Patch: gcj ggc patch

[Per Bothner]

Tom Tromey wrote:
> I don't really know whether this patch is necessary.  I found other
> occurrences of ggc_add_tree_root in similar situations in jcf-write.c.
> And as far as I can tell these static variables weren't otherwise
> marked as roots.
> 
> Is this required?  If so, is it ok to commit?

 From what I can tell:  No, and no.

init_stringpool (in stringpool.c) marks the entire stringpool
hash table,

But it would be good to get an authoritative ruling:
Does an identifier node stashed in a global variable need
to be explicitly ggc-registerd?
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/