RTPUtils, inaccurate difference of sequence numbers

[flurbi]

org.jitsi.util.RTPUtils.sequenceNumberDiff(...)
    line 34
    incorrect: diff > 1<<15
    problem  : 1<<15 is not in the short range (-2^15 .. 2^15 - 1)
    correct  : diff >= 1<<15

[bgrozev]

Well, if the difference between two sequence numbers is exactly 2^15, then we don't know which one comes first. I think both 2^15 and -2^15 are reasonable values to return, and I don't know of a requirement for the return value to be in the short range.

[flurbi]

Casted back to an short leads to the same result anyway (2^15 -> -2^15). I do not know if its matter, but since the two inputs are in (-2^15 .. 2^15 - 1), I thought it would be more consistent if the computed value lays in the same range (unique representative modulo 2^16). Note that the docu says "modulo 2^16", without specifying the range (which is not ambiguous in C with uint16_t). That the inputs are also "modulo 2^16" is not explicitly stated neither, but otherwise the code does not make much sense.

I suggest this simpler implementation, which I had in mind as I opened this ticket:

    public static short sequenceNumberDiff_loic(short a, short b) {
        return (short) (a - b);
    }

Then it is clear what "modulo 2^16" means. It is (up to the handling of 2^15) equivalent to the original code, shorter and possibly faster. I even question if a function is needed for that.

Up to you judgment to adapt the code and/or the documentation, it is not critical in my opinion.

[bgrozev]

Good point, I'll address it when I get a chance.