jni_generator: Wrap all native methods in stubs.

jni_generator: Wrap all native methods in stubs.

Instead of only wrapping native methods which call C++ methods, wrap
them all. This doesn't require any changes to the implementations, but
makes the handling of the two kinds of method more consistent in
preparation for introducing return type/parameter wrapping in future
changes.

This also changes the way that some of the stub code is generated to
make it less different depending whether native exports are in use, and
moves "extern "C"" declarations to be on only the specifically exported
symbols instead of applying to all forward declarations.

BUG=519562
R=rmcilroy@chromium.org

Committed: https://chromium.googlesource.com/chromium/src/+/b37ad14e10041f15a4467022d5169be4dac73ceb

[Torne, 2015-08-11 16:58:20]

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

[Torne, 2015-08-11 16:58:21]

Some refactoring before I try out wrapping parameters/return values for all jni
methods.. no functional change intended here. (the stubs should always have the
real method inlined into them so shouldn't cost anything).

[Torne, 2015-08-12 10:18:50]

On 2015/08/11 16:58:21, Torne wrote:
> Some refactoring before I try out wrapping parameters/return values for all
jni
> methods.. no functional change intended here. (the stubs should always have
the
> real method inlined into them so shouldn't cost anything).

Checked, and this change increases libchrome.so .text size by 32 bytes. Not sure
why; possibly one of the methods is too big to inline even though it's only
called once, but that seems within the scope of random variation :)

[rmcilroy, 2015-08-12 10:35:35]

Thanks for checking the size implications. 32 bytes seems reasonable :).

lgtm

https://codereview.chromium.org/1279163006/diff/1/base/android/jni_generator/...
File base/android/jni_generator/jni_generator.py (right):

https://codereview.chromium.org/1279163006/diff/1/base/android/jni_generator/...
base/android/jni_generator/jni_generator.py:1116: stub_visibility = 'static '
nit - pull this out to a helper function to avoid duplicating with above.

[Torne, 2015-08-12 10:37:56]

https://codereview.chromium.org/1279163006/diff/1/base/android/jni_generator/...
File base/android/jni_generator/jni_generator.py (right):

https://codereview.chromium.org/1279163006/diff/1/base/android/jni_generator/...
base/android/jni_generator/jni_generator.py:1116: stub_visibility = 'static '
On 2015/08/12 10:35:35, rmcilroy wrote:
> nit - pull this out to a helper function to avoid duplicating with above.

Actually, I could probably just merge these two functions at this point and have
it take a parameter for whether to treat it like a method or function, which is
the general goal anyway (to get to where the only difference is whether it
handles the "this" pointer for you). Will fiddle with that in a bit.

[Torne, 2015-08-12 16:36:35]

OK, I merged the two functions (in a way that doesn't change the generator
output at all, so it should be a safe refactor) in PS#2.

I then removed the separate forward declaration section entirely in PS#3 (so all
the stubs are just together in one place instead of separated by whether they
are calling C++ methods vs functions) which does change the output, but is a
much simpler change to the script.

Hopefully that makes this easier to review ;)

[rmcilroy, 2015-08-13 10:50:23]

On 2015/08/12 16:36:35, Torne wrote:
> OK, I merged the two functions (in a way that doesn't change the generator
> output at all, so it should be a safe refactor) in PS#2.
> 
> I then removed the separate forward declaration section entirely in PS#3 (so
all
> the stubs are just together in one place instead of separated by whether they
> are calling C++ methods vs functions) which does change the output, but is a
> much simpler change to the script.
> 
> Hopefully that makes this easier to review ;)

lgtm, thanks!

[Torne, 2015-08-13 10:58:08]

The CQ bit was checked by torne@chromium.org

[commit-bot: I haz the power, 2015-08-13 10:58:25]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1279163006/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1279163006/40001

[commit-bot: I haz the power, 2015-08-13 11:06:59]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-08-13 11:06:59]

Try jobs failed on following builders:
  chromium_presubmit on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/chromium_presub...)

[Torne, 2015-08-14 12:42:50]

Committed patchset #3 (id:40001) manually as
b37ad14e10041f15a4467022d5169be4dac73ceb (presubmit successful).