Paul Evans

> For XS accessor generation modules this'd break the following use case (extensively present in DBIx::Class) ... New devel perl will do things that CPAN modules that poke at optrees...

> Could you just update the top item on the mark stack (set to `svp+1`), instead of doing this Move? > (I haven't thought this through, just musing.) That was...

> 1) I'm vaguely nervous that this is touching heavily in areas related to my current and next projects: making the stack reference counted, and doing more signature work (especially...

Plus I should add that all of the comments above about entersub on coderefs also apply to method calls: ```perl my $z = $obj?->meth( gen_args() ); my $i = objA()->meth(...

> One small doc nit. Wording fixed. > Otherwise, it'd be nice to have tests... Yeeeah. Hrm. A test could be written but that's basically going to be a lot...

I've added an entry to `perldelta.pod` as well, before merge.

This seems a good idea, as it's already a mechanism familiar to CPAN authors, being the standard way those dists usually work.

What about attributes on signatured lexical subs? Currently they complain: ```perl use feature 'signatures'; sub MODIFY_CODE_ATTRIBUTES($p, $t, @at) {} my sub f ($arg) :Final {} ``` ``` The signatures feature...

I should clarify my comment: The fix skips the warning conditionally on `in_my`. But `my sub ...` will also be `in_my`. So I believe the fix will stop warning about...

I suspect fixing this would be relatively simple, involving a flag on `OP_ENTERSUB` to tell it to look at the other end of the stack for the CV to invoke,...