Re: Bug with vec() and bitwise OR?

[p5pRT]

Migrated from rt.perl.org#3223 (status was 'resolved')

Searchable as RT3223$

[p5pRT]

From rmb1@cise.npl.co.uk

Hey all,

The following snippet just appeared on c.l.p.misc, and I think it unveils a
bug (at least on 5.005_03)​:

\#\!/usr/bin/perl \-w

use strict;
my $x = my $y = my $z = 0;

vec\($x\, 0\, 1\) = 1;
vec\($y\, 1\, 1\) = 1;

$z = $x | $y;

print "$x $y $z\\n";
\_\_END\_\_

The above prints​:

1 2 0

The values of $x and $y are fine, but $z should be 3, no?
If you explicitly set $x to 1 and $y to 2 (without using vec()), then you
get the expected result for $z.

Anybody with a good explanation? I don't have 5.6 handy. Anybody can try it
there?

% perl -we 'for (qw(0 "0" "@​" "A" ""), q(" ")) { my $x = eval $_; die if $@​; vec($x,0,1)=1; print qq($_ => "$x", ), ($x == 1 ? "yes" : "no"), "\n"; }'
0 => "1", no
"0" => "1", yes
Argument "A" isn't numeric in eq at -e line 1.
"@​" => "A", no
Argument "A" isn't numeric in eq at -e line 1.
"A" => "A", no
Argument "^A" isn't numeric in eq at -e line 1.
"" => "", no
Argument "!" isn't numeric in eq at -e line 1.
" " => "!", no

It appears that after C<$x = 0; vec($x,0,1)=1;>,
  $x is numeric (no warning)
  is not numerically equal to 1,
  but is the string "1".

Robin

[p5pRT]

From [Unknown Contact. See original ticket]

I have a patch for this​:

Hey all,

The following snippet just appeared on c.l.p.misc, and I think it unveils a
bug (at least on 5.005_03)​:

\#\!/usr/bin/perl \-w

use strict;
my $x = my $y = my $z = 0;

vec\($x\, 0\, 1\) = 1;
vec\($y\, 1\, 1\) = 1;

$z = $x | $y;

print "$x $y $z\\n";
\_\_END\_\_

The above prints​:

1 2 0

The values of $x and $y are fine, but $z should be 3, no?
If you explicitly set $x to 1 and $y to 2 (without using vec()), then you
get the expected result for $z.

Anybody with a good explanation? I don't have 5.6 handy. Anybody can try it
there?

% perl -we 'for (qw(0 "0" "@​" "A" ""), q(" ")) { my $x = eval $_; die if $@​; vec($x,0,1)=1; print qq($_ => "$x", ), ($x == 1 ? "yes" : "no"), "\n"; }'
0 => "1", no
"0" => "1", yes
Argument "A" isn't numeric in eq at -e line 1.
"@​" => "A", no
Argument "A" isn't numeric in eq at -e line 1.
"A" => "A", no
Argument "^A" isn't numeric in eq at -e line 1.
"" => "", no
Argument "!" isn't numeric in eq at -e line 1.
" " => "!", no

It appears that after C<$x = 0; vec($x,0,1)=1;>,
$x is numeric (no warning)
is not numerically equal to 1,
but is the string "1".

Robin

Perl-Friends​: To unsubscribe, mail <perl-friends-unsubscribe@​perlsupport.com>

With my patch the original script prints "1 2 3" and my script print "yes".

Robin

--- doop.c 2000/05/11 12​:55​:39 1.1
+++ doop.c 2000/05/11 13​:30​:41 1.2
@​@​ -897,6 +897,7 @​@​
  }
#endif
  }
+ SvNIOK_off(targ); /* see t/op/vec.t test 16 */
  SvSETMAGIC(targ);
}

Patch​:
--- t/op/vec.t 2000/05/11 13​:18​:33 1.1
+++ t/op/vec.t 2000/05/11 13​:30​:41 1.2
@​@​ -4,5 +4,5 @​@​

-print "1..15\n";
+print "1..16\n";

print vec($foo,0,1) == 0 ? "ok 1\n" : "not ok 1\n";
print length($foo) == 0 ? "ok 2\n" : "not ok 2\n";
@​@​ -24,4 +24,6 @​@​
vec($Vec, 0, 32) = 0xbaddacab;
print $Vec eq "\xba\xdd\xac\xab" ? "ok 14\n" : "not ok 14\n";
print vec($Vec, 0, 32) == 3135089835 ? "ok 15\n" : "not ok 15\n";
+vec($foo = 0,0,1) = 1; # ensure $foo forgets its original numerical value
+print $foo == 1 ? "ok 16\n" : "not ok 16\n";

End of patch

[p5pRT]

From @tamias

On Thu, May 11, 2000 at 03​:35​:41PM +0100, Robin Barker wrote​:

--- doop.c 2000/05/11 12​:55​:39 1.1
+++ doop.c 2000/05/11 13​:30​:41 1.2
@​@​ -897,6 +897,7 @​@​
}
#endif
}
+ SvNIOK_off(targ); /* see t/op/vec.t test 16 */

That comment won't make sense if the tests are changed around. It would be
better to put a real comment here.

 SvSETMAGIC\(targ\);

}

Ronald

[p5pRT]

From rmb1@cise.npl.co.uk

Fixed by patch supplied by Robin Barker,
the additional test (t/op/vec.t #16) is still in the test suite.

[p5pRT]

rmb1@cise.npl.co.uk - Status changed from 'open' to 'resolved'