Improve javac.py --incremental logic by having it recompile more files

Fix javac.py --incremental logic

Fix #1 - It now recompiles .java files that reference changed files.
Fix #2 - It now extracts all needed .class files (was messing up on
         Foo.class vs. FooImpl.class due to prefix matching)

Also renames the debug environment variable:
  PRINT_MD5_DIFFS -> PRINT_INCREMENTAL_EXPLANATIONS

BUG=533442

[agrieve, 2015-09-25 17:32:31]

agrieve@chromium.org changed reviewers:
+ yfriedman@chromium.org

[agrieve, 2015-09-25 17:33:30]

On 2015/09/25 17:32:31, agrieve wrote:
> mailto:agrieve@chromium.org changed reviewers:
> + mailto:yfriedman@chromium.org

(-’๏_๏’-)

[agrieve, 2015-09-25 17:50:49]

On 2015/09/25 17:33:30, agrieve wrote:
> On 2015/09/25 17:32:31, agrieve wrote:
> > mailto:agrieve@chromium.org changed reviewers:
> > + mailto:yfriedman@chromium.org
> 
> (-’๏_๏’-)

Note: I did measure, and this new step adds about 100ms.

[agrieve, 2015-09-25 17:59:00]

agrieve@chromium.org changed reviewers:
+ cjhopman@gmail.com

[agrieve, 2015-09-25 17:59:25]

On 2015/09/25 17:50:49, agrieve wrote:
> On 2015/09/25 17:33:30, agrieve wrote:
> > On 2015/09/25 17:32:31, agrieve wrote:
> > > mailto:agrieve@chromium.org changed reviewers:
> > > + mailto:yfriedman@chromium.org
> > 
> > (-’๏_๏’-)
> 
> Note: I did measure, and this new step adds about 100ms.

+ cjhopman2

[Yaron, 2015-09-25 18:34:10]

lgtm per offline discussion provided:
1) we add an assert / build failure if the flag is used in release mode
2) this doesn't become the default

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py
File build/android/gyp/javac.py (right):

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
build/android/gyp/javac.py:130: class_name_pattern = re.compile(r'\b(?:%s)\b' %
'|'.join(changed_class_names))
I'm having trouble thinking this through but make sure you've considered inner
classes here? I guess changed_paths would only ever refer to the top-level java
file so should be ok on that end. In terms of the grep, I think it should also
match top-level class or inner classes so seems ok?

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
build/android/gyp/javac.py:158: return ret
merge with line above

[agrieve, 2015-09-25 18:49:44]

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py
File build/android/gyp/javac.py (right):

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
build/android/gyp/javac.py:130: class_name_pattern = re.compile(r'\b(?:%s)\b' %
'|'.join(changed_class_names))
On 2015/09/25 18:34:09, Yaron wrote:
> I'm having trouble thinking this through but make sure you've considered inner
> classes here? I guess changed_paths would only ever refer to the top-level
java
> file so should be ok on that end. In terms of the grep, I think it should also
> match top-level class or inner classes so seems ok?

You can't reference an inner class without typing the outer-class's name.

However...

Just thought of the case where you do:  obj.getFoo().doThing()

You never reference the type of getFoo(), yet still depend on it :( 

Back to javap...

https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
build/android/gyp/javac.py:158: return ret
On 2015/09/25 18:34:09, Yaron wrote:
> merge with line above

Done.

[agrieve, 2015-09-25 20:24:34]

On 2015/09/25 18:49:44, agrieve wrote:
> https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py
> File build/android/gyp/javac.py (right):
> 
>
https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
> build/android/gyp/javac.py:130: class_name_pattern = re.compile(r'\b(?:%s)\b'
%
> '|'.join(changed_class_names))
> On 2015/09/25 18:34:09, Yaron wrote:
> > I'm having trouble thinking this through but make sure you've considered
inner
> > classes here? I guess changed_paths would only ever refer to the top-level
> java
> > file so should be ok on that end. In terms of the grep, I think it should
also
> > match top-level class or inner classes so seems ok?
> 
> You can't reference an inner class without typing the outer-class's name.
> 
> However...
> 
> Just thought of the case where you do:  obj.getFoo().doThing()
> 
> You never reference the type of getFoo(), yet still depend on it :( 
> 
> Back to javap...
> 
>
https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
> build/android/gyp/javac.py:158: return ret
> On 2015/09/25 18:34:09, Yaron wrote:
> > merge with line above
> 
> Done.

oooooookay. After more research on the topic, I believe the best path forward is
to make use of:
https://github.com/pantsbuild/jmake/blob/master/src/org/pantsbuild/jmake/Main...

Good article on the subject:
http://blog.ltgt.net/most-build-tools-misuse-javac/

[agrieve, 2015-09-28 15:20:07]

On 2015/09/25 20:24:34, agrieve wrote:
> On 2015/09/25 18:49:44, agrieve wrote:
> > https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py
> > File build/android/gyp/javac.py (right):
> > 
> >
>
https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
> > build/android/gyp/javac.py:130: class_name_pattern =
re.compile(r'\b(?:%s)\b'
> %
> > '|'.join(changed_class_names))
> > On 2015/09/25 18:34:09, Yaron wrote:
> > > I'm having trouble thinking this through but make sure you've considered
> inner
> > > classes here? I guess changed_paths would only ever refer to the top-level
> > java
> > > file so should be ok on that end. In terms of the grep, I think it should
> also
> > > match top-level class or inner classes so seems ok?
> > 
> > You can't reference an inner class without typing the outer-class's name.
> > 
> > However...
> > 
> > Just thought of the case where you do:  obj.getFoo().doThing()
> > 
> > You never reference the type of getFoo(), yet still depend on it :( 
> > 
> > Back to javap...
> > 
> >
>
https://codereview.chromium.org/1373443004/diff/1/build/android/gyp/javac.py#...
> > build/android/gyp/javac.py:158: return ret
> > On 2015/09/25 18:34:09, Yaron wrote:
> > > merge with line above
> > 
> > Done.
> 
> oooooookay. After more research on the topic, I believe the best path forward
is
> to make use of:
>
https://github.com/pantsbuild/jmake/blob/master/src/org/pantsbuild/jmake/Main...
> 
> Good article on the subject:
> http://blog.ltgt.net/most-build-tools-misuse-javac/

Abandoning this patch. Filed a dedicated by for jmake:
https://code.google.com/p/chromium/issues/detail?id=536817