From Alan.Bateman at Sun.COM  Tue Mar  2 05:06:17 2010
From: Alan.Bateman at Sun.COM (Alan Bateman)
Date: Tue, 02 Mar 2010 13:06:17 +0000
Subject: Need reviewer for 6931216: TEST_BUG:
	test/java/nio/file/WatchService/LotsOfEvents.java failed with NPE
Message-ID: <4B8D0D49.4020902@sun.com>

One of the WatchService tests failed with a NullPointerException in one 
of Kelly's test runs. It's a test issue and the fix is trivial. The 
reason for the NPE is that the test incorrectly assumes that the first 
batch of events will be read retrieved within 15s. The fix is to 
restructure the loops so that it checks the key at the beginning rather 
than the end.

Thanks,

-Alan.


diff -r 893034df4ec2 test/java/nio/file/WatchService/LotsOfEvents.java
--- a/test/java/nio/file/WatchService/LotsOfEvents.java Mon Mar 01 
18:00:47 2010 +0000
+++ b/test/java/nio/file/WatchService/LotsOfEvents.java Tue Mar 02 
12:58:07 2010 +0000
@@ -102,7 +102,7 @@ public class LotsOfEvents {
 
         int nread = 0;
         boolean gotOverflow = false;
-        do {
+        while (key != null) {
             List<WatchEvent<?>> events = key.pollEvents();
             for (WatchEvent<?> event: events) {
                 WatchEvent.Kind<?> kind = event.kind();
@@ -122,7 +122,7 @@ public class LotsOfEvents {
             if (!key.reset())
                 throw new RuntimeException("Key is no longer valid");
             key = watcher.poll(2, TimeUnit.SECONDS);
-        } while (key != null);
+        }
 
         // check that all expected events were received or there was an 
overflow
         if (nread < count && !gotOverflow)
@@ -168,7 +168,7 @@ public class LotsOfEvents {
                 // process events and ensure that we don't get repeated 
modify
                 // events for the same file.
                 WatchKey key = watcher.poll(15, TimeUnit.SECONDS);
-                do {
+                while (key != null) {
                     Set<Path> modified = new HashSet<Path>();
                     for (WatchEvent<?> event: key.pollEvents()) {
                         WatchEvent.Kind<?> kind = event.kind();
@@ -186,7 +186,7 @@ public class LotsOfEvents {
                     if (!key.reset())
                         throw new RuntimeException("Key is no longer 
valid");
                     key = watcher.poll(2, TimeUnit.SECONDS);
-                } while (key != null);
+                }
             }
 
         } finally {


From Christopher.Hegarty at Sun.COM  Tue Mar  2 08:20:03 2010
From: Christopher.Hegarty at Sun.COM (Christopher Hegarty - Sun Microsystems)
Date: Tue, 02 Mar 2010 16:20:03 +0000
Subject: Need reviewer for 6931216: TEST_BUG:
	test/java/nio/file/WatchService/LotsOfEvents.java failed with NPE
In-Reply-To: <4B8D0D49.4020902@sun.com>
References: <4B8D0D49.4020902@sun.com>
Message-ID: <4B8D3AB3.7030702@sun.com>

Looks fine.

BTW, I don't think the first part of the change is necessary since the 
enclosing method is never called with count <= 0, but it is certainly 
good practice and may prevent future failures if this tests evolves.

-Chris.

Alan Bateman wrote:
> One of the WatchService tests failed with a NullPointerException in one 
> of Kelly's test runs. It's a test issue and the fix is trivial. The 
> reason for the NPE is that the test incorrectly assumes that the first 
> batch of events will be read retrieved within 15s. The fix is to 
> restructure the loops so that it checks the key at the beginning rather 
> than the end.
> 
> Thanks,
> 
> -Alan.
> 
> 
> diff -r 893034df4ec2 test/java/nio/file/WatchService/LotsOfEvents.java
> --- a/test/java/nio/file/WatchService/LotsOfEvents.java Mon Mar 01 
> 18:00:47 2010 +0000
> +++ b/test/java/nio/file/WatchService/LotsOfEvents.java Tue Mar 02 
> 12:58:07 2010 +0000
> @@ -102,7 +102,7 @@ public class LotsOfEvents {
> 
>         int nread = 0;
>         boolean gotOverflow = false;
> -        do {
> +        while (key != null) {
>             List<WatchEvent<?>> events = key.pollEvents();
>             for (WatchEvent<?> event: events) {
>                 WatchEvent.Kind<?> kind = event.kind();
> @@ -122,7 +122,7 @@ public class LotsOfEvents {
>             if (!key.reset())
>                 throw new RuntimeException("Key is no longer valid");
>             key = watcher.poll(2, TimeUnit.SECONDS);
> -        } while (key != null);
> +        }
> 
>         // check that all expected events were received or there was an 
> overflow
>         if (nread < count && !gotOverflow)
> @@ -168,7 +168,7 @@ public class LotsOfEvents {
>                 // process events and ensure that we don't get repeated 
> modify
>                 // events for the same file.
>                 WatchKey key = watcher.poll(15, TimeUnit.SECONDS);
> -                do {
> +                while (key != null) {
>                     Set<Path> modified = new HashSet<Path>();
>                     for (WatchEvent<?> event: key.pollEvents()) {
>                         WatchEvent.Kind<?> kind = event.kind();
> @@ -186,7 +186,7 @@ public class LotsOfEvents {
>                     if (!key.reset())
>                         throw new RuntimeException("Key is no longer 
> valid");
>                     key = watcher.poll(2, TimeUnit.SECONDS);
> -                } while (key != null);
> +                }
>             }
> 
>         } finally {
> 

From Alan.Bateman at Sun.COM  Tue Mar  2 08:36:04 2010
From: Alan.Bateman at Sun.COM (Alan Bateman)
Date: Tue, 02 Mar 2010 16:36:04 +0000
Subject: Need reviewer for 6931216: TEST_BUG:
	test/java/nio/file/WatchService/LotsOfEvents.java failed with NPE
In-Reply-To: <4B8D3AB3.7030702@sun.com>
References: <4B8D0D49.4020902@sun.com> <4B8D3AB3.7030702@sun.com>
Message-ID: <4B8D3E74.5090701@sun.com>

Christopher Hegarty - Sun Microsystems wrote:
> Looks fine.
>
> BTW, I don't think the first part of the change is necessary since the 
> enclosing method is never called with count <= 0, but it is certainly 
> good practice and may prevent future failures if this tests evolves.  
That's right and I should have mentioned it - I only changed 
drainAndCheckOverflowEvents to keep it consistent. Thanks for reviewing.

-Alan.