From @masak

<masak> std​: when 42 { say "This should fail" }
<p6eval> std 29347​: ok 00​:01 106m␤
<masak> ...because S04 says that a C<when> block outside of a
topicalizer block is a mortal sin.
<moritz_> at compile time?
<masak> aye.
<masak> it explicitly says that.
<moritz_> rakudo​: when 32 { say "ooops" }
<p6eval> rakudo 55f51d​: Use of uninitialized value␤
<masak> maybe I should submit a rakudobug, too? :)
<moritz_> sure
* masak submits

From @bbkr

NOM​: still broken

bbkr​:nom bbkr$ ./perl6 -e 'when 42 { say "This should fail" }'
Use of uninitialized value in numeric context

The RT System itself - Status changed from 'new' to 'open'

From @bbkr

2012.10 - still broken, but with different message

$ perl6 -e 'when 42 { say "This should fail" }'
use of uninitialized variable $a of type Any in numeric context in block at -e​:1

From @jnthn

On Wed Dec 16 08​:18​:11 2009, masak wrote​:

<masak> std​: when 42 { say "This should fail" }
<p6eval> std 29347​: ok 00​:01 106m␤
<masak> ...because S04 says that a C<when> block outside of a
topicalizer block is a mortal sin.
<moritz_> at compile time?
<masak> aye.
<masak> it explicitly says that.
<moritz_> rakudo​: when 32 { say "ooops" }
<p6eval> rakudo 55f51d​: Use of uninitialized value␤
<masak> maybe I should submit a rakudobug, too? :)
<moritz_> sure
* masak submits

Tried making this a mortal sin in a number of ways a month or two back. Actually discovered that there were loads of useful uses of when outside of a given/for-like construct that took advantage of every sub and block getting a fresh $_, setting it, and using when to check against it​:

sub foo(...) {
  ...
  $_ = something();
  when ... {
  }
  when ... {
  }
}

In discussion, nobody seemed to want to forbid this.

So, the design has been revised​:

Raku/old-design-docs@5f132abb41

Tests in S04-statements/when.t cover this behavior.

@jnthn - Status changed from 'open' to 'resolved'