[PATCH 1/2] dmime/tests: Added tests for IDirectMusicPerformance.

Vitaliy Margolen wine-devel at kievinfo.com
Thu Aug 12 00:48:11 CDT 2010


On 08/11/2010 11:11 PM, Austin Lund wrote:
> On 12 August 2010 14:23, Vitaliy Margolen<wine-devel at kievinfo.com>  wrote:
>> On 08/11/2010 09:14 PM, Austin Lund wrote:
>>>
>>> On 11 August 2010 20:12, David Adam<david.adam.cnrs at gmail.com>    wrote:
>>> I wasn't sure of the best way to do this.  But basically I want to
>>> check that the Vtbl had been set.  I cannot cast to
>>> IDirectMusicPortImpl and I'm still confused by the COM macros.
>>
>> You shouldn't be checking for any vtbl, it's all object's internal stuff. If
>> you must check a pointer if it's null or not, then do that explicitly.
>> Pointers have different size in 32-bit and 64-bit architectures.
>>
>> Also you don't need two things pointing to one another. IDirectMusicPort is
>> a pointer in a first place. So no need for memset either. And no need for
>> all the extra traces that print the same stuff. You already checking if it's
>> null or not. The value of the pointer won't be of any use.
>
> The setting (or not) of the Vtbl is exactly why the crash in bug 22598
> occurs.  So if I'm not allowed to test for this, then this bug
> shouldn't have a test written for it?
The IDirectMusicPort is a pointer. To zero it you do "ptr = NULL". Don't 
mention any vtbs in test. Think of it an (void*) pointer - what it points to 
shouldn't matter.

> The test I wrote for this bug just ensures that the object has
> something put into it to avoid the segfault.
You can't test that. All you can test is that you got the IDirectMusicPort 
back or not. What exactly inside is irrelevant.

And if IDirectMusicPerformance8_PChannelInfo succeeds and you getting not 
null pDirectMusicPort back but it points to bogus data or is unchanged (as 
it is now), it's outright invalid behavior that should be fixed. No tests 
needed for that.

Vitaliy.



More information about the wine-devel mailing list