[m-dev.] Re: for review: new method of handling failures, part 1 of 6
zs at cs.mu.OZ.AU
Tue Jul 7 17:40:13 AEST 1998
> > I don't know the reason. I suppose we could reverse the order, see what
> > breaks, and document that :-(
> Well, goal_info.m has a slightly more detailed comment:
> % ... Note that when traversing forwards
> % through a goal, deaths must be applied before births;
> % this is necessary to handle certain circumstances where a
> % variable can occur in both the post-death and post-birth sets,
> % or in both the pre-death and pre-birth sets.
Yes, but this doesn't say what those circumstances are, which is the crux
of the issue.
>> The design above requires the allocation and filling of only a one-word cell;
> Does it even require that? Won't the position_info be a `no_tag' type
> for which there is no runtime overhead at all?
Yes, you are right. I always forget about no_tag, perhaps because I didn't
design it :-)
More information about the developers