Re: MPS signals and Emacs

[Gerd Möllmann]

Eli Zaretskii <eliz@gnu.org> writes:

>> I think I've also seen at least one other example, but I forget where
>> that was (maybe specpdl, or bytecode?). I think the function called
>> is mark_mem or so.
>
> So any Lisp object referenced from the C stack or from byte-code will
> not be copied by MPS?  

Yes.

> What are the implications of that for Emacs? AFAIU, the current GC
> considers any such object as live, marks it, and doesn't collect it as
> garbage during the sweep step, but being unable to copy such an object
> means something very different for a copying GC, doesn't it? Because
> AFAIU live objects _are_ copied.

No, it's a bit different. Being able to copy and objects being live are
orthogonal. At the heart of the mostly-copying idea is simply that
ambigupusly references aren't moved in memory (so they aren't copied).
That doesn't mean these objects aren't live. They are.