[Spice-devel] [PATCH spice-html5] Review the webm audio track header and remove the fixmes.

[Jeremy White]

On 8/11/19 4:32 AM, Frediano Ziglio wrote:
> Hi,
>    why this was not merged ?

I completely missed the ack at the time.

I do think the patch is fundamentally correct and an improvement.  This 
one arguably deserves a reprise of the audit I performed at the time to 
ensure that the results are still correct.  I was planning to do that in 
September (and resubmit the uid patch at the same time).

For the record, I have an internal todo list of patches that were missed 
to circle back to.  You've now caught most of them, thank you 
(websockets being the big one).  I believe I still have one or two 
patches to argument parsing on libcacard, and I do not have a patch, but 
a concern, on the removal of the spice-controller.  I also have a large 
separate patch set against the kernel providing usbredir support.

Cheers,

Jeremy

>
> Frediano
>
>> Ack
>>
>> Thanks,
>> Pavel
>>
>> On Fri, 2016-12-23 at 10:53 -0600, Jeremy White wrote:
>>> This involved a review of the Firefox parsing code along
>>> with the official specifcation, and setting these fields
>>> to the specified default values.
>>>
>>> Most notably, I have found that recent browsers do not need the
>>> 8 ms pre skip, and it seems to remove some audio lag to
>>> switch to 0.
>>>
>>> Signed-off-by: Jeremy White <jwhite at codeweavers.com>
>>> ---
>>>   webm.js | 27 +++++++++++++++++++--------
>>>   1 file changed, 19 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/webm.js b/webm.js
>>> index 789da14..72c1853 100644
>>> --- a/webm.js
>>> +++ b/webm.js
>>> @@ -393,28 +393,39 @@ webm_SeekHead.prototype =
>>>   
>>>   function webm_AudioTrackEntry()
>>>   {
>>> +    /*
>>> +    ** In general, we follow this specification:
>>> +    **   https://www.matroska.org/technical/specs/index.html
>>> +    ** we supply the mandatory values, and a comment notes
>>> +    ** where we differ from the default
>>> +    **   There is one RECOMMENDED guideline we are omitting;
>>> +    ** the OPUS codec_delay is recommended to be 80ms.
>>> +    ** However, the spice server does not currently provide time
>>> +    ** stamps that are offset by 80ms, so you effectively get an
>>> +    ** 80ms lag with this setting.
>>> +    */
>>>       this.id = WEBM_TRACK_ENTRY;
>>>       this.number = 1;
>>>       this.uid = 1;
>>>       this.type = 2; // Audio
>>>       this.flag_enabled = 1;
>>>       this.flag_default = 1;
>>> -    this.flag_forced = 1;
>>> -    this.flag_lacing = 0;
>>> -    this.min_cache = 0; // fixme - check
>>> +    this.flag_forced = 1;  // Different than default; we wish to
>>> force
>>> +    this.flag_lacing = 1;
>>> +    this.min_cache = 0;
>>>       this.max_block_addition_id = 0;
>>> -    this.codec_decode_all = 0; // fixme - check
>>> -    this.seek_pre_roll = 0; // 80000000; // fixme - check
>>> -    this.codec_delay =   80000000; // Must match
>>> codec_private.preskip
>>> +    this.seek_pre_roll = 0;
>>> +    this.codec_delay =   0; // Must match codec_private.preskip
>>>       this.codec_id = "A_OPUS";
>>> +    this.codec_decode_all = 1;
>>>       this.audio = new webm_Audio(OPUS_FREQUENCY);
>>>   
>>>       // See:  http://tools.ietf.org/html/draft-terriberry-oggopus-01
>>>       this.codec_private = [ 0x4f, 0x70, 0x75, 0x73, 0x48, 0x65,
>>> 0x61, 0x64,  // OpusHead
>>>                              0x01, // Version
>>>                              OPUS_CHANNELS,
>>> -                           0x00, 0x0F, // Preskip - 3840 samples -
>>> should be 8ms at 48kHz
>>> -                           0x80, 0xbb, 0x00, 0x00,  // 48000
>>> +                           0x00, 0x00, // Preskip - 3840 samples
>>> would be 8ms at 48kHz
>>> +                           0x80, 0xbb, 0x00, 0x00,  // nominal rate
>>> - 48000
>>>                              0x00, 0x00, // Output gain
>>>                              0x00  // Channel mapping family
>>>                              ];