Front page | perl.perl5.porters |
Postings from May 2012
On Tue, Sep 06, 2011 at 12:29:00PM +0100, Dave Mitchell wrote:
> There is a second issue that is brought up by the original bug report,
> and that is that $a =~ /foo/ closes over any outer lexical $_, even though
> it doesn't use $_. This is because when /foo/ is compiled, it initially
> assumes that it wants to use $_ and so closes over it (if there's a lexical
> version in scope); only later does it see the =~ bind operator, and
> undo that assumption (by setting OPf_STACKED). By then the closure
> has already been set up (a FAKE entry in the pad).
>
> I don't see a simple way to fix that.
Meaning that the third section of this code is buggy?
$ cat ../test/94682.pl
#!/usr/bin/perl -w
sub DESTROY {
print "DESTROY for $_[0]\n";
}
sub should_close {
my $_ = shift;
return sub {
print "I'm really in a closure with $_\n";
}
}
sub should_not_close {
my $_ = shift;
return sub {
print "I'm not a closure\n";
}
}
sub buggy {
my $_ = shift;
return sub {
my $m = "I'm not to be a closure\n";
$m =~ s/not/not meant/;
print $m;
}
}
print "Should close\n";
my $s = should_close(bless []);
print "(object should live)\n";
$s->();
undef $s;
print "\n";
print "Should not close\n";
$s = should_not_close(bless []);
print "(object should be gone)\n";
$s->();
undef $s;
print "\n";
print "Should not close either\n";
$s = buggy(bless []);
print "(object should be gone)\n";
$s->();
undef $s;
__END__
$ ./perl -Ilib ../test/94682.pl
Should close
(object should live)
I'm really in a closure with main=ARRAY(0x100802eb8)
DESTROY for main=ARRAY(0x100802eb8)
Should not close
DESTROY for main=ARRAY(0x1008114a8)
(object should be gone)
I'm not a closure
Should not close either
(object should be gone)
I'm not meant to be a closure
DESTROY for main=ARRAY(0x1008114c0)
In that the sub buggy() should not return a closure, hence the third
paragraph of output should be the same as the second, not the first, as
currently is?
(and this should become a TODO test somewhere?)
Nicholas Clark
Thread Previous
|
Thread Next