This is a regression test for issue 201968.
First, download the sample project provided here. Open this project and sync it with the Gradle files.
Run the “ExampleInstrumentedT...” configuration with an AVD, and verify that all 10 tests pass. There should be a green “OK” icon before each entry in the test console, which should depict a tree of the following structure:
The order of the useAppContext entries does not matter, but all 10 must be present, passing, and nested under the ExampleInstrumentedTest entry.
Finally, Debug “ExampleInstrumentedT...”, and verify that it also produces the same output.
If you are QA, you may ignore this section.
Because the original issue was a race condition, it may not occur every time the above manual test is performed. It is possible to reduce the chance of a false negative by slowing the execution of one of the threads, making it more likely to lose the race if the race condition still exists.
To do this, open the Android Studio source, and navigate to the stdout method of ProcessHandlerConsolePrinter. This class is located in the following file: tools/adt/idea/android/src/com/android/tools/idea/run/ProcessHandlerConsolePrinter.java
At line 57, where stdout calls print, create a breakpoint. Set the breakpoint to ‘Evaluate and log’ the following:
int baz = 0;
for (int i = 0; i < 10000; i++)
baz += 10;
return baz;
Make sure the breakpoint is not set to suspend any threads. Increasing the 10000 will make the thread even slower, lowering the probability of a false negative but increasing the test run time.
Debug Android Studio with the breakpoint enabled. Then perform the normal reproduction steps in this debugged instance.