I haven't tried out the repro test yet, but the error message is pretty clear:  you've called setValueAtTime at time 0.2119 but your setValueCurveAtTime runs from time 0.20317 to 0.21317.  Thus, the error message is correct: you can't set an automation in the middle of a setValueCurveAtTime.

Thanks for your comment!

Yes the error is clear but the problem is that the event ends up to be scheduled at a time that is already passed because the whole application slows down when the array is too long.

On my pretty fast Linux Z840, your test case uses up 85% CPU.  So you could possibly be running out of CPU for the render quantum, causing your timing to be messed up.

I'll need to take a closer look at this, though.

Oh, and for the record, is it really crashing for you?  That is, a sad tab? Or is it actually producing a console error message about setValueAtTime overlapping a setValueCurve?

The distinction matters a lot.

Ok, so it would be possible that the delay comes from a too high CPU. Is it the case with older version of Chrome as well?

As for the crash, in some cases it freezes and I have to quit Chrome and restart it. Sometimes I can refresh the page. In any case, I have hundreds of error message in the console.

Ok.  Couple of things here.  First, there is a cost for long curves
because we produce a warning if any curve value is outside the nominal
range and because we have to copy the curve to an internal storage
area.

Second, I can reproduce this on my mac.  We probably shouldn't be
generating this error.  On my mac (after adding some logs to
toneCloud.js), I see this:

setValueCurve(..., 0.005333333333333333, 0.01)
setvalueCurve(..., 0.04533333333333334, 0.01)

Failed to execute setValueCurveAtTime on AudioParam:
setValueAtTime(0,0.005333333333333333) overlaps
setValueCurveAtTime(..., 0.04533333333333334, 0.01)

You never call setValueAtTime, so this setValueAtTime is an internally
generated event that is placed at the end of a setValueCurveAtTime.

This needs to be fixed.

Oh, and it's not a crash.  It's a Javascript error.  That's a big difference.

Oops. The actual error message on OSX is:

Failed to execute 'setValueCurveAtTime' on 'AudioParam': setValueAtTime(0, 0.05533333333333334) overlaps setValueCurveAtTime(..., 0.05333333333333334, 0.01)

so the setValueCurveAtTiem is starting at the same time as the setValueAtTime.

I have a new warning in Chrome 64.0.3275.0 (Canary). Maybe this is related: 

`Gain.gain.value setter called at time 0 overlaps event setValueCurveAtTime(..., 0, 0.01) to setValueAtTime(0.1, 0.01)`

You can try to reproduce it from the file `target.js` in the same github repo (https://github.com/hadrienj/testSetValueCurveAtTime). There is no warning in Chrome 62.0.3202.94.

Thanks!

Sorry for the delay.  

The new warning is a note that you're setting the gain.value at the same time as when you've scheduled a setValueCurveAtTime.  In the very near future, this is going to cause an error because the value setter will be come exactly equivalent to setValueAtTime and you can't do a setValueAtTime that overlaps a setValueCurveAtTime.

Hi,

I still have this error with newer version of Chrome and now even with a smaller amount of values in `setValueCurveAtTime`. Any update about this?

Thanks!

Please send us the error message that you're getting the in dev console.

It is still the same error (like `Uncaught DOMException: Failed to execute 'setValueCurveAtTime' on 'AudioParam': setValueAtTime(0, 0.2119423349114291) overlaps setValueCurveAtTime(..., 0.2031746031746032, 0.01)`)

This message says that you've scheduled (probably) multiple setValueCurveAtTime() automations that overlap.  The error message is not very clear.  

I am working to clarify the message so it's more obvious exactly what the problem is.  It's not going to help you though. I think you need to change how you schedule the automations so that they occur a little further into the future.

Yes I see thank you for your time! It is indeed very frustrating because it was working well on older versions of Chrome (until Chrome 58.0.3029.81)!

Hmm.  It's possible we regressed how setValueCurveAtTime is done between Chrome 58 and now.  One change is that we now have to copy the curve array into internal storage; this wasn't done before. I'll need to investigate when this change is made.

Not sure what's going on.  We've been doing a copy since at least Chrome 57.

Using ToT chrome with my patch for better messages.  I still get messages like:

Uncaught DOMException: Failed to execute 'setValueCurveAtTime' on 'AudioParam': setValueCurveAtTime(..., 187.3888700599857, 0.01) overlaps setValueAtTime(0.0166667, 187.3966439909297)

So it seems you are actually scheduling setValueCurve that overlaps setValueAtTime (or vice versa).

I also run perf on this.  Very little time is spent on running the timeline including the time to process setValueCurveAtTime which include the time to do the overlap check and the array copy:

   0.12%  chrome           libblink_modules.so          [.] blink::AudioParam::setValueCurveAtTime
   0.04%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::ValuesForFrameRange
   0.03%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::ValuesForFrameRangeImpl
   0.01%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::ProcessSetValueCurve
   0.01%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::HandleAllEventsInThePast
   0.01%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamHandler::CalculateFinalValues
   0.01%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::HandleFirstEvent
   0.01%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::HasValues
   0.01%  chrome           libblink_modules.so          [.] blink::AudioParamHandler::AudioParamHandler
   0.00%  chrome           libblink_modules.so          [.] blink::AudioParamTimeline::SetValueCurveAtTime
   0.00%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParam::value
   0.00%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamHandler::CalculateSampleAccurateValues
   0.00%  chrome           libblink_modules.so          [.] blink::AudioParamTimeline::InsertEvent
   0.00%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamTimeline::ClampNewEventsToCurrentTime
   0.00%  AudioOutputDevi  libblink_modules.so          [.] blink::AudioParamHandler::Value

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/9390f8e06d90465c467c8c72138bf4c3cd327206

commit 9390f8e06d90465c467c8c72138bf4c3cd327206
Author: Raymond Toy <rtoy@chromium.org>
Date: Mon Feb 12 19:05:20 2018

Use special internal event for marking the end of a curve

Previously, creation of a SetValueCurve event would automatically
insert a SetValue event at the end of the SetValueCurve to establish
an end value for the curve so that following events would start from
the end of the curve instead of the beginning.

However, this causes incorrect errors about overlapping SetValueCurve
events if a SetValueCurve is scheduled to start at the very end of a
preceeding SetValueCurve.  The implicitly inserted SetValue would
overlap the new SetValueCurve.  Using an internal event avoids this
conflict, but still establishes the appropriate event needed to manage
the timeline.

Bug: 768740
Test: 
Change-Id: Ibc798e0427bb49c3e163f5a7535f26c49b166303
Reviewed-on: https://chromium-review.googlesource.com/731186
Reviewed-by: Hongchan Choi <hongchan@chromium.org>
Commit-Queue: Raymond Toy <rtoy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#536150}
[modify] https://crrev.com/9390f8e06d90465c467c8c72138bf4c3cd327206/third_party/WebKit/LayoutTests/webaudio/AudioParam/audioparam-setValueCurve-exceptions-expected.txt
[modify] https://crrev.com/9390f8e06d90465c467c8c72138bf4c3cd327206/third_party/WebKit/Source/modules/webaudio/AudioParamTimeline.cpp
[modify] https://crrev.com/9390f8e06d90465c467c8c72138bf4c3cd327206/third_party/WebKit/Source/modules/webaudio/AudioParamTimeline.h

hadrianjean@: Can you retry with Chrome canary?  There have been a few changes to make the messages clearer (and a few minor speed-ups in some cases).

Hi rtoy,

Thanks for your help!

Unfortunately, the problem remains with the last Canary version. Do you have the same problem when you copy and paste the content of the `toneCloud.js` file into the console ? It works on Chrome 58 and older but not yet on the last Canary.

Thanks again

I haven't tried it yet.  I still think it's because you don't schedule the events far enough in advance. but I will test again locally.