From @masak

<masak> m​: CHECK { say "switching to runtime" }; foo(); macro foo() {
say "OH HAI" }
<camelia> rakudo-moar 3fa1bd​: OUTPUT«switching to runtime␤OH HAI␤»
<masak> TimToady​: what should the semantics be of a macro that's
called before its declaration? error at CHECK time?
<masak> TimToady​: I think the above semantics is basically an emergent
accident, and should be fixed.
<masak> it's too surprising that the exact same statement might meant
two wildly different things depending on whether it comes before or
after its definition.
<jnthn> masak​: I'm inclined to agree
<masak> ok, that's good enough for me.
<masak> spec clarification coming up.
<jnthn> m​: say A; class A { }
<camelia> rakudo-moar 3fa1bd​: OUTPUT«===SORRY!=== Error while
compiling /tmp/5iWGahghKE␤Illegally post-declared type​:␤ A used at
line 1␤␤»
<jnthn> I think given a macro is als something effective at compile
time, we can argue it along the same lines as this
<jnthn> And maybe even implement it using the same mechanism.
<dalek> specs​: 272d12f | masak++ | S06-routines.pod​:
<dalek> specs​: clarify that macros can't be called before they're defined
<dalek> specs​: review​: Raku/old-design-docs@272d12fc84
<masak> ...and because it's now spec... :)
* masak submits rakudobug

To be clear, I would expect calling the macro `foo()` to result in an
error similar to the use of `A` before its definition.

@coke - Status changed from 'new' to 'stalled'