On 12/23/2017 08:23 PM, Eirik Berg Hanssen wrote: > On Sat, Dec 23, 2017 at 5:38 PM, Sawyer X <xsawyerx@gmail.com > <mailto:xsawyerx@gmail.com>> wrote: > > > We can document that every compilation-affecting subroutine attribute > should be used before. Another way is to add a mechanism to > register an > attribute as "requiring compilation," allowing us to croak about > them if > we see them after signatures. Not an optimal solution, to say the > least. > > > If you're really considering this, please go the other way: Require > such registration of the attribute before allowing it after the signature. Wouldn't it depend on how likely attributes are to affect or not affect compilation?Thread Previous | Thread Next