You need to look at the discussions of die() and warn() and the %SIG hash. You can set up your own handlers for errors (via die) using $SIG{__DIE__} = sub { ... } and for warnings (via warn) using $SIG{__WARN__} = sub { ... }. See the perlvar and perlfunc manpages. You can also run code that you think might die inside an eval { } construction and keep your program from dying.

#!/usr/bin/perl -w
use strict;

$SIG{__DIE__} = sub { die("My error: ", @_); };
$SIG{__WARN__} = sub { warn("My warning: ", @_); };

warn("some warning");
die("aarggh!");
[download]

Using eval {}; and $@ (which, if set, contains the error message that would otherwise be fatal -- see perldoc perlvar) will allow you to catch most errors:

eval {
 #code you want to catch errors from
};
custom_error_handler($@) if $@;
[download]

bikeNomad's answer is correct, with an addendum.

Using signals is not always straightforward. Signals are handled asynchronously and can interrupt each other. If you are, for example, redirecting those messages to a file, one signal might come in while the previous one is still processing, and stop it from going any further. You could loose some warnings that way, especially if they are being emitted from a tight loop and the handler is slow to execute.

In this case, it may be easier to redirect STDERR, which can be done with something like:

my $old_stderr = STDERR;
open STDERR, ">mylogfile.log";
[download]