New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
7.4 backports: fix build in nspawn, bug fixes #7705
Conversation
martinpitt
commented
Sep 18, 2017
•
edited
edited
- Fix building in nspawn (otherwise the build fails on current images)
- Fix non-standard ssh ports (https://bugzilla.redhat.com/show_bug.cgi?id=1478828)
- Fix timers (https://bugzilla.redhat.com/show_bug.cgi?id=1478832)
Sigh, we just got rhel-7.4 branch tests working a few weeks ago, and now it seems the package build got broken (there's no build log in the results directory). I can actually reproduce that locally, and it appears to be a really strange tar error:
|
On second run I didn't hit the weird tar EOF error, but got some test failures instead; these sound like issue #7395, we aren't prepared yet for building in nspawn, at least not on the rhel-7.4 branch. Our rhel-7 images now have nspawn enabled in mock, so I take it these got fixed on master, but I'm not sure if that's worth backporting. |
This is necessary to be able to build Cockpit in an nspawn-enabled mock (which RHEL 7.4 now enables by default). Otherwise the unit tests fail with Message: couldn't lookup group: mockbuild: Success ERROR:src/ws/test-remotectlcertificate.c:127:test_combine_good: assertion failed (test->ret == 0): (1 == 0) as that group doesn't seem to exist in mock's nspawn. Cherry-picked from b186e4c (PR cockpit-project#7594)
and not into foo.service. This allows us to enable and disable the actual timer, and not the service that runs when the timer expires. Also, avoid partial lines in the unit files. Put timers into WantedBy=timers.target. This is the recommended thing to do. Also, WantedBy=default.target does not seem to work with certain versions of systemd. Cherry-picked from 5ba8277 (PR cockpit-project#6324) Bug https://bugzilla.redhat.com/show_bug.cgi?id=1478832
The connection properties' "port" field can be an integer, so stringify it (or any other field except "visible" which is a boolean and already special-cased) when building the GVariant wrapper in update_saved_machine(). Adjust TestMultiMachineAdd to reproduce/verify this case by putting the second external machine (m3) onto port 2222. Fixes cockpit-project#6291 Cherry-picked from 4f6c6ae (PR cockpit-project#6466) Closes cockpit-project#7705 Bug https://bugzilla.redhat.com/show_bug.cgi?id=1478828
dff84ca
to
bb4e1ea
Compare
I backported the test fix for nspawn, which seems harmless enough - in exchange cockpit can be built on the default RHEL 7.4 mock configuration, which is no small win. I also backported the timers fix, so this PR now contains both changes which we want to make to 7.4 base. |
So this is interesting, the rhel-7-4 tests fail because this PR is patching packages that are part of base on so not installed on that system. So it looks like we might need changes to the way that image is built when it is run on the rhel-7-4 branch. |
@petervo: Right; I was going to explain that once the rhel-7 tests pass (there is a flake). The rhel-7 tests prove that the fixes (and accompanying test changes) work once these fixes land in RHEL 7.4. While the rhel-7-4 test failures prove that the test cases actually detect/reproduce the bug and that these two features are really broken in current RHEL. My gut feeling is that for updates to RHEL base (i. e. this branch) the "rhel-7" tests are definitive and should stay green, while the rhel-7-4 tests should be definitive for PRs to RHEL extras (we don't currently have/need a branch for that, but we might again), and of course for master. I would like that better than crippling the rhel-7-4 tests in that branch, only to uncripple them in a couple of weeks when the fixes land. WDYT? |
That makes sense. I kind of would like something better than having red x for 7.4 meaning things are good here. But I don't have a better suggestion ATM. |
These two special cases covered known bugs in RHEL 7.4 base's cockpit 138 packages. These are more suitable as known issues (cockpit-project#7710 and cockpit-project#7711), particularly as they are being fixed in RHEL too (PR cockpit-project#7705). Closes cockpit-project#7712