On 12/12/12 9:47 AM, Alan Bateman wrote:
On 12/12/2012 16:36, Daniel D. Daugherty wrote:
For this item:
> test/java/util/logging/LoggingDeadlock4.java
> Test case was simplified to avoid AWT class loading.
Negative test
> result was tested on early JDK7 build.
if I remember correctly, the whole point of that test was to
check for a logging deadlock relative to AWT's usage of logging.
If you avoid loading AWT classes, doesn't that make the test
rather useless?
Dan
java.awt.Container:
private static final PlatformLogger log =
PlatformLogger.getLogger("java.awt.Container");
private static final PlatformLogger eventLog =
PlatformLogger.getLogger("java.awt.event.Container");
and the updated test is just using PlatformLogger directly,
I thought the deadlock had to do with locks grabbed on the way
to getting into the underlying PlatformLogger, but my memory is
hazy and I don't have the cycles to research this.
I hope it demonstrates the same issue with a JDK that doesn't have the
fix.
That would be the right way to see if the test still "works".
(BTW: Just as background, with compact profiles coming then we need to
beat our tests into shape so that the tests for the APIs supported in
each profile can be run. Alexey is addressing some of the low-hanging
fruit, clearly it won't be possible to remove the dependency from all
tests. Also care is required to ensure that the test continues to test
what it was created to test. Expect efforts like this ^10 once modules
comes).
Thanks for the background.
Dan
