<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    EAP uses *-proposed branches for exactly these purposes.<br>
    <br>
    Once we get green lights from all CI stations, the respective main
    branch is fast forwarded to the proposed state.<br>
    <br>
    Carlo<br>
    <br>
    <div class="moz-cite-prefix">On 07-12-17 08:29, Scott Marlow wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAJtdNN6UQFFeC7x0atUFc8tU=_xYXgQYpatFhDBywiJbYt4jAw@mail.gmail.com">
      <div dir="ltr">Excellent suggestions Rostislav!  There are some
        trade offs due to the amount of time needed for each test run. 
        Whatever the implementation, it would be good to have a
        prioritized way of running the tests that accomplish a-c.
        <div><br>
        </div>
        <div>Scott  
          <div><br>
          </div>
          <div>
            <div>
              <div class="gmail_extra">
                <div class="gmail_quote">On Wed, Dec 6, 2017 at 7:31 AM,
                  Rostislav Svoboda <span dir="ltr">&lt;<a
                      href="mailto:rsvoboda@redhat.com" target="_blank"
                      moz-do-not-send="true">rsvoboda@redhat.com</a>&gt;</span>
                  wrote:<br>
                  <blockquote class="gmail_quote" style="margin:0 0 0
                    .8ex;border-left:1px #ccc solid;padding-left:1ex">I
                    see 3 main use-cases for TCK runs outside WFLY
                    master<br>
                    <br>
                     a) pre-checks before final merging to WFLY master -
                    master-ignore way discussed here<br>
                     b) checks on big feature(s) development branches<br>
                          something like ladybird or RFE development
                    across multiple components<br>
                          selected TCK modules can be executed, depends
                    on scope of changes<br>
                     c) regular component checks on component master<br>
                          early / regression checks on component level
                    that they do not regress<br>
                          TCK modules related to the component and
                    layered components should be executed<br>
                          I know only few components which run related
                    TCK modules with their master<br>
                    <br>
                    With proposed changes for quick WF delivery I
                    believe use-cases b) and c) will need to be
                    addressed.<br>
                    <br>
                    Use-cases b) and c) could be done on component level
                    or via central pipeline.<br>
                      component level - prepare automated way to run TCK
                    with custom build - e.g. bash script, docker image<br>
                      central pipeline - set of jobs can be prepared and
                    linked via Jenkins pipeline so the end user (or curl
                    request) provides just URL of custom WFLY build zip
                    + list of modules to be executed<br>
                    <span class="HOEnZb"><font color="#888888"><br>
                        Rostislav<br>
                      </font></span>
                    <div class="HOEnZb">
                      <div class="h5"><br>
                        ----- Original Message -----<br>
                        &gt; On Tue, Dec 5, 2017 at 4:37 AM, Scott
                        Marlow &lt; <a href="mailto:smarlow@redhat.com"
                          moz-do-not-send="true">smarlow@redhat.com</a>
                        &gt; wrote:<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; It would be great if we could have a branch
                        that includes all of the commits<br>
                        &gt; that we are considering to merge at a
                        particular time of day, such that we<br>
                        &gt; would run the TCK against that branch, only
                        once a day.<br>
                        &gt;<br>
                        &gt; Can this be done that often? I had in my
                        mind that if we did one of these it<br>
                        &gt; would amount to stealing one of the regular
                        runs, but perhaps that's not the<br>
                        &gt; case.<br>
                        &gt;<br>
                        &gt; Now, I don't think we'd want to do these
                        anywhere near that often, but it's<br>
                        &gt; good to know what the limits would be. For
                        example, I could imagine doing 10<br>
                        &gt; of these over the course of a WF release,
                        but by luck or whatever 3 of them<br>
                        &gt; come in the same week.<br>
                        &gt;<br>
                        &gt; +1 to using a branch. We have a branch like
                        that, master-ignore, that we use<br>
                        &gt; for batching up PRs to test as a group
                        before merging. I wouldn't want to<br>
                        &gt; use master-ignore for this, but a
                        differently named branch run the same way<br>
                        &gt; sounds good.<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; If one of the changes cause a TCK failure,
                        none of them get merged<br>
                        &gt; (investigation follows that to determine
                        which change caused the<br>
                        &gt; failure(s)), if the test succeeds, we can
                        then merge that batch of changes<br>
                        &gt; into WildFly master.<br>
                        &gt;<br>
                        &gt; We likely would want to avoid running the
                        testing, on days when we haven't<br>
                        &gt; merged any changes to the WF testing
                        branching.<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; Can the TCK be set up to run based on a
                        check for a change in the sha of the<br>
                        &gt; head of a branch? So every day at a fixed
                        time it checks the branch, and<br>
                        &gt; does nothing if there is no change. If we
                        want a run, we force push the<br>
                        &gt; branch before that time. We have CI jobs
                        that check master-ignore that way,<br>
                        &gt; except they poll regularly, not just once a
                        day. That works for those as<br>
                        &gt; they aren't so resource intensive that we
                        worry a lot about limiting how<br>
                        &gt; often they run.<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; Would that approach help how we merge PRs
                        on master?<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; I think it could be helpful earlier in the
                        release cycle before merging big<br>
                        &gt; changes, and then perhaps late in the
                        release cycle if we're worried about<br>
                        &gt; possible regressions.<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; Scott<br>
                        &gt;<br>
                        &gt; On 12/04/2017 09:33 PM, Stuart Douglas
                        wrote:<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; On Tue, Dec 5, 2017 at 3:40 AM, Brian
                        Stansberry &lt;<br>
                        &gt; <a
                          href="mailto:brian.stansberry@redhat.com"
                          moz-do-not-send="true">brian.stansberry@redhat.com</a>
                        &lt;mailto: <a
                          href="mailto:brian.stansberry@redhat.com"
                          moz-do-not-send="true">brian.stansberry@redhat.com</a>
                        &gt;&gt; wrote:<br>
                        &gt;<br>
                        &gt; Great. :)<br>
                        &gt;<br>
                        &gt; One thing I think we need to do is figure
                        out how to get custom TCK<br>
                        &gt; runs for PR branches. The TCK is a big part
                        of our test coverage,<br>
                        &gt; and one way to not "use master as a test
                        bed" is to get a check of a<br>
                        &gt; branch on the TCK before we merge it.<br>
                        &gt;<br>
                        &gt; I know we've gotten TCK runs of ad-hoc
                        branches before, so by<br>
                        &gt; "figure out" I mean work out how to make
                        that not overly painful,<br>
                        &gt; come to some sort of consensus on when it's
                        worthwhile, etc.<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; I think if we were going to do this it
                        should probably be something reviewers<br>
                        &gt; can ask for on specific PR. The TCK uses a
                        *lot* more resources than a<br>
                        &gt; standard CI run, so we need to make sure we
                        limit it to cases where it is<br>
                        &gt; required.<br>
                        &gt;<br>
                        &gt; Stuart<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; On Fri, Dec 1, 2017 at 10:04 AM, Alessio
                        Soldano<br>
                        &gt; &lt; <a href="mailto:asoldano@redhat.com"
                          moz-do-not-send="true">asoldano@redhat.com</a>
                        &lt;mailto: <a
                          href="mailto:asoldano@redhat.com"
                          moz-do-not-send="true">asoldano@redhat.com</a>
                        &gt;&gt; wrote:<br>
                        &gt;<br>
                        &gt; There you go... PR updated to consume the
                        same api jar now<br>
                        &gt; released as final.<br>
                        &gt;<br>
                        &gt; Cheers<br>
                        &gt; Alessio<br>
                        &gt;<br>
                        &gt; On Fri, Dec 1, 2017 at 3:30 PM, David Lloyd<br>
                        &gt; &lt; <a
                          href="mailto:david.lloyd@redhat.com"
                          moz-do-not-send="true">david.lloyd@redhat.com</a>
                        &lt;mailto: <a
                          href="mailto:david.lloyd@redhat.com"
                          moz-do-not-send="true">david.lloyd@redhat.com</a>
                        &gt;&gt; wrote:<br>
                        &gt;<br>
                        &gt; On Thu, Nov 30, 2017 at 5:50 PM, Alessio
                        Soldano<br>
                        &gt; &lt; <a href="mailto:asoldano@redhat.com"
                          moz-do-not-send="true">asoldano@redhat.com</a>
                        &lt;mailto: <a
                          href="mailto:asoldano@redhat.com"
                          moz-do-not-send="true">asoldano@redhat.com</a>
                        &gt;&gt; wrote:<br>
                        &gt; &gt; As suggested by Brian, I'd like to
                        draw attention to the discussion on<br>
                        &gt; &gt; <a
                          href="https://github.com/wildfly/wildfly/pull/10604"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://github.com/wildfly/<wbr>wildfly/pull/10604</a><br>
                        &gt; &lt; <a
                          href="https://github.com/wildfly/wildfly/pull/10604"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://github.com/wildfly/<wbr>wildfly/pull/10604</a>
                        &gt; .<br>
                        &gt; &gt; The PR is an upgrade of the
                        webservices stack, including JBossWS, Apache<br>
                        &gt; &gt; CXF, JAXB-RI and JAXB API. In
                        particular, the JAXB upgrade is for EE8 and<br>
                        &gt; &gt; better JDK 9 compatibility.<br>
                        &gt; &gt; Now, due to the upgrade of the JAXB
                        API spec jar, the PR is essentially<br>
                        &gt; &gt; stalled since 20 days; the new spec is
                        released as an alpha (as it's been<br>
                        &gt; &gt; tested within JBossWS only) and that
                        does not satisfy a rule that requires<br>
                        &gt; &gt; any artifact being pulled to be Final.<br>
                        &gt; &gt; We're talking about a spec jar, we
                        could simply re-tag that as Final,<br>
                        &gt; &gt; chances are we won't need changes any
                        time soon there anyway, but as Tomaz<br>
                        &gt; &gt; pointed out, in principle that would
                        be dishonest.<br>
                        &gt;<br>
                        &gt; My opinion is that you should go ahead and
                        make a .Final<br>
                        &gt; tag. In the<br>
                        &gt; (unlikely?) event that the spec has to be
                        modified for some<br>
                        &gt; reason, I<br>
                        &gt; think you could make a 1.0.1.Final tag and
                        call it a "bug fix".<br>
                        &gt;<br>
                        &gt; The alternative is to simply wait. I don't
                        think there is<br>
                        &gt; any middle position.<br>
                        &gt;<br>
                        &gt; &gt; While I see the point in requiring
                        that only sufficiently stable upgrades<br>
                        &gt; &gt; are applied to the codebase, I'm
                        wondering whether, maybe, we're going a<br>
                        &gt; &gt; bit<br>
                        &gt; &gt; too far with the rules. Brian wrote on
                        this topic: "how to determine that<br>
                        &gt; &gt; something is good enough to go in
                        without using master as a test bed" ?<br>
                        &gt;<br>
                        &gt; I don't think we are; I agree with the
                        policy as it stands. If you<br>
                        &gt; look at it in terms of being able to
                        release at any time,<br>
                        &gt; then it<br>
                        &gt; follows that everything _must_ be stable.<br>
                        &gt;<br>
                        &gt; --<br>
                        &gt; - DML<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; ______________________________<wbr>_________________<br>
                        &gt; wildfly-dev mailing list<br>
                        &gt; <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a>
                        &lt;mailto: <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a>
                        &gt;<br>
                        &gt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a><br>
                        &gt; &lt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; -- Brian Stansberry<br>
                        &gt; Manager, Senior Principal Software Engineer<br>
                        &gt; Red Hat<br>
                        &gt;<br>
                        &gt; ______________________________<wbr>_________________<br>
                        &gt; wildfly-dev mailing list<br>
                        &gt; <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a>
                        &lt;mailto: <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a>
                        &gt;<br>
                        &gt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a><br>
                        &gt; &lt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; ______________________________<wbr>_________________<br>
                        &gt; wildfly-dev mailing list<br>
                        &gt; <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a><br>
                        &gt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a><br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt;<br>
                        &gt; --<br>
                        &gt; Brian Stansberry<br>
                        &gt; Manager, Senior Principal Software Engineer<br>
                        &gt; Red Hat<br>
                        &gt;<br>
                        &gt; ______________________________<wbr>_________________<br>
                        &gt; wildfly-dev mailing list<br>
                        &gt; <a
                          href="mailto:wildfly-dev@lists.jboss.org"
                          moz-do-not-send="true">wildfly-dev@lists.jboss.org</a><br>
                        &gt; <a
                          href="https://lists.jboss.org/mailman/listinfo/wildfly-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.jboss.org/<wbr>mailman/listinfo/wildfly-dev</a><br>
                      </div>
                    </div>
                  </blockquote>
                </div>
                <br>
              </div>
            </div>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
wildfly-dev mailing list
<a class="moz-txt-link-abbreviated" href="mailto:wildfly-dev@lists.jboss.org">wildfly-dev@lists.jboss.org</a>
<a class="moz-txt-link-freetext" href="https://lists.jboss.org/mailman/listinfo/wildfly-dev">https://lists.jboss.org/mailman/listinfo/wildfly-dev</a></pre>
    </blockquote>
    <br>
  </body>
</html>