Re: Bogus reports from coverage.postgresql.org

I wrote:
> Huh.  7.3.1 shows the bug.

Wait, scratch that, I misread it.  The gcov output on 7.3.1 looks like

     4127:  478:                                case CLASS_ATOM:
        -:  479:
        -:  480:                                        /*
        -:  481:                                         * atom => atom is the base case
        -:  482:                                         */
        -:  483:                                        return
     4127:  484:                                                predicate_implied_by_simple_clause((Expr *\
) predicate,
call    0 returned 100%
        -:  485:                                                                                          \
                         clause,
        -:  486:                                                                                          \
                         weak);
        -:  487:                        }
    #####:  488:                        break;
        -:  489:        }
        -:  490:
        -:  491:        /* can't get here */
    #####:  492:        elog(ERROR, "predicate_classify returned a bogus value");
call    0 never executed
call    1 never executed
call    2 never executed
        -:  493:        return false;
        -:  494:}
        -:  495:
        -:  496:/*----------
        -:  497: * predicate_refuted_by_recurse

which is correct: it's showing that the elog line is never executed.

So apparently, this bug is specific to gcc 6.3.0 and maybe a few
versions on either side of that.

            regards, tom lane