Actually, it is a feature of how Perl's global destruction is not done in an orderly manner in order to save time. Whether or not you get the error depends on some rather random things.

I don't see how not using -w would help as trying to call a method on an undefined value does not require -w in order to trigger. Perhaps you are just getting lucky.

You don't have to upgrade Perl to solve this problem. Simply upgrade Win32::TieRegistry. If you go to http://search.cpan.org/ and search for "TieRegistry", you'll find version 0.24. Click on the very first link and then click on the Changes file to see:

Revision history for Perl extension Win32::TieRegistry.

0.24  2001-02-06 19:00
        - Silenced error that could appear "at random" during "cleanup":
                (Can't call method "FETCH" on an undefined value)

In any case, I was able to download the module. Then unzip the file and install it like any CPAN module (or, if you don't have a version of "make", run the included Install.bat). The latest version may also be available in PPM form from some repository (I don't keep track much since I don't seem to have the ability to affect what gets made available).

As for the error being harmless, you are mostly correct. The only thing related to that error that is a problem is fairly minor, especially considering that it only affects a feature that is rarely used. From the Bugs section of manual:

Current implementations of Perl's "global destruction" phase can cause objects returned by Load() to be destroyed while keys within the hive are still open, if the objects still exist when the script starts to exit. When this happens, the automatic UnLoad() will report a failure and the hive will remain loaded in the Registry.

Thank you Brother Tye for an excellent exposition of the root problem! I only received this error when exiting each program. I do have a follow-on question for you:

Do I now have to manually delete entries from my registry (using regedit) to clean it up? Your last sentence indicates "the hive will remain loaded in the Registry?" Thanks!

No. You can get the error once for each Registry key object that you (or anyone) stored in a global variable. But you can only run into the bug if you used the Load() in such a way that you expected the module to automatically unload once you close the key (and also in such a way that the error could be triggered for that key or some key within the hive you Load()ed). Since you probably didn't Load() any keys at all, then there would be no chance of leaving any keys loaded.

If you'd left keys loaded, it is easy to notice when you run regedt32 (there'd be some unexpected keys with names that include "TieRegistry" and some sequence number hanging out). Other than the clutter, leaving the keys loaded does little harm (it leaves the hive file locked) and the "problem" goes away when you reboot.

        - tye (but my friends call me "Tye")