Adding a non-templatized version of QI for times when the IID and type aren't...

Adding a non-templatized version of QI for times when the IID and type aren't associated.

Also fixing the ScopedComPtrVector unit test.
The problem with the test was that it expected AddRef to be called twice and Release() once when pushing a pointer to the vector.  However that only happens when a push causes a reallocation.  When there's no reallocation, there's only a single call to AddRef and no call to Release, causing the test to report failure.  The fix is to simply use std::list instead and fix the expected values.

TEST=Run *ScopedComPtr* unit tests.
BUG=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=20183

[M-A Ruel, 2009-07-08 19:15:10]

lgtm

http://codereview.chromium.org/149345/diff/5/7
File base/scoped_comptr_win_unittest.cc (right):

http://codereview.chromium.org/149345/diff/5/7#newcode8
Line 8: #include "base/scoped_comptr_win.h"
nit: Can you put this include first since there is no .cc equivalent.

http://codereview.chromium.org/149345/diff/5/7#newcode82
Line 82: // Use std::list instead of std::vector to avoid hitting a reallocation
nit: Bah, that's quite basic C++; I don't think this need a comment :P

[tommi (sloooow) - chröme, 2009-07-08 19:36:33]

http://codereview.chromium.org/149345/diff/5/7
File base/scoped_comptr_win_unittest.cc (right):

http://codereview.chromium.org/149345/diff/5/7#newcode8
Line 8: #include "base/scoped_comptr_win.h"
On 2009/07/08 19:15:10, Marc-Antoine Ruel wrote:
> nit: Can you put this include first since there is no .cc equivalent.

done... (I take it you mean above the <> includes)

http://codereview.chromium.org/149345/diff/5/7#newcode82
Line 82: // Use std::list instead of std::vector to avoid hitting a reallocation
On 2009/07/08 19:15:10, Marc-Antoine Ruel wrote:
> nit: Bah, that's quite basic C++; I don't think this need a comment :P

hmm... well... I don't think it's that immediately obvious that hitting a
relocation in the container class will affect the test results.  comment doesn't
hurt. what do you think?

[tommi (sloooow) - chröme, 2009-07-08 19:41:35]

so, according to the buildbots, switching to list isn't
enough<http://build.chromium.org/buildbot/try-server/builders/win/builds/9122/steps/base_unittests/logs/ScopedComPtrVector>so
I'm just switching back to vector as it was before and calling
reserve()
to avoid the reallocation.

On Wed, Jul 8, 2009 at 3:36 PM, <tommi@chromium.org> wrote:

>
> http://codereview.chromium.org/149345/diff/5/7
> File base/scoped_comptr_win_unittest.cc (right):
>
> http://codereview.chromium.org/149345/diff/5/7#newcode8
> Line 8: #include "base/scoped_comptr_win.h"
> On 2009/07/08 19:15:10, Marc-Antoine Ruel wrote:
>
>> nit: Can you put this include first since there is no .cc equivalent.
>>
>
> done... (I take it you mean above the <> includes)
>
> http://codereview.chromium.org/149345/diff/5/7#newcode82
> Line 82: // Use std::list instead of std::vector to avoid hitting a
> reallocation
> On 2009/07/08 19:15:10, Marc-Antoine Ruel wrote:
>
>> nit: Bah, that's quite basic C++; I don't think this need a comment :P
>>
>
> hmm... well... I don't think it's that immediately obvious that hitting
> a relocation in the container class will affect the test results.
> comment doesn't hurt. what do you think?
>
>
> http://codereview.chromium.org/149345
>

[M-A Ruel, 2009-07-08 20:07:45]

lgtm