On Thursday 14 February 2008 18:16:03 Jerry Gay wrote:
> 'cotto' on #parrot came to me with a problem today. he is unable to
> call VTABLE_is_equal(...) inside a .pmc file. i'll use
> src/pmc/exporter.pmc as an example pmc file, because i know its
> internals. here's a patch:
>
> Index: src/pmc/exporter.pmc
> ===================================================================
> --- src/pmc/exporter.pmc (revision 25720)
> +++ src/pmc/exporter.pmc (working copy)
> @@ -100,6 +100,17 @@
>
> */
>
> +static void test(PARROT_INTERP) {
> + PMC *p1 = pmc_new(interp, enum_class_String);
> + PMC *p2 = pmc_new(interp, enum_class_String);
> + INTVAL cmp;
> + VTABLE_set_string_native(interp, p1, const_string(interp, "test"));
> + VTABLE_set_string_native(interp, p2, const_string(interp, "test"));
> +
> + cmp = VTABLE_is_equal(interp, p1, p2);
> + printf("<<<%d>>>\n", cmp);
> +}
> +
> pmclass Exporter
> need_ext {
>
> @@ -117,6 +128,7 @@
> void init() {
> Parrot_Exporter *exp = NULL;
>
> + test(INTERP);
> /* Set up the object. */
> exp = mem_allocate_zeroed_typed(Parrot_Exporter);
> exp->ns_src = PMCNULL;
>
>
> compiling with this patch produces an error like:
> exporter.obj : error LNK2019: unresolved external symbol
> _VTABLE_is_equal referenced in function _test
>
> err, wha? it's in docs/pdds/pdd17_pmc.pod. it should be there. why on
> earth isn't it callable?
>
> i went digging in include/parrot/vtable.h, tools/build/vtable_h.pl,
> src/vtable.tbl, lib/Parrot/Vtable.pm. i can't figure out what's going
> on. i can see that there's no VTABLE_ macro for is_equal (or cmp, for
> that matter). i just can't put my finger on *why*. i wondered if it
> was just the missing VTABLE_ accessor macros, so i unrolled the macro.
> + cmp = (p1)->vtable->is_equal(interp, p1, p2);
>
> didn't work:
> .\src\pmc\exporter.pmc(110) : error C2039: 'is_equal' : is not a
> member of '_vtable'
>
>
> i was able to work around the problem, looking at how 'eq' is
> implemented in src/ops/cmp.ops gave me an idea. if i replace the line
> + cmp = VTABLE_is_equal(interp, p1, p2);
> with
> + cmp = mmd_dispatch_i_pp(interp, p1, p2, MMD_EQ);
> then it works. so, mmd dispatch is working. the functions are there.
>
>
> what am i missing? i'm guessing it's something in the perl that
> generates the c, but i have to step away for a while. i'd be happy if
> somebody else put their eyes on it, too. chromatic? anyone?
They're marked as MMD in vtable.tbl, so my guess is that they're not directly
callable by vtable pointer from C. F<docs/mmd.pod> (though admittedly out of
date) suggests that mmd_dispatch_* is the right approach.
-- c
