Re^16: supporting quads on 32 bit Perl

[salva]

ok, I have uploaded version 0.20. Let's see how it does at the ActiveState PPM builders this time.

In any case, if you find any other problem with the module, don't hesitate to report it.

[bulk88]

I got a strange croak from Int64. I'm not sure who to blame. I'm leaning on Test::More is to blame.

use strict;
use warnings;
use Math::Int64 qw( hex_to_int64 );
use Test::More;

my $normalnum = 0;
is($normalnum, hex_to_int64("0x8000200030004000"), "some test");

sub _unoverload {
    my $self = shift;
    my $type = shift;

    $self->_try(sub { require overload; }, die_on_fail => 1);

    foreach my $thing (@_) {
        if( $self->_is_object($$thing) ) {
            if( my $string_meth = overload::Method( $$thing, $type ) )
+ {
>>>>>>>>>>>>>>>                $$thing = $$thing->$string_meth();
            }
        }
    }

    return;
}

which was called from

sub _unoverload_str {
    my $self = shift;

    return $self->_unoverload( q[""], @_ );
}

which was called from Test::Builder::cmp_ok which was called from Test::Builder::is_eq which was called from Test::More::is. When I put a breakpoint right before Int64's _string XSUB is called in _unoverload, the command line says

#   Failed test 'some test'
not ok 1 - some test
#   at C:\Documents and Settings\Owner\Desktop\n13.pl line 7.

If I run the whole script the output is

C:\Documents and Settings\Owner\Desktop>perl n13.pl
not ok 1 - some test
#   Failed test 'some test'
#   at n13.pl line 7.
Usage: Math::Int64::_string(self, other, rev) at C:/perl512/lib/Test/B
+uilder.pm
line 876.
# Tests were run but no plan was declared and done_testing() was not s
+een.

C:\Documents and Settings\Owner\Desktop>

In XS, in Math::Int64::_string, from Test::Builder::_unoverload, items variable was 1. That will cause the croak obviously.

[salva]

Yes, this is an error in Test::More calling some overload method directly with an incorrect number of arguments.

I have changed Math::Int64 to support it anyway and also solved a problem with MS compiler not being able to handle uint64 to double conversions.

Version 0.22 is now available from CPAN.

[bulk88]

0.22 from CPAN will not compile. All the errors follow the same format.

Int64.xs(974) : error C2275: 'int64_t' : illegal use of this type as a
+n expressi
on
        Int64.xs(22) : see declaration of 'int64_t'

Given this XS code

SV *mi64_right(self, other, rev = &PL_sv_no)
    SV *self
    SV *other
    SV *rev
CODE:
    int64_t a;
    uint64_t b;
    if (SvTRUE(rev)) {
        a = SvI64(aTHX_ other);
        b = SvU64x(self);
    }
    else {
        a = SvI64x(self);
        b = SvU64(aTHX_ other);
    }

which produces this C code

XS(XS_Math__Int64__right); /* prototype to pass -Wmissing-prototypes *
+/
XS(XS_Math__Int64__right)
{
#ifdef dVAR
    dVAR; dXSARGS;
#else
    dXSARGS;
#endif
    if (items < 2 || items > 3)
       croak_xs_usage(cv,  "self, other, rev = &PL_sv_no");
    {
    SV *    self = ST(0);
    SV *    other = ST(1);
    SV *    rev;
    SV *    RETVAL;

    if (items < 3)
        rev = &PL_sv_no;
    else {
        rev = ST(2);
    }
#line 974 "Int64.xs"
    int64_t a;
    uint64_t b;
    if (SvTRUE(rev)) {
        a = SvI64(aTHX_ other);
        b = SvU64x(self);
    }
    else {
        a = SvI64x(self);
        b = SvU64(aTHX_ other);
    }

The error is obvious. You need to use PREINIT. Never rely on typemap entries being 1 liners and initialization not being deferred by XSPP. You can open new blocks, create new C autos, and run 1/4 page of code from a typemap entry. You can even run perl code to generate the type entry in the XS sub. For example to get accept a reference and its target.

T_REFANDSV
    ".(unshift(@line, (
    'PREINIT:',
    '    SV * '.$var.'Ref;',
    'INPUT:')),'').
    "${\$var}Ref = $arg;
    if (SvROK(${\$var}Ref))
        $var = (SV*)SvRV(${\$var}Ref);
    else
        Perl_croak(aTHX_ \"%s: %s is not a reference\",
            ${$ALIAS?\q[GvNAME(CvGV(cv))]:\qq[\"$pname\"]},
            \"$var\");