We're getting extra helpings of humble pie today, that's for sure. My bad.

In a prior version of the code, I'd put reverse in a few of the functions, for example:

sub returns_rtwo { return reverse qw[ foo bar ]; }

To cut down on duplication, I'd left them out.

Here's a little snippet that takes that one step further:

#!/usr/bin/perl -w

use strict;

sub foo_p($) { print "$_[0]\n" }
sub foo      { print "$_[0]\n" }

sub baz { qw[ foo bar ]     }
sub zab { reverse baz()     }
sub bam { return (baz(),zab())  }
sub bar { return bam()      }

foo(bar());
foo_p(bar());
[download]

On a regular, unprotyped function, you get your regular 'foo', but a prototyped version of same is a whole different thing.

If you've got prototyped functions, you're going to have to get serious, I suppose, and stick to the letter of the law. No extra parameters, no mis-typed values, and no "random junk".

Update:
Considering the normal, unmolested return value of bar() is 'foo','bar','bar','foo', it is odd that you get 'raboof' because of the forced scalar(). The last function is trying to return an array with four values, so you'd think, at least on some plane of existence, that you'd get 4 in the end, but this is not the case. The only way you get this is if you assign to a temporary array:

sub bar { return my @foo = bam(); }