The jps command is used on Windows by the VM/NSK run_sajdi.sh script to translate from Cygwin pids to Windows pids for the purpose of attaching to the target VM. Sometimes this command fails for unknown reasons when the target VM is known to be up and running. The failure mode looks like: [2010-05-20T05:18:31.05] Debuggee class started [2010-05-20T05:18:31.05] Debuggee class ready for debugging [2010-05-20T05:18:31.05] DEBUGGEE_READY_FOR_DEBUGGING [2010-05-20T05:18:31.05] script> Debuggee VM got ready for debugging [2010-05-20T05:18:31.05] script> # ERROR: jps could not find Windows debuggee VM pid [2010-05-20T05:18:31.05] script> # ERROR: DEBUGGEE_CLASS='nsk.sajdi.ReferenceType.fieldByName.fldbyname001a' [2010-05-20T05:18:31.05] script> # ERROR: <begin_jps_and_ps_output> [2010-05-20T05:18:31.05] 5480 [2010-05-20T05:18:31.05] 29772 sun.tools.jps.Jps [2010-05-20T05:18:31.05] 3956 -- process information unavailable [2010-05-20T05:18:31.05] 10736 Tonga [2010-05-20T05:18:31.05] 13376 com.sun.javaws.Main [2010-05-20T05:18:31.05] 17172 10700 28952 3956 con 1004 23:16:58 /cygdrive/c/work/local/baseline_jdks/1.7.0_b93/windows-i586/bin/java [2010-05-20T05:18:35.14] script> # ERROR: <end_jps_and_ps_output> [2010-05-20T05:18:35.14] script> Kill debuggee VM process for pid: 17172 The java process is present and has written the DEBUGGEE_READY_FOR_DEBUGGING token before waiting on a monitor for the next stage of the test. I don't know why jps fails to get the process information and I'm filing this bug to make note of the fact that this issue needs investigation. No specific tests are named in this bug intentionally because this issue shows up during VM/NSK sajdi test setup and thus can affect any test.
|