RFR: 8052403 java/util/logging/CheckZombieLockTest.java fails with NoSuchFileException
Please find below a fix for 8052403: java/util/logging/CheckZombieLockTest.java fails with NoSuchFileException https://bugs.openjdk.java.net/browse/JDK-8052403 The issue here is that CheckZombieLockTest and CheckLockLocationTest use (create and delete) the same writable-dir under /tmp. When jtreg is run with a high concurrency factor this sometimes results in one of the tests deleting the directory that the other is running. I could verify this by triggering the same failure with: jtreg -verbose:all -jdk $MYJDK -conc:12 \ jdk/test/java/util/logging/Check* The fix simply arranges for CheckZombieLockTest to use a different directory than that used by CheckLockLocationTest (and also improves the diagnostic by arranging for setUp() to fail fast if the test lock file cannot be created). Webrev: http://cr.openjdk.java.net/~dfuchs/webrev_8052403/webrev.00/ best regards, -- daniel
Looks good. Nit: adding a space between catch and ( would be good. Mandy On 8/13/2014 5:43 AM, Daniel Fuchs wrote:
Please find below a fix for
8052403: java/util/logging/CheckZombieLockTest.java fails with NoSuchFileException https://bugs.openjdk.java.net/browse/JDK-8052403
The issue here is that CheckZombieLockTest and CheckLockLocationTest use (create and delete) the same writable-dir under /tmp.
When jtreg is run with a high concurrency factor this sometimes results in one of the tests deleting the directory that the other is running. I could verify this by triggering the same failure with:
jtreg -verbose:all -jdk $MYJDK -conc:12 \ jdk/test/java/util/logging/Check*
The fix simply arranges for CheckZombieLockTest to use a different directory than that used by CheckLockLocationTest (and also improves the diagnostic by arranging for setUp() to fail fast if the test lock file cannot be created).
Webrev: http://cr.openjdk.java.net/~dfuchs/webrev_8052403/webrev.00/
best regards,
-- daniel
On 8/15/14 10:32 PM, Mandy Chung wrote:
Looks good. Nit: adding a space between catch and ( would be good.
Thanks Mandy. I took that opportunity to also fix the spaces before and after '+' in print statements. For the record the new webrev I'm going to push is here: http://cr.openjdk.java.net/~dfuchs/webrev_8052403/webrev.01/ (no need to re-review it's mostly noise due to the added white spaces now). best regards, -- daniel
Mandy
On 8/13/2014 5:43 AM, Daniel Fuchs wrote:
Please find below a fix for
8052403: java/util/logging/CheckZombieLockTest.java fails with NoSuchFileException https://bugs.openjdk.java.net/browse/JDK-8052403
The issue here is that CheckZombieLockTest and CheckLockLocationTest use (create and delete) the same writable-dir under /tmp.
When jtreg is run with a high concurrency factor this sometimes results in one of the tests deleting the directory that the other is running. I could verify this by triggering the same failure with:
jtreg -verbose:all -jdk $MYJDK -conc:12 \ jdk/test/java/util/logging/Check*
The fix simply arranges for CheckZombieLockTest to use a different directory than that used by CheckLockLocationTest (and also improves the diagnostic by arranging for setUp() to fail fast if the test lock file cannot be created).
Webrev: http://cr.openjdk.java.net/~dfuchs/webrev_8052403/webrev.00/
best regards,
-- daniel
participants (2)
-
Daniel Fuchs
-
Mandy Chung