Regression with SMT2 `distinct` about boolector

[mpreiner]

Thanks for reporting! This is now fixed on master with a slightly different fix than @andrewvaughanj suggested.

[aytey]

Hi @jamesbornholt,

I took a quick look at this. It seems that this error is your classic "off-by-one" issue in the encoding of distinct: https://github.com/Boolector/boolector/blob/master/src/parser/btorsmt2.c#L2125

        for (i = 1; i < nargs; i++)
        {
          for (j = i + 2; j <= nargs; j++)

This inner loop, unfortunately, doesn't encode the distinct correctly (it doesn't tie arg[1] with arg[2], hence your SAT result -- and when encoding arg[2], the rest of the distinct is not considered).

I made a fix in this branch https://github.com/andrewvaughanj/boolector/tree/avj_boolector_distinct_issue_6 and have also opened-up a PR against Boolector master [#7].