TEST: java/util/logging/LoggerWeakRefLeak.sh

[Volker Simonis]

On Wed, Dec 18, 2013 at 5:58 PM, Alan Bateman <Alan.Bateman at oracle.com> wrote:
> On 18/12/2013 16:44, Volker Simonis wrote:
>>
>> Hi,
>>
>> the test 'java/util/logging/LoggerWeakRefLeak.sh' checks for the
>> availability of the '-histo:live' option in jmap which is only
>> available if  the SA is available.
>>
>> However I don't think that it should be an error if the option isn't
>> supported because of a missing SA implementation on a certain
>> platfrom.
>
> The tests have a lot of implementation dependencies and I'm sure you'll run
> into others too.
>

Yes, but the situation is starting to improve slowly:)

With the following two additional small changes:

diff -r 1baadb741a33 test/sun/tools/jinfo/Basic.sh
--- a/test/sun/tools/jinfo/Basic.sh     Tue Dec 17 19:01:21 2013 +0100
+++ b/test/sun/tools/jinfo/Basic.sh     Wed Dec 18 19:23:21 2013 +0100
@@ -45,7 +45,7 @@

 runSA=true

-if [ $isMacos = true ]; then
+if [ $isMacos = true -o $isAIX = true -o `uname -m` = ppc64 ]; then
     runSA=false
 fi

diff -r 1baadb741a33 test/tools/launcher/Settings.java
--- a/test/tools/launcher/Settings.java Tue Dec 17 19:01:21 2013 +0100
+++ b/test/tools/launcher/Settings.java Wed Dec 18 19:23:21 2013 +0100
@@ -73,16 +73,20 @@
     }

     static void runTestOptionDefault() throws IOException {
+        String stackSize = "256"; // in kb
+        if (getArch().equals("ppc64")) {
+            stackSize = "800";
+        }
         TestResult tr = null;
         tr = doExec(javaCmd, "-Xms64m", "-Xmx512m",
-                "-Xss256k", "-XshowSettings", "-jar",
testJar.getAbsolutePath());
+                "-Xss" + stackSize + "k", "-XshowSettings", "-jar",
testJar.getAbsolutePath());
         containsAllOptions(tr);
         if (!tr.isOK()) {
             System.out.println(tr.status);
             throw new RuntimeException("test fails");
         }
         tr = doExec(javaCmd, "-Xms65536k", "-Xmx712m",
-                "-Xss256000", "-XshowSettings", "-jar",
testJar.getAbsolutePath());
+                "-Xss" + stackSize + "000", "-XshowSettings", "-jar",
testJar.getAbsolutePath());
         containsAllOptions(tr);
         if (!tr.isOK()) {
             System.out.println(tr.status);

I get the same pass rate on Linux/PPC64 like on Linux/x86_64 (4672
passed / 1 Failed / 0 Errors / 900 Not run). The one that fails also
fails in jdk8b120 so it's nothing I'm going to worry about in the next
days or so:)

Admittedly, Linux/PPC64 was the easy part and for AIX I still have
about 100 failures and 50 errors, but I hope to cut that drastically
down with a few changes until the end of the week.

>
>>
>> I suggest the following fix:
>>
>> --- a/test/java/util/logging/LoggerWeakRefLeak.sh       Tue Dec 17
>> 19:01:21 2013 +0100
>> +++ b/test/java/util/logging/LoggerWeakRefLeak.sh       Wed Dec 18
>> 17:40:46 2013 +0100
>> @@ -83,9 +83,9 @@
>>       fi
>>
>>       if [ "$status" != 0 ]; then
>> -        echo "ERROR: 'jmap $jmap_option' is not supported so this test"
>> -        echo "ERROR: cannot work reliably. Aborting!"
>> -        exit 2
>> +        echo "WARNING: 'jmap $jmap_option' is not supported on this
>> platform"
>> +        echo "WARNING: so this test cannot work reliably. Aborting!"
>> +        exit 0
>>       fi
>>   fi
>>
>> I want to put this fix into the patch for "8028537: PPC64: Updated
>> jdk/test scripts to understand the AIX os and environment" so I don't
>> think we need a special BugID for it.
>>
>> What do you think?
>>
> I'd prefer to see this one (and AnonLoggerWeakRefLeak) changed to not use
> jmap to detect leaks but that is beyond the scope of what you are doing.
> What you have is fine and shouldn't have any impact on other platforms.
>

Thanks!

> -Alan