Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-8205966

[testbug] New Nestmates JDI test times out with Xcomp on sparc

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Fixed
    • Icon: P4 P4
    • 11
    • 11, 12
    • core-svc
    • 11
    • b22
    • sparc
    • solaris

        This has been seen in tier 4 testing - details in comments.

        Running the tests locally on Solaris x86 there are two interesting observations. Here's a normal test run:

        TEST: com/sun/jdi/RedefineNestmateAttr/TestNestmateAttr.java
          build: 0.001 seconds
          compile: 2.409 seconds
          compile: 2.317 seconds
          build: 0.0 seconds
          main: 7.226 seconds
          compile: 2.257 seconds
          build: 0.0 seconds
          main: 102.009 seconds
          compile: 2.232 seconds
          build: 0.0 seconds
          main: 7.322 seconds
          compile: 2.473 seconds
          build: 0.0 seconds
          main: 7.603 seconds
        TEST RESULT: Passed. Execution successful

        This test have 4 @run commands, hence the 4 "main" entries. But note that the 2nd one takes 102 seconds compared to the 7-8 seconds of all the others. This huge difference is currently inexplicable.

        Here's a Xcomp -XX:CompileThreshold=100 -XX:-TieredCompilation run:

        TEST: com/sun/jdi/RedefineNestmateAttr/TestNestmateAttr.java
          build: 0.001 seconds
          compile: 2.437 seconds
          compile: 2.336 seconds
          build: 0.001 seconds
          main: 133.23 seconds
          compile: 2.28 seconds
          build: 0.0 seconds
          main: 288.456 seconds
          compile: 2.285 seconds
          build: 0.0 seconds
          main: 133.609 seconds
          compile: 2.242 seconds
          build: 0.0 seconds
          main: 135.068 seconds
        TEST RESULT: Passed. Execution successful

        Interesting observation here is that in general it takes 18x long to run this way: 135s versus 8. But the really slow test only takes a bit over twice as long.

        While I need to investigate the execution anomaly, it seems evident that we need to ensure the timeout allows for a 18x slow down in general.

        section:main
        ----------messages:(7/289)----------
        command: main TestNestmateAttr HostA
        reason: User specified action: run main/othervm TestNestmateAttr HostA
        Mode: othervm [/othervm specified]
        Additional options from @modules: --add-modules java.compiler
        Timeout information:
        --- Timeout information end.
        elapsed time (seconds): 1586.496
        ----------configuration:(3/41)----------
        Boot Layer
          add modules: java.compiler

        ----------System.out:(1155/69563)----------
        vmOpts: '-XX:MaxRAMPercentage=3 -ea -esa -Xmx512m'
        javaOpts: '-Xcomp -XX:+CreateCoredumpOnCrash -ea -esa -XX:CompileThreshold=100 -XX:+UnlockExperimentalVMOptions -server -XX:-TieredCompilation'
        JVM version:11-internal
        JDI version: 11.0
        JVM description: Java Debug Interface (Reference Implementation) version 11.0
        Java Debug Wire Protocol (Reference Implementation) version 11.0
        JVM Debug Interface version 11.0
        JVM version 11-internal (Java HotSpot(TM) 64-Bit Server VM, compiled mode)
        Target: Testing original Host class from HostA
        TestNestmateAttr: Testing original Host class from HostA
        Trying bad retransform from Host/redef
        Timeout refired 1200 times
        ...

              dholmes David Holmes
              dholmes David Holmes
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: