Add visibility of proguard rules to README or wiki please

[isuPatches]

Are there examples of proguard rules that will work with minifyEnabled set to true? I'm currently getting:

 java.lang.NoSuchMethodError: No virtual method doOnSubscribe(Lrx/functions/Action0;)Lrx/Observable

[ZacSweers]

+1

The unsafe package does some reflective lookups of fields that get lost in obfuscation, and the resulting NoSuchFieldException gets thrown up.

[julienbanse]

+1
Actually, I have NoSuchFieldException: producerIndex with the release 1.0.14 (not happens in 1.0.13)

My current proguard settings for RxJava is

-keep class rx.schedulers.Schedulers {
    public static <methods>;
}
-keep class rx.schedulers.ImmediateScheduler {
    public <methods>;
}
-keep class rx.schedulers.TestScheduler {
    public <methods>;
}
-keep class rx.schedulers.Schedulers {
    public static ** test();
}

[ZacSweers]

I'm able to avoid it by keeping everything in rx.internal.util.unsafe, but would be nice to avoid the reflection entirely if possible.

[julienbanse]

Yes, this is what I did to patch this issue.

[gokhanbarisaker]

Full stack trace for NoSuchFieldException, that @julienbanse mentioned.

08-14 14:58:28.619    7362-7429/com.markavip.app E/AndroidRuntime﹕ FATAL EXCEPTION: pool-1-thread-1
    Process: com.markavip.app, PID: 7362
    java.lang.IllegalStateException: Fatal Exception thrown on Scheduler.Worker thread.
            at rx.internal.schedulers.ScheduledAction.run(SourceFile:62)
            at rx.schedulers.ExecutorScheduler$ExecutorSchedulerWorker.run(SourceFile:98)
            at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1112)
            at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:587)
            at java.lang.Thread.run(Thread.java:818)
     Caused by: java.lang.InternalError
            at rx.internal.util.unsafe.UnsafeAccess.a(SourceFile:103)
            at rx.internal.util.unsafe.SpscArrayQueueProducerFields.<clinit>(SourceFile:39)
            at rx.internal.operators.OperatorObserveOn$ObserveOnSubscriber.<init>(SourceFile:102)
            at rx.internal.operators.OperatorObserveOn.a(SourceFile:64)
            at rx.internal.operators.OperatorObserveOn.a(SourceFile:44)
            at rx.Observable$2.a(SourceFile:158)
            at rx.Observable$2.a(SourceFile:154)
            at rx.Observable.a(SourceFile:7710)
            at rx.internal.operators.OperatorSubscribeOn$1$1.a(SourceFile:62)
            at rx.internal.schedulers.ScheduledAction.run(SourceFile:55)
            at rx.schedulers.ExecutorScheduler$ExecutorSchedulerWorker.run(SourceFile:98)
            at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1112)
            at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:587)
            at java.lang.Thread.run(Thread.java:818)
     Caused by: java.lang.NoSuchFieldException: producerIndex
            at java.lang.Class.getDeclaredField(Class.java:890)
            at rx.internal.util.unsafe.UnsafeAccess.a(SourceFile:100)
            at rx.internal.util.unsafe.SpscArrayQueueProducerFields.<clinit>(SourceFile:39)
            at rx.internal.operators.OperatorObserveOn$ObserveOnSubscriber.<init>(SourceFile:102)
            at rx.internal.operators.OperatorObserveOn.a(SourceFile:64)
            at rx.internal.operators.OperatorObserveOn.a(SourceFile:44)
            at rx.Observable$2.a(SourceFile:158)
            at rx.Observable$2.a(SourceFile:154)
            at rx.Observable.a(SourceFile:7710)
            at rx.internal.operators.OperatorSubscribeOn$1$1.a(SourceFile:62)
            at rx.internal.schedulers.ScheduledAction.run(SourceFile:55)
            at rx.schedulers.ExecutorScheduler$ExecutorSchedulerWorker.run(SourceFile:98)
            at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1112)
            at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:587)
            at java.lang.Thread.run(Thread.java:818)

[desseim]

Thing is, this used to JustWork(tm) because the reflective call was done directly on the target class (e.g. SpscArrayQueueProducerFields.class.getDeclaredField("producerIndex")) and proguard could figure out that this field needed to be kept (see the "reflection" chapter of the proguard documentation).
Starting with 1.0.14 the reflective call is now indirected through a utility class (UnsafeAccess#addressOf) and I guess proguard don't detect the field to keep anymore.

For now, we can either keep everything in package unsafe or selectively keep reflected field with

-keepclassmembers class rx.internal.util.unsafe.*ArrayQueue*Field* {
    long producerIndex;
    long consumerIndex;
}
-keepclassmembers class rx.internal.util.unsafe.BaseLinkedQueueProducerNodeRef {
    long producerNode;
    long consumerNode;
}

while waiting for them to be annotated appropriately in the source.

[JakeWharton]

FWIW for Android at least this is going to be handled by ReactiveX/RxAndroid#219 automatically for consumers. This obviously does not solve if you are ProGuarding for another target (or simply not using RxAndroid on Android).

[arturonaredo]

@julienbanse + @desseim responses together worked for me.