Inline constructors/destructor for JavaRef.

Inline constructors/destructor for JavaRef.

JavaRef<object> only contains a POD type (jobject); the destructor does
nothing and the default constructor just NULL-initializes. Inline both.
Also inline the non-default constructor when !DCHECK_IS_ON (but leave it
out of line otherwise as it contains a nontrivial DCHECK).

This saves 8736 bytes in the .text section as of the time of writing,
but will save 40kB or more as we wrap more JNI method parameters in
JavaRef.

BUG=519562
Committed: https://crrev.com/cf27355590f9fff414fdcd86ac4fee074d1e6271
Cr-Commit-Position: refs/heads/master@{#345114}

[Torne, 2015-08-24 16:30:52]

torne@chromium.org changed reviewers:
+ rmcilroy@chromium.org

[Torne, 2015-08-24 16:30:53]

Yes, I know what the C++ dos & donts guide says about inlining
constructors/destructors, but for smartpointer types that are used in many
places I think we should make an exception. I verified that each of the three
inlinings is individually a code size reduction in release, and that the
generated ARM code is as I expect. :)

[rmcilroy, 2015-08-24 16:40:19]

lgtm with a suggestion.

https://codereview.chromium.org/1308413003/diff/1/base/android/scoped_java_ref.h
File base/android/scoped_java_ref.h (right):

https://codereview.chromium.org/1308413003/diff/1/base/android/scoped_java_re...
base/android/scoped_java_ref.h:50: JavaRef() : obj_(NULL) {}
Could you add a comment here that no other code should be added to the
constructors to avoid causing a size increase due to inlining (for all three
methods).

[Torne, 2015-08-24 16:47:26]

On 2015/08/24 16:40:19, rmcilroy wrote:
> lgtm with a suggestion.
> 
>
https://codereview.chromium.org/1308413003/diff/1/base/android/scoped_java_ref.h
> File base/android/scoped_java_ref.h (right):
> 
>
https://codereview.chromium.org/1308413003/diff/1/base/android/scoped_java_re...
> base/android/scoped_java_ref.h:50: JavaRef() : obj_(NULL) {}
> Could you add a comment here that no other code should be added to the
> constructors to avoid causing a size increase due to inlining (for all three
> methods).

Commented all three explaining what's going on.

[Torne, 2015-08-24 16:50:20]

The CQ bit was checked by torne@chromium.org

[Torne, 2015-08-24 16:50:20]

The patchset sent to the CQ was uploaded after l-g-t-m from
rmcilroy@chromium.org
Link to the patchset: https://codereview.chromium.org/1308413003/#ps40001
(title: "also comment .cc file")

[commit-bot: I haz the power, 2015-08-24 16:50:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1308413003/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1308413003/40001

[commit-bot: I haz the power, 2015-08-24 17:41:39]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-08-24 17:42:19]

Patchset 3 (id:??) landed as
https://crrev.com/cf27355590f9fff414fdcd86ac4fee074d1e6271
Cr-Commit-Position: refs/heads/master@{#345114}