Re: [PATCH v2 0/1] MR8436: mfplat: Optimise 1d Lock of 2d buffer when pitch equals width.

[Nikolay Sivov (＠nsivov)]

Nikolay Sivov (@nsivov) commented about dlls/mfplat/buffer.c:
> EnterCriticalSection(&buffer->cs);
>
> -    if (!buffer->_2d.linear_buffer && buffer->_2d.locks)
> +    if (!buffer->_2d.linear_buffer && buffer->_2d.width == buffer->_2d.pitch)
> +    {
> +        BYTE *scanline;
> +        LONG pitch;
> +
> +        /* width and pitch are the same, so this avoids a potentially expensive copy */
> +        hr = memory_2d_buffer_lock(buffer, &scanline, &pitch, data, NULL);
> +    }
> +    else if (!buffer->_2d.linear_buffer && buffer->_2d.locks)
> +    {
>          hr = MF_E_INVALIDREQUEST;
> +    }

Is it correct to swap those and ignore "!linear_buffer && locks" condition that was a failing path before?

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/8436#note_108034