My reading is that OP did want to be interrupted, but didn't want to jump straight to the end of the program. He wants the interrupt to kill the current eval-block, but then to do some more processing. I could make his example a bit more tricky by requiring that we recover from the interrupt, rather than dying. I.e.

START:
my $interrupted = 0;
eval {
  local $SIG{INT} = sub { $interrupted = 1; die };
  ...
}
if ($@) {
  if ($interrupted) {
    print "Operation was interrupted by ^C. Press <return> to retry, ^
+C to die\n";
    scalar <STDIN>;
    goto START;
  }
}
[download]

If the interrupt hits when the code is in a DTOR, then it will not kill the eval block, and the user will not be asked to retry. In fact, the only effect of the interrupt will be to set $interrupted; and to cause some resource to not be properly cleaned up. We could detect that this happenned by checking:

if ($@) {
  ...
} elsif ($interrupted) {
  print "interrupted but didn't die. Must have hit ^C during DTOR. Sor
+ry.\n";
  die "better late than never!\n";
}
[download]