Re: [bug-gawk] GAWK 4.1.60 DIV() Remainder Bug

[Aharon Robbins]

Hello again.

> From: Aharon Robbins <address@hidden>
> Date: Tue, 05 Aug 2014 16:25:23 +0300
> To: address@hidden, address@hidden
> Cc: address@hidden
> Subject: Re: [bug-gawk] GAWK 4.1.60 DIV() Remainder Bug
>
> Hi Andy & Katie.
>
> Some quick experimentation convinced me that MPFR was getting this wrong
> also. Andy, your analysis down to mpz_mod is great and has saved me some
> time.  Now we have to figure out WHY mpz_mod isn't doing what it is
> supposed to be doing.
>
> In the worst case, I'll just switch out to use the div_mod function that
> div() uses, but I'd like to try for a bit to understand why mpz_mod
> isn't working.

So, after web searching and reading, I'm stumped. mpz_mod() should be doing
the trick, but it's not.  So I have punted and moved to mpz_tdiv_qr() which
produces the right result, and I just throw away the quotient.

Here is the patch that I will be pushing shortly.

Katie - feel free to keep poking.  You may be the first user to seriously
exercise the MPFR code.  I think this is like the 3rd or 4th bug fix in
the space of 2 weeks!

Thanks!

Arnold
-------------------------------------------------------------
diff --git a/mpfr.c b/mpfr.c
index 37a1217..e53af61 100644
--- a/mpfr.c
+++ b/mpfr.c
@@ -1376,8 +1376,27 @@ mpg_mod(NODE *t1, NODE *t2)
        int tval;
 
        if (is_mpg_integer(t1) && is_mpg_integer(t2)) {
+               /*
+                * 8/2014: Originally, this was just
+                *
+                * r = mpg_integer();
+                * mpz_mod(r->mpg_i, t1->mpg_i, t2->mpg_i);
+                *
+                * But that gave very strange results with negative numerator:
+                *
+                *      $ ./gawk -M 'BEGIN { print -15 % 7 }'
+                *      6
+                *
+                * So instead we use mpz_tdiv_qr() to get the correct result
+                * and just throw away the quotient. We could not find any
+                * reason why mpz_mod() wasn't working correctly.
+                */
+               NODE *dummy_quotient;
+
                r = mpg_integer();
-               mpz_mod(r->mpg_i, t1->mpg_i, t2->mpg_i);
+               dummy_quotient = mpg_integer();
+               mpz_tdiv_qr(dummy_quotient->mpg_i, r->mpg_i, t1->mpg_i, 
t2->mpg_i);
+               unref(dummy_quotient);
        } else {
                mpfr_ptr p1, p2;
                p1 = MP_FLOAT(t1);