[m-dev.] diff: accurate gc typeinfo variable bug.

[Fergus Henderson]

Tyson Richard DOWD, you wrote:
> 
> 	Fergus, can you please review this?

> +		% If a variable is unused in the goal, it shouldn't be
> +		% in the initial liveness. (If we allowed it to start
> +		% live, it wouldn't ever become dead, because it would
> +		% have to be used to be killed).
> +		% So we intersect the headvars with the non-locals.
> +	proc_info_goal(ProcInfo, _Goal - GoalInfo),
> +	goal_info_get_nonlocals(GoalInfo, NonLocals),
> +	set__intersect(Liveness2, NonLocals, Liveness3),
> +		% If doing accurate garbage collection, the corresponding
> +		% typeinfos need to be added to these.
> +	module_info_globals(ModuleInfo, Globals),
> +	globals__get_gc_method(Globals, GCmethod),
> +	(
> +		GCmethod = accurate
> +	->
> +		proc_info_get_typeinfo_vars_setwise(ProcInfo, Liveness3,
> +			TypeInfoVars),
> +		set__union(Liveness3, TypeInfoVars, Liveness)
> +	;
> +		Liveness = Liveness3

Are you sure that these two steps are done in the right order?
Shouldn't the typeinfo vars be intersected with the non-locals too?

-- 
Fergus Henderson <fjh at cs.mu.oz.au>   |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>   |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3         |     -- the last words of T. S. Garp.