This is not one of my stronger areas, but my initial guess is that you haven't given us enough information. What sort of object is $webcrawler ? Have you tried stepping through with the debugger to find out why that sub in HTTP::Response is getting an undefined $base? Shouldn't you be assigning something to $url_name before using as an arg to  $webcrawler->get ?

While not being able to answer your real question, I'll point out that your grep regex should probably be like this:

my $exts = join( "|", qw/pdf doc ps txt/;
@links = grep( !/\.(?:$exts)$/, @links );
[download]

The difference is in the period being treated as a literal period (your version treats it as a wildcard), and having the the set of alternative extension strings match only at the end of a link string (your version only matchs $ext3 (.txt) at the end of the string, and the others can match anywhere in the string).

(updated to fix grammar)