Re: Assignments into FUNCTAB resulting either strange output or hanging

[Andrew J. Schorr]

On Wed, Sep 08, 2021 at 11:55:18AM -0400, Andrew J. Schorr wrote:
> Interesting. The docs say:
> 
> "Any attempt to assign to an element of 'FUNCTAB' also causes a fatal error."
> 
> So the main question in my mind is why these programs don't throw a fatal
> error.
> 
> bash-4.2$ ./gawk 'BEGIN {print FUNCTAB[0] = 1}'
> gawk: cmd. line:1: fatal: cannot assign to elements of FUNCTAB
> bash-4.2$ ./gawk 'BEGIN {print FUNCTAB[0]}'
> 0

The debug dump of that last program is basically:

[     2:0x1171188] Op_push_array       : FUNCTAB
[     2:0x11711b0] Op_push_i           : 0 [MALLOC|NUMCUR|NUMBER|NUMINT]
[     2:0x1171200] Op_subscript        : [sub_count = 1]
[     2:0x1171160] Op_K_print          : [expr_count = 1] [redir_type = ""]

In interpret.h:Op_subscript, we have:

                case Op_subscript:
                        t2 = mk_sub(pc->sub_count);
                        t1 = POP_ARRAY(false);

...
                        /* for FUNCTAB, get the name as the element value */
                        if (t1 == func_table) {
                                static bool warned = false;

                                if (do_lint_extensions && ! warned) {
                                        warned = true;
                                        lintwarn(_("FUNCTAB is a gawk extension"
));
                                }
                                r = t2;
                        } else {
                                /* make sure stuff like NF, NR, are up to date 
*/
                                if (t1 == symbol_table)
                                        update_global_values();

                                r = *assoc_lookup(t1, t2);
                        }
                        DEREF(t2);
...

                        if (r->type == Node_val)
                                UPREF(r);
                        PUSH(r);
                        break;

So when FUNCTAB[<blah>] is referenced, it returns <blah> as the value,
regardless of whether <blah> actually exists in the FUNCTAB table.
Is that what's wanted? Or should it throw a fatal error if <blah>
is not in the FUNCTAB table?

Possible patch attached.

Regards,
Andy

functab_subscript.patch
Description: Text document