On Mon, Dec 08, 2008 at 02:32:14PM -0600, Patrick R. Michaud wrote:
: A very interesting question came up on #perl today, so I'm
: forwarding it to p6l for discussion/decision.
:
: Given the following code:
:
: sub foo() { return 1; }
: sub bar() { warn "oops"; }
:
: {
: CONTROL { ... }
: foo();
: bar();
: }
:
: S04 seems to clearly indicate that the CONTROL block above would
: be invoked by the C<warn> exception thrown by bar(). The #perl6
: question is, do the same semantics apply for C<return> -- i.e.,
: would the return exception thrown in foo() also invoke the
: CONTROL block in the caller?
:
: Both interpretations have validity, which is why I'm bringing
: it here for further reflection and guidance.
foo()'s implicit CONTROL handles the return after checking that the
exception's target context happens to match &?ROUTINE, so the outward
search for a handler terminates at that point. Doing it the other way
would make it impossible to optimize return into a goto end-of-routine.
Unlike warn, return is normal control flow, and we need to be careful
not to mandate largely useless overhead.
Larry
