On 1/25/21 10:09 AM, Nikolay Sivov wrote:
On 1/18/21 10:30 PM, Derek Lesho wrote:
Signed-off-by: Derek Lesho dlesho@codeweavers.com
dlls/winegstreamer/audioconvert.c | 26 ++++++++++++++++++++++---- 1 file changed, 22 insertions(+), 4 deletions(-)
diff --git a/dlls/winegstreamer/audioconvert.c b/dlls/winegstreamer/audioconvert.c index e16fc6f1a78..a95fc5506e7 100644 --- a/dlls/winegstreamer/audioconvert.c +++ b/dlls/winegstreamer/audioconvert.c @@ -123,16 +123,34 @@ static HRESULT WINAPI audio_converter_GetStreamIDs(IMFTransform *iface, DWORD in
static HRESULT WINAPI audio_converter_GetInputStreamInfo(IMFTransform *iface, DWORD id, MFT_INPUT_STREAM_INFO *info) {
- FIXME("%p %u %p.\n", iface, id, info);
- TRACE("%p %u %p.\n", iface, id, info);
- return E_NOTIMPL;
if (id != 0)
return MF_E_INVALIDSTREAMNUMBER;info->dwFlags = MFT_INPUT_STREAM_WHOLE_SAMPLES | MFT_INPUT_STREAM_DOES_NOT_ADDREF;
info->cbMaxLookahead = 0;
info->cbAlignment = 0;
info->hnsMaxLatency = 0;
/* TODO: this can be calculated using MFCalculateImageSize */
info->cbSize = 0;
return S_OK; }
static HRESULT WINAPI audio_converter_GetOutputStreamInfo(IMFTransform *iface, DWORD id, MFT_OUTPUT_STREAM_INFO *info) {
- FIXME("%p %u %p.\n", iface, id, info);
- TRACE("%p %u %p.\n", iface, id, info);
- return E_NOTIMPL;
if (id != 0)
return MF_E_INVALIDSTREAMNUMBER;info->dwFlags = MFT_OUTPUT_STREAM_PROVIDES_SAMPLES | MFT_OUTPUT_STREAM_WHOLE_SAMPLES;
info->cbAlignment = 0;
/* TODO: this can be calculated using MFCalculateImageSize */
info->cbSize = 0;
return S_OK; }
static HRESULT WINAPI audio_converter_GetAttributes(IMFTransform *iface, IMFAttributes **attributes)
Was that verified with what converter returns on Windows? I could work on tests for that if you want, because returning something like that looks rather arbitrary. And comment for image size probably does not apply for audio.
No, that was not verified, I think the reason I kept this commit around is that the session needs some implementation of it. As for the specific flags, I think that they match the implementation of the transform, we do allocate the output samples, and the whole_samples flag seems to be meaningless in the case of uncompressed audio.
It seems like the tests for fixing this function would be simple enough, so maybe I should just include them in the updated/fixed patch.