[FFmpeg-devel] [PATCH 2/7] avutil: add av_jni_{get, set}_android_app_ctx helper

[Matthieu Bouron]

On Tue, Feb 27, 2024 at 02:42:01PM +0100, Andreas Rheinhardt wrote:
> Matthieu Bouron:
> > This will allow users to pass the Android ApplicationContext which is mandatory
> > to retrieve the ContentResolver responsible to resolve/open Android content-uri.
> > ---
> >  libavutil/jni.c | 28 ++++++++++++++++++++++++++++
> >  libavutil/jni.h | 17 +++++++++++++++++
> >  2 files changed, 45 insertions(+)
> > 
> > diff --git a/libavutil/jni.c b/libavutil/jni.c
> > index 541747cb20..d6f96717b0 100644
> > --- a/libavutil/jni.c
> > +++ b/libavutil/jni.c
> > @@ -34,6 +34,7 @@
> >  #include "libavutil/jni.h"
> >  
> >  static void *java_vm;
> > +static void *android_app_ctx;
> >  static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
> >  
> >  int av_jni_set_jvm(void *vm, void *log_ctx)
> > @@ -63,6 +64,24 @@ void *av_jni_get_jvm(void *log_ctx)
> >      return vm;
> >  }
> >  
> > +void av_jni_set_android_app_ctx(void *app_ctx)
> > +{
> > +    pthread_mutex_lock(&lock);
> > +    android_app_ctx = app_ctx;
> > +    pthread_mutex_unlock(&lock);
> > +}
> > +
> > +void *av_jni_get_android_app_ctx(void)
> > +{
> > +    void *ctx;
> > +
> > +    pthread_mutex_lock(&lock);
> > +    ctx = android_app_ctx;
> > +    pthread_mutex_unlock(&lock);
> > +
> > +    return ctx;
> > +}
> > +
> >  #else
> >  
> >  int av_jni_set_java_vm(void *vm, void *log_ctx)
> > @@ -75,4 +94,13 @@ void *av_jni_get_java_vm(void *log_ctx)
> >      return NULL;
> >  }
> >  
> > +void av_jni_set_android_app_ctx(void *app_ctx)
> > +{
> > +}
> > +
> > +void *av_jni_get_android_app_ctx(void)
> > +{
> > +    return NULL;
> > +}
> > +
> >  #endif
> > diff --git a/libavutil/jni.h b/libavutil/jni.h
> > index 700960dbb8..630f4074a1 100644
> > --- a/libavutil/jni.h
> > +++ b/libavutil/jni.h
> > @@ -43,4 +43,21 @@ int av_jni_set_jvm(void *vm, void *log_ctx);
> >   */
> >  void *av_jni_get_jvm(void *log_ctx);
> >  
> > +/*
> > + * Set the Android application context which will be used to retrieve the Android
> > + * content resolver to resolve content uris.
> > + *
> > + * @param app_ctx global JNI reference to the Android application context
> > + */
> > +void av_jni_set_android_app_ctx(void *app_ctx);
> > +
> > +/*
> > + * Get the Android application context that has been set with
> > + * av_jni_set_android_app_ctx.
> > + *
> > + * @return a pointer the the Android application context
> > + */
> > +void *av_jni_get_android_app_ctx(void);
> > +
> > +
> >  #endif /* AVUTIL_JNI_H */
> 
> This adds global state in order to pass the application context to your
> protocol (in lieu of an option to pass pointers to a protocol). Can
> there be scenarios in which multiple application contexts are useful in
> the same process?

This is intended. In general, there is only one application context that
lives for the entire application lifetime. I don't think there is any
useful scenario where multiple application contexts are useful (and I
haven't encountered any), even while unit testing (where you can easily
craft multiple application contexts).