[OpenJDK Rasterizer] Marlin renderer contribution for review

[Phil Race]

On 03/25/2015 02:39 PM, Laurent Bourgès wrote:
>
> Hi Jim,
>
> > Briefly I checked out the line lengths only in Renderer.java and 
> they seem good now.
> Ok, I fixed also MarlinConst / RendererStats you pointed out.
>
> > But you changed a number of other things that I'm now going to have 
> to review.  Can you summarize not just the "what changed", but also 
> the "where it changed" so I don't have to look over all parts of all 
> files trying to find the changes?
> Ok,  sorry. I pushed also my changes onto github that has a nice diff 
> tool... you can not use.
>
> You could maybe compare both marlin.x/jdk.patch files.
>
> > Also, if we can hold off on further development work until we get a 
> stake in the ground it would avoid this "having to review a huge 
> changeset over again from scratch" due to ancillary fixes.
>
> Agreed. It was minor changes compared to the huge marlin patch: I will 
> give you hints below.
>
> > Also, the very first file I brought up still had /** */ single line 
> faux-doc comments (AAShapePipe, on private fields no less).  That was 
> the second thing you said that you fixed below...?
>
> I disagree: this is a typical javadoc comment for a field (even 
> private). Maybe you dislike javadoc for fields.
> What's clearly the problem ? Do you want me to remove all new comments 
> / javadoc ?
>

Javadoc actually has an option to generate the doc for private members ( 
-private) :-
http://docs.oracle.com/javase/6/docs/technotes/tools/windows/javadoc.html#javadocoptions
so I suppose its not considered illegal and won't cause any problems 
that I know of,
although its not the default.

-phil.

> I said I removed many of them that were redundant or not appropriate 
> javadoc:
> See RendererContext.
>
> I do my best to make it right and spent many hours on preparing & 
> testing the latest webrev.
>
> I agree it can be painful on your side.
>
> >>     - Fixed Unsafe access
>
> See Renderer lines 52 to 75.
>
> >>     - Use PhantomReference / ReferenceQueue and the new OffHeapDisposer
> >>     thread to free off-heap memory to avoid finalization in Renderer
>
> See Renderer lines 52 to 75 and lines 1379 to 1415.
>
> I forgot to remove lines 530 to 537 = deprecated finalize method.
>
> >>     - added the ArrayCachesHolder class in RendererContext to 
> gather all
> >>     ArrayCache instances: use it wrapped using a WeakReference to 
> reduce
> >>     the memory footprint of large array caches
> In RendererContext lines 70 and 140-170 and 340 to eof.
>
> >>     - Use AccessController.doPrivileged(new GetPropertyAction(key)) to
> >>     properly get System properties in a secure environment in
> >>     MarlinRenderingEngine
>
> MarlinRenderingEngine:  Line 830 to eof.
>
> If you prefer, I can send a diff between webrevs or do more cleanup ?
>
> Laurent
>