Re: setupapi: Implement pSetupGetField, with tests.

[Paul Vriens]

Hi James,

I'm looking into a crash of the setupapi/parser test on Vista.

It crashes in this piece:

465         ret = HeapFree( GetProcessHeap(), 0, (LPVOID)field );
  466         ok( !ret, "Expected HeapFree to fail\n" );
  467         ok( GetLastError() == ERROR_INVALID_PARAMETER,
  468             "Expected ERROR_INVALID_PARAMETER, got %u\n", GetLastError() );

when trying the second field.

It's been added by a patch of yours:

http://source.winehq.org/git/wine.git/?a=commit;h=7588b669d8331ff64fc3b75ceb...

What's the reason for checking the HeapFree here? Or what's the reason for doing 
it at all?

setupapi: Implement pSetupGetField, with tests.

[James Hawkins]

On Thu, Mar 27, 2008 at 7:06 AM, Paul Vriens  wrote:
> Hi James,
>
> I'm looking into a crash of the setupapi/parser test on Vista.
>
> It crashes in this piece:
>
> 465         ret = HeapFree( GetProcessHeap(), 0, (LPVOID)field );
>   466         ok( !ret, "Expected HeapFree to fail\n" );
>   467         ok( GetLastError() == ERROR_INVALID_PARAMETER,
>   468             "Expected ERROR_INVALID_PARAMETER, got %u\n", GetLastError() );
>
> when trying the second field.
>
> It's been added by a patch of yours:
>
> http://source.winehq.org/git/wine.git/?a=commit;h=7588b669d8331ff64fc3b75ceb...
>
> What's the reason for checking the HeapFree here? Or what's the reason for doing
>  it at all?

To show that the value returned is not in the heap.

setupapi: Implement pSetupGetField, with tests.

[Dmitry Timoshkov]

"James Hawkins"  wrote:
> > What's the reason for checking the HeapFree here? Or what's the reason for doing
> >  it at all?
>
> To show that the value returned is not in the heap.

Isn't any of HeapValidate or HeapSize better in that case?

setupapi: Implement pSetupGetField, with tests.

[James Hawkins]

On Thu, Mar 27, 2008 at 10:52 AM, Dmitry Timoshkov
 wrote:
> "James Hawkins"  wrote:
> > > What's the reason for checking the HeapFree here? Or what's the reason for doing
> > >  it at all?
> >
> > To show that the value returned is not in the heap.
>
> Isn't any of HeapValidate or HeapSize better in that case?

Sure.  It worked on XP and earlier so I didn't really think twice
about it when I wrote the tests.