Re: gperl_callback_invoke userdata copy

[Torsten Schoenfeld <kaffeetisch gmx de>]

Kevin Ryde wrote:
> > It's safer and more correct to use the copy,
>
> Does it also work to SvREFCNT_inc and sv_mortal, instead of a copy as
> such?  I guess the difference is then assigning to $_[-1] would modify
> the stored userdata for subsequent calls, would it?  (Maybe a good
> thing, maybe a very bad thing! :-).

I don't know if it will work -- did you try it?

Prior to changing gperl_callback_invoke to use sv_mortalcopy, we just pushed the
user data SV as is on the stack, so that assigning to $_[-1] would change the
original SV, I think.  With sv_mortalcopy, this changed.  So, if we want to
support assigning to $_[-1], then sv_mortalcopy should indeed be ditched in
favor of something that just adjusts reference counts.

Do we want to support assigning to $_[-1]?

(Note that passing variables by reference and modifying them inside the callback
still works.  sv_mortalcopy does a "shallow" copy.)
--