On Thu, Jan 31, 2019 at 8:09 PM Mark Thomas <ma...@apache.org> wrote:
> On 31/01/2019 16:31, Mark Thomas wrote: > > On 31/01/2019 15:58, Rémy Maucherat wrote: > >> On Wed, Jan 30, 2019 at 8:37 PM Mark Thomas <ma...@apache.org> wrote: > >> > >>> A few failures on this one. I'm tracking progress here: > >>> > >>> https://cwiki.apache.org/confluence/display/TOMCAT/WebSocket+TCK > >>> > >>> > >>> Score so far is: > >>> 1 TCK bug > >>> 38 TBD > >>> > >>> A lot of the TBD appear to have the same root cause so hopefully that > >>> number will come down quickly. > >>> > >> > >> Running the TCK too now. > >> > >> The very last commit is broken (to be honest, I don't understand the > >> meaning of the spec language regarding these negdep tests failures, I > guess > >> there were fewer of these when I last looked at it and I didn't have the > >> problem). > > > > Whoops. I'll take a look. > > Fixed. Sorry for the noise. > > >> I had four issues then that appear to be fixed (three fixes and one test > >> gone, which was > >> com.sun.ts.tests.websocket.ee > .javax.websocket.handshakeresponse.getHeadersHasOriginTest). > >> com/sun/ts/tests/websocket/ee/javax/websocket/session/setTimeout1Test > >> doesn't pass for me at the moment, although you put in your wiki it is > >> fixed. > > > > I'm not 100% sure on that. It might be intermittent. Once I'd been > > through everything once, I was going to do a full run and see where > > things stand. > > Got to the bottom of that one (and 2 others). It was related to how > quickly we checked for session timeout. The default is every 10s which > isn't fast enough for those tests. I've added a system property to > reduce it to every second and that fixes them. > Hum, right, I think I remember that one now. It was a while ago. > > > Having looked at some of the TCK tests I am more in favour of making > > some of the changes I was previously opposed to. Again, I was planning > > on coming back to this after I'd done a new full run. > > Starting the full run now... > > Assuming it takes as long to run as it did last time (~2 hours) I'll > come back to this tomorrow. > So it's all good sorted out now, except the concurrency problem. The spec probably implied concurrency was "ok". Resolving this by waiting for the right state works for me (unsurprisingly), I have a first patch. Note: I cannot edit the confluence, what did I do wrong ? Rémy
