On Thu, 15 Dec 2016, John Spray wrote: > On 15 Dec 2016 16:12, "Jason Dillaman" <jdillama@xxxxxxxxxx> wrote: > The issue is that this is an example of a workunit pulling > support > files. You are suggesting that the workunit task now support > listing > dependent scripts that the teuthology worker task should also > download > and copy over to the test host? If so, it would definitely need > to be > done in such a way that we can still run these workunits locally > (outside of teuthology). > > > Ah, I see what you mean. > > In this particular case the ultimate solution is perhaps to skip the > indirection of having a workunit that downloads and runs a test script, and > instead have a teuthology task that just executes the test explicitly (skip > the bash part) > > However I haven't checked exactly how many tests might need reworking that > way so if its a lot then we might need a CEPH_REPO var into work units to go > with the CEPH_REF This category of tests are fixed by https://github.com/ceph/ceph/pull/12511. The workunit task already does a git checkout. The fix is to run the workunits directly out of the checkout directory, and to fix the tests to use relative paths. CEPH_BASE is also set to the base dir of the checkout, in case that's useful. I'm not sure we need CEPH_REPO since the checkout is already there.. If there are other tests that are still broken, let me know! sage > > John > > > On Thu, Dec 15, 2016 at 11:04 AM, John Spray <jspray@xxxxxxxxxx> > wrote: > > On Thu, Dec 15, 2016 at 3:54 PM, Jason Dillaman > <jdillama@xxxxxxxxxx> wrote: > >> Correct me if I am wrong, but since we are just installing > RPMs, I > >> wouldn't expect the full source tree to be available on the > test host. > >> The test scripts would still need to pull the correct scripts > from > >> git. Preferably, it would be nice if we can get a new > environment > >> variable that includes the full git url with the branch/sha1 > already > >> embedded. That would allow testing any script changes on > teuthology > >> w/o forcing a push to ceph-ci and creating an unnecessary > build. > > > > I'm thinking that rather than teuthology (the process > executing > > workunit.py) calling a git command on the test host, it would > take its > > local copy of the script (i.e. relative to workunit.py > > ../../../qa/workunits etc) and send that directly. > > > > I think we can already avoid pushing test-only changes to > ceph-ci, > > because there are separate options for ceph vs. suite branches > (even > > though both point at ceph.git forks). So when I want to test > a > > qa-suite change, I would use --suite-repo <github /jcsp/ path> > > --suite-branch wip-whatever. That would get the workunits > from > > wip-whatever. > > > > John > > > >> > >> On Thu, Dec 15, 2016 at 10:46 AM, John Spray > <jspray@xxxxxxxxxx> wrote: > >>> On Thu, Dec 15, 2016 at 3:00 PM, Mykola Golub > <trociny@xxxxxxxxxxx> wrote: > >>>> On Thu, Dec 15, 2016 at 04:51:44PM +0200, Mikolaj Golub > wrote: > >>>>> Hi, Sage, > >>>>> > >>>>> It looks like we still need to push testing branches to > ceph.git for > >>>>> things like below? > >>>>> > >>>>> qa/workunits/rbd/test_librbd_python.sh: > >>>>> > >>>>> wget -O test_rbd.py"https://git.ceph.com/?p=ceph.git;a=blob_plain;hb=$CEPH_REF;f=src/test/pybi > nd/test_rbd.py" || \ > >>>>> wget -O test_rbd.py"https://git.ceph.com/?p=ceph.git;a=blob_plain;hb=ref/heads/$CEPH_REF;f=src > /test/pybind/test_rbd.py" > >>>>> > >>>>> Do we have some env variable (CEPH_REPO?) we could use > simirlaly to > >>>>> CEPH_REF here? > >>>> > >>>> Or rather may be now we can avoid such hacks at all? > >>> > >>> Exactly - as you notice bits like that, let's convert them > to just > >>> read their files locally instead of trying to fetch them > from git. I > >>> guess the workunit task will be the major one here, if > nobody already > >>> adjusted it. > >>> > >>> John > >>> > >>>> -- > >>>> Mykola Golub > >>>> -- > >>>> To unsubscribe from this list: send the line "unsubscribe > ceph-devel" in > >>>> the body of a message to majordomo@xxxxxxxxxxxxxxx > >>>> More majordomo info at > http://vger.kernel.org/majordomo-info.html > >>> _______________________________________________ > >>> Sepia mailing list > >>> Sepia@xxxxxxxxxxxxxx > >>> http://lists.ceph.com/listinfo.cgi/sepia-ceph.com > >> > >> > >> > >> -- > >> Jason > > -- > > To unsubscribe from this list: send the line "unsubscribe > ceph-devel" in > > the body of a message to majordomo@xxxxxxxxxxxxxxx > > More majordomo info at > http://vger.kernel.org/majordomo-info.html > > > > -- > Jason > > > >
