As ConvolverNode normalization has been incorrect much of the time due to bug 1024182 and bug 1389638, this might be a good opportunity to align with the spec. It seems the spec has been consistent since at least https://www.w3.org/TR/2012/WD-webaudio-20120802/#attributes-ConvolverNode but the Blink implementation differed by 0.7% Given these are arbitrary and useless numbers (trailing zeros change the scaling - https://github.com/WebAudio/web-audio-api/issues/481), I prefer the simpler form that matches the spec, but I don't really mind. What do you think?

Probably the main reason for doing this would be to prevent Gecko being used as evidence to support any argument that the spec should be made unnecessarily more complicated.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=e25ffac52583a8a2acbd2591f91e436b6432a5c0 I'm also happy to adjust the test to expect specified unit output, if you think that's helpful.

Comment on attachment 8896497 [details] bug 1389641 adjust ConvolverNode buffer normalization gain calibration to match spec https://reviewboard.mozilla.org/r/167754/#review173414

Pushed by ktomlinson@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/e920d33002f8 adjust ConvolverNode buffer normalization gain calibration to match spec r=padenot

Pushed by ktomlinson@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/9b101cdb3935 add 'f' to literal initializer for float const r=trivial

https://hg.mozilla.org/mozilla-central/rev/e920d33002f8 https://hg.mozilla.org/mozilla-central/rev/9b101cdb3935

Test case on bug 1024182.