This is likely related to this change in 5.6 (from perldelta):

=head2 delete(), each(), values() and hash iteration are faster

The hash values returned by delete(), each(), values() and hashes in a
list context are the actual values in the hash, instead of copies.
This results in significantly better performance, because it eliminate
+s
needless copying in most situations.
[download]

Previously, only copies were used, which meant that foreach-aliasing didn't really alias into the real hash values:

my %hash = (name => 'andrew', beer => 'Ale');
foreach my $KV (%hash) {
    print ">>$KV<<\n";
    $hash{$KV} = 'Dark Ale' if $KV eq 'beer';
}
__END__
# with 5.00503:
>>name<<
>>andrew<<
>>beer<<
>>Ale<<

# with 5.6.1
>>beer<<
>>Dark Ale<<
>>name<<
>>andrew<<
[download]

As you can see, with 5.005, even after we've changed the value for the key 'beer', we still get the old value because it was a copy. In 5.6 we get the new value because it is an alias into the hash. Now, in your situation, after deleting the key '*test' the next iteration involves an alias to a non-existing value and this is causing a problem for perl (which with -w gives "attempt to free unreferenced scalar", and a segfault otherwise).

% perl5.6.1 -w test.pl
Use of uninitialized value in substr at test.pl line 14.
Attempt to free unreferenced scalar at test.pl line 13.
Use of uninitialized value in substr at test.pl line 14.
Attempt to free unreferenced scalar at test.pl line 13.

# other invocations....
% perl5.00503 -w test.pl
[nothing]

% perl5.00503 test.pl
[nothing]

% perl5.6.1 test.pl
Segmentation fault (core dumped)
[download]

Update 2: It looks like deleting the key before you get to the corresponding value in the foreach loop is causing problems...

#!/usr/bin/perl
use strict;

### show what we're looping over (a,1,b,2,c,3)
my %hash = (a=>1, b=>2, c=>3);
for (%hash) {
  print "OK: $_\n";
}

### check the value *before* we attempt to delete the key
### i.e. 1 before a, 2 before b, 3 before c
%hash = (a=>1, b=>2, c=>3);
for (reverse %hash) {
  substr $_, 0, 1;
  delete($hash{$_}) if /[a-z]/;
  print "Still-OK: $_\n";
}

### emulate your structure where we can potentially delete
### a key first, then loop over the value associated with it
%hash = (a=>1, b=>2, c=>3);
for (%hash) {
  substr $_, 0, 1; # <== segfaults here when testing value associated 
+with deleted key
  delete($hash{$_}) if /[a-z]/;
  print "Not-OK: $_\n";
}

__END__

=head1 OUTPUT
OK: a
OK: 1
OK: b
OK: 2
OK: c
OK: 3

Still-OK: 3
Still-OK: c
Still-OK: 2
Still-OK: b
Still-OK: 1
Still-OK: a

Not-OK: a
Not-OK:
Segmentation fault (core dumped)
[download]

-Blake

use strict;

my %allusers = (
        'users' => {
                'user' => 'Test Account',
                '*Crudles' => 'Hello World',
                'Crud' => 'Another Test',
                '*test' => 'Crud User'
        }
);

foreach my $key (keys %allusers ) {
        delete $allusers {'users'}{$key}
                if '*' eq substr $key, 0, 1;
}
[download]

#!/usr/bin/perl -w
use strict;

my %allusers = (
        'users' => {
                'user' => 'Test Account',
                '*Crudles' => 'Hello World',
                'Crud' => 'Another Test',
                '*test' => 'Crud User'
        }
);

foreach my $key ( %{ $allusers{ users } } )
{
    delete $allusers{ users }{ $key } if '*' eq substr $key, 0, 1;
}


__END__
F:\dev>perl -c ovidcoredump.pl
ovidcoredump.pl syntax OK

F:\dev>perl ovidcoredump.pl
Use of uninitialized value in substr at ovidcoredump.pl line 14.
Attempt to free unreferenced scalar at ovidcoredump.pl line 13.
Use of uninitialized value in substr at ovidcoredump.pl line 14.
Attempt to free unreferenced scalar at ovidcoredump.pl line 13.

F:\dev>perl -v

This is perl, v5.6.1 built for MSWin32-x86-multi-thread
(with 1 registered patch, see perl -V for more detail)

Copyright 1987-2001, Larry Wall

Binary build 628 provided by ActiveState Tool Corp. http://www.ActiveS
+tate.com
Built 15:41:05 Jul  4 2001


Perl may be copied only under the terms of either the Artistic License
+ or the
GNU General Public License, which may be found in the Perl 5 source ki
+t.

Complete documentation for Perl, including FAQ lists, should be found 
+on
this system using `man perl' or `perldoc perl'.  If you have access to
+ the
Internet, point your browser at http://www.perl.com/, the Perl Home Pa
+ge.


F:\dev>f:\indigoperl\bin\perl ovidcoredump.pl
Use of uninitialized value in substr at ovidcoredump.pl line 15.
Attempt to free unreferenced scalar at ovidcoredump.pl line 13.
Use of uninitialized value in substr at ovidcoredump.pl line 15.
Attempt to free unreferenced scalar at ovidcoredump.pl line 13.


F:\dev>f:\indigoperl\bin\perl -v

This is perl, v5.6.0 built for MSWin32-x86-multi-thread

Copyright 1987-2000, Larry Wall

Perl may be copied only under the terms of either the Artistic License
+ or the
GNU General Public License, which may be found in the Perl 5.0 source 
+kit.

Complete documentation for Perl, including FAQ lists, should be found 
+on
this system using `man perl' or `perldoc perl'.  If you have access to
+ the
Internet, point your browser at http://www.perl.com/, the Perl Home Pa
+ge.


F:\dev>

bash-2.05$ perl ovid.pl
bash-2.05$ perl -c ovid.pl
ovid.pl syntax OK
bash-2.05$ perl -v

This is perl, version 5.005_03 built for i386-freebsd

Copyright 1987-1999, Larry Wall

Perl may be copied only under the terms of either the Artistic License
+ or the
GNU General Public License, which may be found in the Perl 5.0 source 
+kit.

Complete documentation for Perl, including FAQ lists, should be found 
+on
this system using `man perl' or `perldoc perl'.  If you have access to
+ the
Internet, point your browser at http://www.perl.com/, the Perl Home Pa
+ge.

bash-2.05$
[download]

 
___crazyinsomniac_______________________________________
Disclaimer: Don't blame. It came from inside the void
perl -e "$q=$_;map({chr unpack qq;H*;,$_}split(q;;,q*H*));print;$q/$q;"

I was wondering if deleting keys while iterating over a hash was reallocating the buckets and perhaps Perl was then trying to access memory that's not there. According to Adding or removing keys while iterating over a hash, this is not the case. I can't add, but I should have no problem deleting. Both merlyn and Dominus state that. However, what's vexing about this is that it appears similar to a Perl4 issue where deleting from a hash while iterating over it was verboten. This was fixed in Perl5.

Forgetting the keys should certainly cause the code to fail, but generating a core dump? That's certainly going to cause some people to panic :)

Cheers,
Ovid

Join the Perlmonks Setiathome Group or just click on the the link and check out our stats.

That's not the same problem. Your new problem is that you aren't using an iterator (keys(), values(), each()), but just directly evaluating the hash in list context --- and in 5.6.1 that results in the values part of the returned list being the actual values in the hash (instead of copies, as per my previous post). This shouldn't be a real problem because the normal way to do what you are trying to do would be to use keys() in the foreach loop to just iterate over the keys (and thus never be in the position of having an alias to a hash-value that no longer exists). There is no 'forgetting' of keys going on in your problem, your code is deleting a *value* that the foreach-list already has an alias to (because you used foreach (%hash) instead of foreach (keys %hash)).

Granted, this shouldn't segfault --- it doesn't when warnings are turned on (at least on 5.6.1 on linux). However, you can cause the same sort of behaviour on an array if you foreach over a list of array elements and remove (splice()) an element from the array before we get there in the list (segfaults on 5.00503 and 5.6.1, linux):

my @array  = (0,1,2);
foreach my $el ($array[0],$array[1],$array[2]){
    print ">>$el<<\n";
    splice(@array,1,1) if $el == 0;
}
[download]

node #623 -- What happens if I add or remove keys from a hash while iterating over it?

-Blake