From base.pm:

    die if $@ && $@ !~ /^Can't locate .*? at \(eval /;
[download]

So you won't ever hear of a file that couldn't be found, instead you get a useless message about the base class being empty:

                    Carp::croak(<<ERROR);
Base class package "$base" is empty.
    (Perhaps you need to 'use' the module which defines that package f
+irst.)
ERROR
[download]

... and that only if the namespace wasn't ever touched before. The following code will not fail nor warn despite the file Foo/Bar.pm not existing:

use strict;
BEGIN {
    print "Foo::Bar::VERSION : $Foo::Bar::VERSION\n";
};
package Foo::Baz;
use base 'Foo::Bar';
package main;
print "done\n.";
[download]

... because base.pm does not look in %INC to see whether a module was loaded.

That's a petty and trivial reason to not use base, no? And if it is such a problem for everyone ... how did that get out the production door? Why has there not been a "bug" fix?

Because Schwern (who added all those misfeatures) threw his hands up in despair and declared it unfixable. Hence, the fixed version was released under a new name, parent.pm.

And it's not a problem when things work, but once things fail, they fail ways that are hard to debug because whether a subclass gets loaded depends on program flow in a nonobvious way.

• p5p thread

That's a petty and trivial reason to not use base, no?

Considering there's no reason to use base in the first place...

Why has there not been a "bug" fix?

They did fix it with the creation of parent. Fixing it in place would break the following intended use of base:

{
   package Base;
   ...   
}

{
   package Subclass;
   use base 'Base';
   ...
}
[download]