Re: RFR: JDK-8214756: SA should ignore archived java heap objects that are not in use

[Jini George]

Thank you very much, Jiangli! Yes, I did test 
open/test/hotspot/jtreg/serviceability/sa/TestHeapDumpForLargeArray.java 
after applying the changes from:


http://cr.openjdk.java.net/~redestad/8214712/jdk.00/

and it passed after this change.

Thanks,
Jini.

On 2/24/2019 9:38 AM, Jiangli Zhou wrote:

Hi Jini,

The change reflects the states of archived java objects and klass 
metadata properly at runtime and looks good to me.


You might have already done so, for additional testing try running 
open/test/hotspot/jtreg/serviceability/sa/TestHeapDumpForLargeArray.java 
with the planned changes for JDK-8214712 
.


Thanks,
Jiangli

On Sat, Feb 23, 2019 at 10:00 AM Jini George > wrote:


Requesting reviews for a very small change to fix:

https://bugs.openjdk.java.net/browse/JDK-8214756

Webrev: http://cr.openjdk.java.net/~jgeorge/8214756/webrev.00/index.html

The proposed change is to ignore objects whose associated classes are
unloaded while walking the heap to create a heapdump in SA.

Thanks,
Jini.

RFR 8207367: 10 vmTestbase/nsk/jdi tests timed out when running with jtreg

[Daniil Titov]

Please review the change that fixes timeout issues for the following 10 tests 
when running with jtreg and default timeout factor (1.0).

For the following 2 tests the event wait timeout was reduced and adjusted for 
test.timeout.factor.  Method receiveEvents(long,pattern) was fixed to ensure 
that it gracefully exits after the specified wait period elapsed:
  
-vmTestbase/nsk/jdi/ClassUnloadRequest/addClassExclusionFilter/exclfilter001.java
  -vmTestbase/nsk/jdi/ClassUnloadRequest/addClassFilter/filter001.java

For these 3 tests the event wait timeout was reduced and adjusted for 
test.timeout.factor:
  -vmTestbase/nsk/jdi/Event/_itself_/event001.java
  -vmTestbase/nsk/jdi/VirtualMachine/suspend/suspend001/TestDescription.java
  -vmTestbase/nsk/jdi/ThreadReference/suspend/suspend001.java

For next 2 tests the event wait timeout and the thread sleep time were reduced 
and adjusted for test.timeout.factor. Additional synchronization between the 
debugger and the debuggee was added to ensure the debugee process continues as 
soon as the test finishes the timeout related checks and advances to the next 
steps:
  - vmTestbase/nsk/jdi/EventQueue/remove_l/remove_l004/TestDescription.java 
  - vmTestbase/nsk/jdi/EventQueue/remove/remove004/TestDescription.java

Instead of just sleeping for 5 minutes while waiting for the debuggee test 
thread to complete  the tests now check whether the debuggee thread is alive in 
the loop. The total waiting timeout was adjusted for test.timeout.factor:
  -vmTestbase/nsk/jdi/VirtualMachine/dispose/dispose004/TestDescription.java 
  -vmTestbase/nsk/jdi/VirtualMachine/dispose/dispose003/TestDescription.java  
  -vmTestbase/nsk/jdi/VirtualMachine/dispose/dispose002/TestDescription.java

Testing.
The following VM options were used  in Mach5 jobs to  verify these changes:
1. No VM args
2. -XX:+UnlockExperimentalVMOptions -XX:+EnableJVMCI -XX:+TieredCompilation 
-XX:+UseJVMCICompiler -Djvmci.Compiler=grail
3. -Xcomp 

Also tier1, tier2 and tier3 Mach5 jobs succeeded.

To verify that tests succeed with test.timeout.factor set to 1.0 the following 
patch was used before running Mach5 jobs.

--- a/make/RunTests.gmk Thu Feb 21 15:17:42 2019 -0800
+++ b/make/RunTests.gmk Thu Feb 21 15:42:15 2019 -0800
@@ -826,6 +826,7 @@
   else
 JTREG_TIMEOUT_FACTOR ?= 4
   endif
+  JTREG_TIMEOUT_FACTOR = 1
   JTREG_VERBOSE ?= fail,error,summary
   JTREG_RETAIN ?= fail,error

Bug: https://bugs.openjdk.java.net/browse/JDK-8207367 
Webrev: http://cr.openjdk.java.net/~dtitov/8207367/webrev.01

Thanks!
--Daniil

RFR: JDK-8214756: SA should ignore archived java heap objects that are not in use

[Jini George]

Requesting reviews for a very small change to fix:

https://bugs.openjdk.java.net/browse/JDK-8214756

Webrev: http://cr.openjdk.java.net/~jgeorge/8214756/webrev.00/index.html

The proposed change is to ignore objects whose associated classes are 
unloaded while walking the heap to create a heapdump in SA.


Thanks,
Jini.