[opus] NaN sanity check not present in tree.

Jean-Marc Valin jmvalin at jmvalin.ca
Mon Sep 23 14:41:16 UTC 2019

Hi Yves,

Sorry for the delay (I blame the Google spam filter). Have you actually
seen problems that this patch fixes? If so, do you have a way to
reproduce the problem?

While technically your patch would be safe, the simple fact that it
would trigger would mean a bug elsewhere. Note that HB_gain is
essentially 1 - exp(-celt_rate/1024.) where celt_rate is supposed to
always be non-negative because compute_silk_rate_for_hybrid() should
never give SILK more bits than there are in total. That means HB_gain
should be bounded between 0 and 1 and should definitely never be NaN.

So either the patch does nothing (in which case we don't need it), or it
does something and there's something much scarier going on than NaNs. In
the latter case, I think more investigation is required.



On 9/12/19 1:00 PM, Yves Gerey wrote:
> Hi, I'm still Yves.
> This other fix
> <https://chromium-review.googlesource.com/c/chromium/src/+/1061499/> never
> made it to opus repository.
> It seems safe and sane enough to be applied to opus master branch.
> What do you think?
> -- 
> Warm regards,
> Yves
> _______________________________________________
> opus mailing list
> opus at xiph.org
> http://lists.xiph.org/mailman/listinfo/opus

More information about the opus mailing list