<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <br>
    <div class="moz-cite-prefix">On 11/12/2013 17:55, Jonathan Gibbons
      wrote:<br>
    </div>
    <blockquote cite="mid:52A8A72C.7050300@oracle.com" type="cite">Balchandra,
      <br>
      <br>
      The important part of that of interest here is the part about
      "calls jtreg commands
      <br>
      (one each for jdk, langtools and hotspot) with those recommended
      <br>
      options".
      <br>
      <br>
      Is there a way you could post here the segment of the script that
      does that?
      <br>
      <br>
    </blockquote>
    Here it is<br>
    <br>
    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
    <table border="0" height="300" width="749">
      <tbody>
        <tr>
          <td>2.1 Running tests in jdk/test
          </td>
        </tr>
        <tr>
          <td>
            <div style="margin-left: 40px;"> $ jtreg -dir:{openjdk
              source top directory}/jdk/test -verbose:summary
              -exclude:{openjdk source top
              directory}/jdk/test/ProblemList.txt -conc:auto -a
              -ignore:quiet -timeoutFactor:5 -othervm -testjdk:{location
              of the test jdk} `cat <a
                href="http://download.java.net/jdk8/testresults/docs/dir.list">dir.list</a>`
              <br>
              <br>
            </div>
          </td>
        </tr>
        <tr>
          <td>
            2.2 Running tests in langtools/test
          </td>
        </tr>
        <tr>
          <td>
            <div style="margin-left: 40px;"> $ jtreg -dir:{openjdk
              source top directory}/langtools/test -verbose:summary
              -conc:auto -a -ignore:quiet -timeoutFactor:5 -agentvm
              -testjdk:{location of the test jdk} com tools <br>
              <br>
            </div>
          </td>
        </tr>
        <tr>
          <td>
            2.3 Running tests in hotspot/test
          </td>
        </tr>
        <tr>
          <td>
            <div style="margin-left: 40px;"> $ jtreg -dir:{openjdk
              source top directory}/hotspot/test -verbose:summary
              -conc:auto -a -ignore:quiet -timeoutFactor:5 -agentvm
              -testjdk:{location of the test jdk} compiler gc runtime
              sanity serviceability
            </div>
          </td>
        </tr>
      </tbody>
    </table>
    <br>
    The instructions is at
    <a class="moz-txt-link-freetext" href="http://download.java.net/jdk8/testresults/docs/howtoruntests.html">http://download.java.net/jdk8/testresults/docs/howtoruntests.html</a><br>
    and is linked from
    <a class="moz-txt-link-freetext" href="http://www.java.net/download/jdk8/testresults/testresults.html">http://www.java.net/download/jdk8/testresults/testresults.html</a><br>
    <br>
    The caveat in this approach is a human error where one (me) forget
    to update the instruction<br>
    or forget to remove any additional flags used in the script
    temporarily - a mismatch<br>
    of the results could occur. <br>
    <br>
    From the next build (b120), I am going to update "2.1 Running tests
    in jdk/test" to<br>
    <br>
    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
    $ jtreg -dir:{openjdk source top directory}/jdk/test
    -verbose:summary -exclude:{openjdk source top
    directory}/jdk/test/ProblemList.txt -conc:auto -a -ignore:quiet
    -timeoutFactor:5 -agentvm -testjdk:{location of the test jdk}
    :jdk_core :jdk_svc :jdk_beans :jdk_imageio :jdk_sound :jdk_sctp
    javax/accessibility com/sun/java/swing javax/print sun/pisces
    com/sun/awt <br>
    <br>
    <br>
    <blockquote cite="mid:52A8A72C.7050300@oracle.com" type="cite">Alternatively,
      that segment of the script could be a candidate for a target in
      <br>
      one or more test/Makefile files.
      <br>
    </blockquote>
    <br>
    This is good idea, but my experience with the 'make' is that if one
    target critically fail, all<br>
    subsequent targets will not run. I thought it is a restriction of
    'make'.<br>
    <br>
    <br>
    Thanks<br>
    Balchandra<br>
    <br>
    <blockquote cite="mid:52A8A72C.7050300@oracle.com" type="cite">
      <br>
      -- Jon
      <br>
      <br>
      <br>
      On 12/11/2013 08:58 AM, Balchandra Vaidya wrote:
      <br>
      <blockquote type="cite">
        <br>
        Hi Volker,
        <br>
        <br>
        The custom script I use just wrap the instruction posted (except
        <br>
        building jtreg). It just clones the openjdk repository, calls
        jtreg commands
        <br>
        (one each for jdk, langtools and hotspot) with those recommended
        <br>
        options, runs diffs,  runs nashorn tests, and archives results.
        <br>
        <br>
        Such convenient script may not fit under openjdk repo, but think
        of
        <br>
        adding it in code-tools project in the future.
        <br>
        <br>
        <br>
        Thanks
        <br>
        Balchandra
        <br>
        <br>
        On 11/12/2013 13:45, Volker Simonis wrote:
        <br>
        <blockquote type="cite">Hi Balchandra,
          <br>
          <br>
          what do you think about checking in a little script or
          Makefile target
          <br>
          which produces the test results you report for the ea builds?
          <br>
          <br>
          This would be of great help to produce comparable test
          results.
          <br>
          <br>
          Regards,
          <br>
          Volker
          <br>
          <br>
          On Tue, Dec 10, 2013 at 5:39 PM, Balchandra Vaidya
          <br>
          <a class="moz-txt-link-rfc2396E" href="mailto:balchandra.vaidya@oracle.com"><balchandra.vaidya@oracle.com></a> wrote:
          <br>
          <blockquote type="cite">On 12/10/13 03:25 PM, Alan Bateman
            wrote:
            <br>
            <blockquote type="cite">On 10/12/2013 11:20, Balchandra
              Vaidya wrote:
              <br>
              <blockquote type="cite">:
                <br>
                <br>
                2) Testcase do not run with -agentvm option. It might be
                an
                <br>
                     issue in the testcase (or how the testcase was
                run). I will look into
                <br>
                it.
                <br>
                <br>
                     I missed this issue because I had changed my
                scripts to use -othervm
                <br>
                     sometime back for debugging/analyzing some testcase
                <br>
                     failure but forgot to change it back. Thank you for
                pointing it out.
                <br>
              </blockquote>
              For Volker's benefit, othervm is where jtreg spins up a
              new VM for each
              <br>
              test, agentvm is where jtreg re-uses the VM if possible.
              There are tests
              <br>
              that don't or can't clean up and those tests need to
              ensure that their @run
              <br>
              tests have the /othervm option so that the tests runs in
              its own VM.
              <br>
              Alternatively, the TEST.ROOT file has a key that lists
              directories where all
              <br>
              the tests in those directory trees must run in their own
              VM.
              <br>
              <br>
              Balchandra - one thing that be useful is to look at the
              client tests to
              <br>
              see what can run in agentvm and what can't. For the
              jdk_core and jdk_svc
              <br>
              tests then they can be run in either (although we're still
              picking off
              <br>
              issues as they found in agentvm mode). I'm not aware of
              any work done yet on
              <br>
              getting the tests in the jdk_desktop test group to run in
              agentvm mode. It
              <br>
              may be that TEST.ROOT needs to be updated to see a few
              additional top-level
              <br>
              directories so that all the tests in these areas
              temporarily run in othervm
              <br>
              mode.
              <br>
            </blockquote>
            <br>
            In the subset of the tests I have been running, it appears
            only one test [1]
            <br>
            is failing with agentvm option. I will try to run other
            tests in client
            <br>
            groups [Note: I will update the instruction to use jtreg
            groups with b119
            <br>
            results posting] and see what tests can be run in agentvm
            mode, but main
            <br>
            issue
            <br>
            in those remaining client tests are some tests may not
            consistently produce
            <br>
            same results - could stabilizing those tests be a goal for
            jdk9 repo ?
            <br>
            <br>
            <br>
            Thanks
            <br>
            Balchandra
            <br>
            <br>
            [1] com/sun/java/swing/plaf/windows/8016551/bug8016551.java
            <br>
            <br>
            <blockquote type="cite">-Alan
              <br>
            </blockquote>
            <br>
          </blockquote>
        </blockquote>
        <br>
      </blockquote>
      <br>
    </blockquote>
    <br>
  </body>
</html>