sub croak   { die  shortmess @_ }
sub confess { die  longmess  @_ }
[download]

In the case of Carp specifically there is a far easier way. Use Carp::Heavy. Then redefine Carp::longmess_heavy and Carp::shortmess_heavy.

Otherwise you'd have to walk the symbol table. Just to get you started on that, print out the keys of %main::. All of the keys ending in :: are subtables you can look for. Beware of infinite loops. (For instance %main::main::main:: is the same as %main::.)

Strange, why aren't you just setting up your own $SIG{__DIE__} handler? It seems tailor-made to solve your problem for you. Granted, it's a hack, and can get messy, but it's a far better hack than trying to overload another module's subroutines globally.

A long time back I went down this road, as well. I thought it was a great idea to setup a __DIE__ handler and catch all the errors, turning them into pretty output for a web framework I was designing. I suppose it worked, sorta. It's much cleaner to design an actual exception model, with exception classes thrown by code, and possibly a catch routine that can transform string exceptions into a relevant object.

If the object in $@ has stringification it shouldn't break anything that's attempting to match against it. The only thing it should "break" is something that's checking to see if it's an object, in which case that code should know what it's doing.

Because $SIG{__DIE__} is too often used by other packages as well. I think my final solution is cleaner in regards to what I want to do.

Could you just redefine Carp::shortmess and Carp::longmess? If not, you can walk the symbol table starting with something like this:

for (grep /(?<!main)::/, keys %::) {
    # do stuff...
}
[download]