general answer: have a look at the -D switch in perlrun , IIRC it can be used to log all call frames.

Otherwise: it really depends on the structure of your tests, maybe it helps to have only one centralized exit point which is called via a sub xreturn() which replaces your old return .

something like: (untested)

my @data;

sub xreturn {
  @data = @_;
  push @data, [(caller)];
  goto CENTRAL_EXIT;
}

CENTRAL_EXIT:
  return @data;
[download]

HTH! :)