[PATCH v2 4/4] mfplat: Implement MFCreateMemoryBuffer

[Nikolay Sivov]

On 09/07/2018 10:51 AM, Alistair Leslie-Hughes wrote:

> Hi Nikolay,
>
>
> On 07/09/18 14:00, Nikolay Sivov wrote:
>> On 09/07/2018 03:58 AM, Alistair Leslie-Hughes wrote:
>>
>>> +static HRESULT WINAPI mfbuffer_Lock(IMFMediaBuffer *iface, BYTE
>>> **buffer, DWORD *max, DWORD *current)
>>> +{
>>> +    mfbuffer *This = impl_from_IMFMediaBuffer(iface);
>>> +
>>> +    TRACE("%p, %p %p, %p\n", This, buffer, max, current);
>>> +
>>> +    if(!buffer)
>>> +        return E_INVALIDARG;
>>> +
>>> +    *buffer = This->buffer;
>>> +    if(max)
>>> +        *max = This->max_length;
>>> +    if(current)
>>> +        *current = This->current;
>>> +
>>> +    return S_OK;
>>> +}
>> This doesn't lock anything. How did you test this?
>>
> Only with the tests written.
>
> When attempting to lock the buffer twice doesn't result return an error and
> in fact returns the same buffer pointer.
>
> The msdn reference, also suggests that locking the buffer really isn't a
> lock in the traditional sense,
> and allows multiple threads to change the data without error.
>
> https://docs.microsoft.com/en-us/windows/desktop/api/mfobjects/nf-mfobjects-imfmediabuffer-lock
> "Locking the buffer does not prevent other threads from calling Lock, so
> you should not rely on this method to synchronize threads."

Ok, but does Lock -> Unlock -> Lock return same pointer second time? 
This could be a useful test to add later.

>
> Regards
> Alistair
>