TIMOB-6242 Android bitmap memory leak

[billdawson]

http://jira.appcelerator.org/browse/TIMOB-6242

Please test at least these two things (in V8 and Rhino):

• The original failcase in the description of the JIRA item.
• This additional test.

[marshall]

The reason for the V8 / Rhino difference probably has to do with JNI Global references on the native side for the V8 runtime.

Technically MAT is correct in saying that each proxy is a GC root, because in JNI, to hold a reference to a Java object inside a V8 JS Object, we use a Global reference. The Java object also holds a pointer to the underlying Persistent V8 reference, so when either Garbage Collector kicks in, we free both references.

There may be another leak in our JS that causes V8 to hold on to the original view instance, possibly in our window.js?

[marshall]

Code and Functional review passed, Request accepted.

[rusticphilosopher]

Code reviewed and accepted. Merging.