RFR (S): 8048112: G1 Full GC needs to support the case when the very first region is not available

[Thomas Schatzl]

Hi Kim,

  thanks for the review.

On Fri, 2014-07-11 at 17:06 -0400, Kim Barrett wrote:
> On Jul 10, 2014, at 5:18 AM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> > 
> >> g1CollectedHeap.hpp
> >> 
> >> I assume the added const qualifier for n_regions() is because there is
> >> some caller that now requires this.  There are several nearby
> >> functions that look like they ought to be similarly qualified, but
> >> aren't, and that looks strange and confusing when first coming upon
> >> this code.
> > 
> > Done. Thanks for thinking of fitting code improvements beyond the
> > change.
> 
> And now we have a similar issue in HeapRegionSetBase, with length() looking
>lonely in its const-ness.
> 
> Unfortunately, late const-ness fixups do tend to propagate.  I think it would
>stop at HeapRegionSetBase in this case though.  Your call as to whether
>to continue the cleanup chain unrelated to the real purpose of this
>change set, or do cleanups in another changeset.

If you do not mind I would like to stop fixing const'ness at this point;
imo there is so much that I would like to change about the current heap
region sets that I want to revisit them rather sooner than later anyway.

E.g. you could continue arguing that some FreeRegionList methods should
be constified as well etc.

I updated JDK-6949241 with current ideas about that, sort-of-hijacking
that CR...

Thanks,
  Thomas