On Viernes 17 Octubre 2008, Nikodemus Siivola wrote:
> On Thu, Oct 16, 2008 at 6:45 PM, Gábor Melis <mega@...> wrote:
> > But mainly I'd like to get some feedback on whether the above frame
> > handling hack looks reasonable to commit.
> I like the way it's take care of in ALIEN-FUNCALL, as opposed to the
> CALL-OUT VOP.
> I think it's OK, but perhaps the annotation could be omitted from
> call-sites with high speed & low debug? Dunno. If that seems
> necessary it can be take care of later as well, I suppose.
I see the point but that would affect the debuggability of all functions
that call it which could be quite counter-intuitive and annoying.
> Is there a reason why you don't stack-allocate the CONSes on
No, I'll change that.
> Also, I did not check this properly, but it seems to me that
> something like
> (eval `(align-funcall ...))
> would not get the benefit of this patch?
I think all calls to ALIEN-FUNCALL (the defun, not the transformed
variants) end up calling a compiled lambda with another ALIEN-FUNCALL
in it and that is transformed. Else, we'd miss the pinning of arguments
> -- Nikodemus