Re: Feature request and a Chrome 22 bug?

On Tue, Nov 13, 2012 at 11:43 AM, Ray Bellis <ray@bellis.me.uk> wrote:

> On 06/11/2012 16:42, Ray Bellis wrote:
>
>> This works on Chrome 25.0.1317.0 canary but doesn't work on Chrome 22 -
>> the node's "playBackState" appears stuck in "scheduled" even though I
>> called "node.noteOn(0)".
>>
> The problem I reported last week persists in Chrome 23, but I've found
> that it's not related directly to my "Unity Source Node". It's also present
> in Safari 6.0.x
>
> My set up code is below:
>
>             // WebAudio set up
>             var ctx = this.ctx = new webkitAudioContext();
>
>             // Main Audio chain
>             // NB: filters doubled to provide original 24dB/octave cutoff
>             var osc = this.osc = ctx.createOscillator();
>             var vcaenv = this.vcaenv = ctx.createGainNode();
>             var filter1 = this.filter1 = ctx.createBiquadFilter();
>             var filter2 = this.filter2 = ctx.createBiquadFilter();
>             var master = this.master = ctx.createGainNode();
>             var mute = this.mute = ctx.createGainNode();
>
>             osc.connect(filter1);
>             filter1.connect(filter2);
>             filter2.connect(vcaenv);
>             vcaenv.connect(master);
>             master.connect(mute);
>             mute.connect(ctx.destination);
>
>             // VCF chain
>             var unity = this.unity = ctx.createUnitySource();
>             var vcfenv = this.vcfenv = ctx.createGainNode();
>             var vcfgain = this.vcfgain = ctx.createGainNode();
>
>             unity.connect(vcfenv);
>             vcfenv.connect(vcfgain);
>             vcfgain.connect(filter1.**frequency);
>             vcfgain.connect(filter2.**frequency);
>
> The primary chain is all audio: oscillator -> VCA EG -> Filters * 2 ->
> Master Volume -> Audio On/Off
>
> The secondary chain is there to allow an envelope to be applied to the two
> filters.   The filters have a base frequency, and then the VCF EG is
> multiplied by a gain node set to the same frequency as the filters and then
> also connected to the filters.  The effect _should_ be that when the VCF EG
> is at "1" the  filter cutoff goes up by one octave.
>
> But - the unity node _never_ leaves "scheduled" state.   I tried replacing
> the AudioBufferNode it's built from with a standard oscillator and I get
> the same behaviour.
>
> *However* - if I add:
>
>             vcfgain.connect(osc.frequency)**;
>
> at the end, it all starts (but of course doesn't sound right).
>
> It looks like something is confusing the node graph, but I can't tell
> what.  I've looked on the WebKit bugzilla and can't find any fixes that
> look related, although it does appear fine in Chrome 25.   I'd really like
> to get this up and working on standard releases of Chrome and Safari so I
> can finish the project and release it.
>
> Any ideas?
>
> thanks,
>
> Ray
>

Ray, could you please create a reduced/simplified test case with the exact
code you're working with, then write up a WebKit bug.  I can have a look if
I have a working/semi-working test page to start with...

Chris