[asterisk-users] More issues with Siren14 datalen == 0 packets

[Richard Kenner]

> All patches need to go into JIRA with a license agreement to be
> accepted.

Understood, but I was using it as an illustration.  Note, however, that,
from a legal perspective, a patch such as this has no protectable IP (you
can't copyright the only way of doing something) and the GNU projects have
a formal rule that sufficiently-small patches need no assignments for that
reason, which I suggest you may want to adopt as well.

> > Why is samples being used as a length instead of datalen?
> 
> Internally a signed linear factory operates in terms of samples, not the
> data payload itself. I've also commented on your original issue in
> regards to the siren codecs that it should NULL out the data pointer
> itself. That is more commonly used.

But I don't think that it would have helped in either case, this one
or in func_speex.c, because neither tests for a null data pointer either.

Can you explain the difference between "datalen" and "samples" in this
context, shouldn't they always be related by a (small) linear factor?

Should I open a JIRA issue for this as well?

Can you suggest ways of searching for other possible occurrences of
this bug?  These crashes are occurring during important conferences and
are causing significant issues.