Is it guaranteed anywhere in docs that hashes with same keys will also have same order?

Question

There's not much guarantees about order of hash keys in Perl. Is there any mention in docs that I can't find that would say that as long as two hashes use exactly same keys, they will go in exactly same order?

Short test seems to confirm that. Even if I generate some additional keys for internal key table between assigning to two different hashes, their keys are returned in same order:

my %aaa;
my %bbb;
my %ccc;
my %ddd;

@aaa{qw(a b c d e f g h i j k l)}=();
# Let's see if generating more keys for internal table matters
@ccc{qw(m n o p q r s t u v w x)}=();
@bbb{qw(a b c d e f g h i j k l)}=();
# Just to test if different insertion order matters
@ddd{qw(l k c d e f g h i j a)}=(); $ddd{b} = ();

print keys %aaa, "\n";
print keys %bbb, "\n";
print keys %ddd, "\n";

However I wouldn't rely on udocumented behavior and only fact that can be easily found in docs is that keys, values and each all will use same order as long as hash is not modified.

Answer 1

From perlsec:

Perl has never guaranteed any ordering of the hash keys, and the ordering has already changed several times during the lifetime of Perl 5. Also, the ordering of hash keys has always been, and continues to be, affected by the insertion order.

http://perldoc.perl.org/perlsec.html

Answer 2

A longer test disproves.

So, different hashes with the same set of keys won't always have the same order. For me the program below demonstrates that two hashes with keys qw(a b c d e f) can differ in ordering:

v5.16.0
%h1: ecabdf
%h2: eadcbf

Program:

#!/usr/bin/env perl

use strict;
use warnings;
use feature qw(say);

# http://stackoverflow.com/q/12724071/132382

use constant KEYS => qw(a b c d e f);

my %h1 = map { $_ => undef } KEYS;
my %h2 = map { $_ => undef } KEYS;

delete @h2{'b', 'd', 'f'};
@h2{'x', 'y', 'z'} = ();
@h2{'b', 'd', 'f'} = ();
delete @h2{'x', 'y', 'z'};

say $^V;
say '%h1: ', keys(%h1);
say '%h2: ', keys(%h2);

Update

Here's a simpler demonstration that insertion order alone matters:

$ perl -MList::Util=shuffle -E \
> '@keys = ('a'..'z'); @h1{@keys} = @h2{shuffle @keys} = ();
> say keys(%$_) for (\%h1, \%h2)'
wraxdjyukhgftienvmslcpqbzo
warxdjyukhgftienmvslpcqbzo
#^^             ^^  ^^
#||             ||  ||