[PATCH v2] winegstreamer: Poll gst_pad_query_duration before converting from byte length.

[Zebediah Figura (she/her)]

On 4/26/21 12:37 PM, Connor McAdams wrote:
> Wait to query the pad duration until after caps are received, and poll
> gst_pad_query_duration a few times before giving up and trying to
> get duration by converting from byte-length. Fixes a bug in Fallout:
> New Vegas where the duration value is incorrect and causes the audio to
> loop.
> 
> Signed-off-by: Connor McAdams <cmcadams at codeweavers.com>
> ---
>   dlls/winegstreamer/wg_parser.c | 14 +++++++++++++-
>   1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/dlls/winegstreamer/wg_parser.c b/dlls/winegstreamer/wg_parser.c
> index 5529321490e..e07fb7510ac 100644
> --- a/dlls/winegstreamer/wg_parser.c
> +++ b/dlls/winegstreamer/wg_parser.c
> @@ -1499,9 +1499,21 @@ static gboolean src_event_cb(GstPad *pad, GstObject *parent, GstEvent *event)
>   static LONGLONG query_duration(GstPad *pad)
>   {
>       gint64 duration, byte_length;
> +    unsigned int i;
>   
>       if (gst_pad_query_duration(pad, GST_FORMAT_TIME, &duration))
>           return duration / 100;
> +    /*
> +     * Sometimes we query duration too early, try again a few times before
> +     * giving up and trying to convert from byte length. If the duration can't
> +     * be found after 50ms, we probably have a bigger issue.
> +     */
> +    for (i = 0; i < 10; ++i)
> +    {
> +        g_usleep(5000);

We probably don't want to just sleep here, but instead try to wait for 
duration-changed. For elements that send it (e.g. mpegaudioparse) that 
should be more efficient than polling.

> +        if (gst_pad_query_duration(pad, GST_FORMAT_TIME, &duration))
> +            return duration / 100;
> +    }
>   
>       WARN("Failed to query time duration; trying to convert from byte length.\n");
>   
> @@ -1553,9 +1565,9 @@ static HRESULT CDECL wg_parser_connect(struct wg_parser *parser, uint64_t file_s
>       {
>           struct wg_parser_stream *stream = parser->streams[i];
>   
> -        stream->duration = query_duration(stream->their_src);
>           while (!stream->has_caps && !parser->error)
>               pthread_cond_wait(&parser->init_cond, &parser->mutex);
> +        stream->duration = query_duration(stream->their_src);
>           if (parser->error)
>           {
>               pthread_mutex_unlock(&parser->mutex);
> 

This is a good idea, but it's also a separate change and deserves its 
own patch.