Review overriddable vtable interfaces

[whiteknight]

we should go through the list of vtable interfaces to make sure we are allowing overrides on all the interfaces that we need to use. Here is a list of vtable interfaces that currently don't appear to support overriding:

init
init_pmc
destroy*
mark*
get_attr_keyed
set_attr_keyed
can
isa_pmc
isa
does
does_pmc
type*
invoke (we have it, but it doesn't work correctly)
visit*
thaw*
thawfinish*
share_ro*

I've marked with a * some of the vtable entries above that really can't be overridden in PIR or that don't make sense to be. Are there any others from this list that should be overridable?

And while I'm at it, are there any methods currently that are overridable, but really shouldn't be?

[whiteknight]

"does" is now overridable. I suspect that we should probably allow overriding of does_pmc for symmetry. Are there any others that we would like to be able to override? I'm thinking "can" and "isa" might be nice. Any opinions?

[coke]

IMO, everything should be overridable without a documented rationale to the contrary.

[whiteknight]

I'm closing this ticket. It was an RFC that attracted very little attention. The list of overridable VTABLEs we have now is clearly sufficient for the needs of most HLLs. I sincerely hope that we completely redesign our object metamodel and completely reimplement the vtable system in the future, instead of needing to revisit issues like this.