Alan,
The fix to remove the timeout value looks okay to me. While you're
there, I wonder if the test passes on Windows with Cygwin and looks like
this test might have a similar bug:
6981005: TEST BUG: java/lang/ClassLoader/TestCrossDelegate.sh
timeout on windows
that the cygwin path should be converted to windows path. The changeset
for 6981005 is:
http://hg.openjdk.java.net/jdk7/tl/jdk/rev/6586ab5b79f4
Mandy
On 6/20/11 6:21 PM, Alan Bateman wrote:
I need a reviewer for a small update to a test that has timed out
several times for me when running many tests concurrently. The problem
is the specified timeout (10s) is too low unless scaled with the
-timeoutFactor option. I don't see any any reason for this test to
override the default timeout so the following patch removes it.
-Alan
diff --git a/test/java/lang/ClassLoader/deadlock/TestOneWayDelegate.sh
b/test/java/lang/ClassLoader/deadlock/TestOneWayDelegate.sh
--- a/test/java/lang/ClassLoader/deadlock/TestOneWayDelegate.sh
+++ b/test/java/lang/ClassLoader/deadlock/TestOneWayDelegate.sh
@@ -25,7 +25,7 @@
# @summary (cl) ClassLoader.loadClass locks all instances in chain
# when delegating
#
-# @run shell/timeout=10 TestOneWayDelegate.sh
+# @run shell TestOneWayDelegate.sh
# if running by hand on windows, change TESTSRC and TESTCLASSES to "."
if [ "${TESTSRC}" = "" ] ; then
