Github user NicoK commented on a diff in the pull request:
https://github.com/apache/flink/pull/5105#discussion_r156397312
--- Diff:
flink-runtime/src/test/java/org/apache/flink/runtime/io/network/buffer/LocalBufferPoolDestroyTest.java
---
@@ -104,11 +104,10 @@ public void testDestroyWhileBlockingRequest() throws
Exception {
* @return Flag indicating whether the Thread is in a blocking buffer
* request or not
*/
- private boolean isInBlockingBufferRequest(StackTraceElement[]
stackTrace) {
+ public static boolean isInBlockingBufferRequest(StackTraceElement[]
stackTrace) {
if (stackTrace.length >= 3) {
return stackTrace[0].getMethodName().equals("wait") &&
-
stackTrace[1].getMethodName().equals("requestBuffer") &&
-
stackTrace[2].getMethodName().equals("requestBufferBlocking");
+
stackTrace[1].getClassName().equals(LocalBufferPool.class.getName());
--- End diff --
I'm a bit unsure whether this is a good change since the test is verifying
a specific scenario but the new check broadens it to "waiting anywhere in
`LocalBufferPool`" rather than "waiting on a buffer request". We do, however,
know that it will be waiting for the buffer at some point but if we ever waited
on something else, we may test another scenario than the intended one.
---
