/14 21:11:36, mochalatte wrote:
> are there constants you can use for this and the weights?

I restructured another place in the code that has similar logic (i think I wrote
that originally as well) and made an enum for these next to the windows
exception definitions.

http://breakpad.appspot.com/188001/diff/2001/3004#newcode162
Line 162: } // namespace google_breakpad
On 2010/09/14 21:11:36, mochalatte wrote:
> two spaces before inline comments, e.g. "}  // namespace..."

Done.

[Cris Neckar]

hold off on re-reviewing I think I uploaded from the wrong checkout

[Cris Neckar]

My checkout state was screwed up and I had to resubmit this as a new CL. The
comments from this issue apply to the following

http://breakpad.appspot.com/189001/show

If you don't mind can you review/land that one.

Sorry for the mix up.

[mochalatte]

can you close this issue now?
On 2010/09/14 23:15:00, Cris Neckar wrote:
> My checkout state was screwed up and I had to resubmit this as a new CL. The
> comments from this issue apply to the following
> 
> http://breakpad.appspot.com/189001/show
> 
> If you don't mind can you review/land that one.
> 
> Sorry for the mix up.