WIP: workmanager and bound services support #2464
Conversation
…into workmanager
| 'src/main/java/{}/{}Worker.java'.format( | ||
| args.package.replace(".", "/"), | ||
| name.capitalize() | ||
| ) |
There was a problem hiding this comment.
Minor: we could make a one-liner helper function/method that does that, so we have a single source of truth.
def get_target_path(package, name):
return 'src/main/java/{}/Service{}.java'.format(package.replace(".", "/"), name.capitalize())| @@ -71,7 +71,7 @@ int file_exists(const char *filename) { | |||
| } | |||
|
|
|||
| /* int main(int argc, char **argv) { */ | |||
| int main(int argc, char *argv[]) { | |||
| int main_(int argc, char *argv[], int call_exit) { | |||
There was a problem hiding this comment.
What happened to the main?
I've never played with that part, but won't this get linking issues that the main is not exposed?
There was a problem hiding this comment.
main only gets called from within the JNICALL functions, and my editor complained about the signature. so i renamed it, maybe start_python or so woud better fit.
| // * so main() will run main.py from this dir | ||
| // */ | ||
| // main_(1, argv, 0); | ||
| // } |
There was a problem hiding this comment.
Probably because it's still a WIP pull request, but I would not keep any commented out code in the final pull request
There was a problem hiding this comment.
Sure, i played around with running the python thread within the calling process (which is not the case for services). As long as no concurrent python gets started this works fine, but crashes otherwise. So this approach cannot be used when starting workers from e.g. kivy. It is required to use RemoteWorkerService/RemoteWorker in this case, which again uses a dedicated process. The commented function is a relict and will be removed during cleanup.
| @@ -356,7 +358,91 @@ int main(int argc, char *argv[]) { | |||
| return ret; | |||
| } | |||
|
|
|||
| JNIEXPORT void JNICALL Java_org_kivy_android_PythonService_nativeStart( | |||
| JNIEXPORT int JNICALL Java_org_kivy_android_PythonService_nativeStart( | |||
There was a problem hiding this comment.
Only thing that changes between Java_org_kivy_android_PythonService_nativeStart(), Java_org_kivy_android_PythonWorker_nativeStart() and Java_org_kivy_android_PythonBoundService_nativeStart() is the function name.
Maybe we should share the body via a commun function
…e to access the java class from python
|
I see all tests are passing - does anything else need to be done with this PR before it's ready to go? Very keen to see this merged. |
|
Any objections if I rebase this and polish it up? We'd like to be able to use this functionality. |
Unless @rnixx wants to take care of it, I don't see any blocking. @AndreMiras considering that you already started the review, any thoughts? |
|
Sure go for it. I'm not as active as I used to be for test/reviews unfortunately |
|
Sorry for leaving this PR abandoned for so long. Of course if have no problem if the work done so far gets picked up by someone else. Referring to #2613 We had the use case to use p4a for a service only inside a native Kotlin application. In an accepted solutions this case ideally continues to work. Referring the removal of the bound service - honestly i am a little out of our original context right now - this need to be double checked about the original intention. |
|
Hi, |
This is a work in progress PR addressing upcoming restrictions with foreground services as of android 12.
See https://developer.android.com/about/versions/12/foreground-services
2 major improvements are implemented:
Currently the implementation lives in service_library bootstrap as this is what's needed right now for a project.
Maybe it makes sense to create a dedicated bootstrap for this features, as especially work managers might be very useful if called from a kivy app via jnius (e.g. tracking GPS periodically). Therefor we'd need a mechanism to optionally "merge" several bootstraps, e.g. a kivy app might depend on sdl and workmanager bootstraps or you would like to create a aar library including workmanager feature. We also might just add workmanager and bound service support to common bootstrap, but this implies some more build dependencies (androidx.work:work-runtime, androidx.work:work-multiprocess, androidx.concurrent:concurrent-futures)
While playing around with starting python in main.c, we came along the problem with injecting sys.exit when python ends. When using workmanager workers this would just kill the entire application so we made it optional. If someone can explain why injecting sys.exit is needed at all this would be nice.
TODO