970083 - audio from ScriptProcessorNode stops after a random while

Status: RESOLVED WORKSFORME

(Core :: Web Audio, defect, P2)

Description

[p.t.deboer]

Attachment: webaudio-scriptprocessor-stops-working.html

What did you do?
Open the attached html file; while it's beeping (or even with volume off)
do random things on the computer, e.g., resizing windows, start other applications.
  
What happened?
The beeping sound produced by this webpage stops (or: doesn't return when
clicking "volume on", if it had previously been turned off).

What should have happened?
The beeping should have continued.

Notes:
- I verified that the onaudioprocess handler is still being called, so
  it's not a duplicate of bug 916387.
- It seems very similar to bug 932621, which apparently has been fixed, 
  but I still see this problem on nightly (on Linux).
- It's not always easy to reproduce, hence the volume on/off buttons
  to just leave it running in the background.
- I've seen the problem also on Windows and Mac (but haven't tested the
  latest nightly on those).

Comment 1

[Maire Reavy [:mreavy]]

For the bug reporter: Are you sure you tested with the fix from Bug 932621?  Your last comment ("It seems very similar to bug 932621, which apparently has been fixed, but I still see this problem on nightly") is from Feb 9, and the fix didn't go into the code until Feb 16.

In any case, can you retest with today's Nightly and tell us if you're still having a problem?  Thanks.

Comment 2

[p.t.deboer]

Oh, sorry, I had silently assumed that bug fixes such as the one for bug 932621 go into Nightly immediately.

Anyway, I've tested with current Nightly and (unfortunately) still see the problem occur, both on a Linux and a Windows system.

Some more information that perhaps might be of help in debugging this: by feeding the output of the scriptprocessor node to another scriptprocessor node, I find that when the problem occurs, the second scriptprocessor node sees the contents of its incoming inputBuffer as zero, even though the first scriptprocessor node had just written non-zeroes into its outputBuffer.
(I now use this in my application to detect the problem and restart the first scriptprocessor node as needed.)

If you'd like me to do any specific further tests, please let me know.

Comment 3

[Maire Reavy [:mreavy]]

Karl - is this still an issue?

Comment 4

[Karl Tomlinson (:karlt)]

I haven't been able to reproduce, but please feel free to reopen if it is still happening.