jni_generator: Always generate native exports.

jni_generator: Always generate native exports.

Native exports work better (faster startup) and generally only have one
real downside: incompatibility with the chromium custom linker. We
compile all code with --native_exports --native_exports_optional for a
while now, which provides the native exports while also having explicit
registration functions for the custom linker case.

We may want to build with only native exports in future, but it doesn't
seem likely we'll want to go back to only having explicit registration,
so simplify the code (and improve the test coverage) by:

1) Removing the native_exports flag and making it always true.
2) Changing the tests to use optional native exports by default,
   matching Chromium behaviour, except for a specific test to verify
   that native-exports-only still works.

This will make it easier to eliminate RegisterNativesImpl methods in
cases where it's currently a no-op in crbug.com/603936, as there will be
fewer cases to consider.

BUG=603936
Committed: https://crrev.com/0cf1db3d8a236c11ecad2201e78981dfb7cc951c
Cr-Commit-Position: refs/heads/master@{#406815}

[Torne, 2016-07-19 13:17:20]

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

[Torne, 2016-07-19 13:17:21]

Ross, does this look reasonable? We've used only native_exports_optional in
chromium for a while now.

Dan, when this lands you can rebase your other change onto this and it should be
easier to think about :)

[rmcilroy, 2016-07-21 09:23:51]

On 2016/07/19 13:17:21, Torne wrote:
> Ross, does this look reasonable? We've used only native_exports_optional in
> chromium for a while now.

Looks reasonable to me. LGTM.

[Torne, 2016-07-21 09:30:41]

The CQ bit was checked by torne@chromium.org

[commit-bot: I haz the power, 2016-07-21 09:30:57]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-21 10:42:14]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-07-21 10:43:36]

Description was changed from

==========
jni_generator: Always generate native exports.

Native exports work better (faster startup) and generally only have one
real downside: incompatibility with the chromium custom linker. We
compile all code with --native_exports --native_exports_optional for a
while now, which provides the native exports while also having explicit
registration functions for the custom linker case.

We may want to build with only native exports in future, but it doesn't
seem likely we'll want to go back to only having explicit registration,
so simplify the code (and improve the test coverage) by:

1) Removing the native_exports flag and making it always true.
2) Changing the tests to use optional native exports by default,
   matching Chromium behaviour, except for a specific test to verify
   that native-exports-only still works.

This will make it easier to eliminate RegisterNativesImpl methods in
cases where it's currently a no-op in crbug.com/603936, as there will be
fewer cases to consider.

BUG=603936
==========

to

==========
jni_generator: Always generate native exports.

Native exports work better (faster startup) and generally only have one
real downside: incompatibility with the chromium custom linker. We
compile all code with --native_exports --native_exports_optional for a
while now, which provides the native exports while also having explicit
registration functions for the custom linker case.

We may want to build with only native exports in future, but it doesn't
seem likely we'll want to go back to only having explicit registration,
so simplify the code (and improve the test coverage) by:

1) Removing the native_exports flag and making it always true.
2) Changing the tests to use optional native exports by default,
   matching Chromium behaviour, except for a specific test to verify
   that native-exports-only still works.

This will make it easier to eliminate RegisterNativesImpl methods in
cases where it's currently a no-op in crbug.com/603936, as there will be
fewer cases to consider.

BUG=603936

Committed: https://crrev.com/0cf1db3d8a236c11ecad2201e78981dfb7cc951c
Cr-Commit-Position: refs/heads/master@{#406815}
==========

[commit-bot: I haz the power, 2016-07-21 10:43:37]

Patchset 1 (id:??) landed as
https://crrev.com/0cf1db3d8a236c11ecad2201e78981dfb7cc951c
Cr-Commit-Position: refs/heads/master@{#406815}