[8u] RFR 8254683: [TEST_BUG] jdk/test/sun/tools/jconsole/WorkerDeadlockTest.java fails

[hedongbo]

Hi, we spent some time investigating this issue.

This looks like a jtreg bug. The fix is ​​as follows:
diff -r ff9fdaab4344 src/share/classes/com/sun/javatest/regtest/config/Locations.java
--- a/src/share/classes/com/sun/javatest/regtest/config/Locations.java  Tue Sep 15 15:23:52 2020 -0700
+++ b/src/share/classes/com/sun/javatest/regtest/config/Locations.java  Sat Oct 17 02:38:18 2020 +0800
@@ -253,7 +253,7 @@
                 String name = lib.substring(2, end);
                 File dir = null;
                 if (name.equals("java.home")) {
-                    dir = testJDK.getFile();
+                    dir = testJDK.getAbsoluteFile();
                 } else if (name.equals("jtreg.home")) {
                     dir = jtpath.asList().get(0).getParentFile().getParentFile();
                 }
Do I need to report a bug to jtreg?( Because the working directory of jtreg is JTwork/scratch, the relative path specified on the command line is wrong)

Currently, we can run this test case through a script, the new webrev is as follows:
webrev: http://cr.openjdk.java.net/~dongbohe/8254683/webrev.01/
We extracted the common code from ResourceCheckTest.sh to common.sh, then modified the ResourceCheckTest.sh, and added WorkerDeadlockTest.sh for this case.

Is this ok?

Thanks,
dongbohe
-----Original Message-----
From: Severin Gehwolf [mailto:sgehwolf at redhat.com] 
Sent: Wednesday, October 14, 2020 4:56 PM
To: hedongbo <hedongbo at huawei.com>; jdk8u-dev <jdk8u-dev at openjdk.java.net>
Subject: Re: [8u] RFR 8254683: [TEST_BUG] jdk/test/sun/tools/jconsole/WorkerDeadlockTest.java fails

On Wed, 2020-10-14 at 01:30 +0000, hedongbo wrote:
> Hi,
> 
> Please review this 8u test-only changes. The problem was introduced 
> with JDK- 8237836
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8254683
> webrev: http://cr.openjdk.java.net/~dongbohe/8254683/webrev.00/
> 
> Testing:  Worked correctly after patch.

One issue I've noticed is that if a relative path is being passed to jtreg via -jdk, like '-jdk:../build/linux...' the test still fails with the patch. Absolute path works. Would there be a way to make it work more reliably?

Thanks,
Severin