Hi all,
A note on this v5 CI report -- apologies in advance for the thread
noise after several revisions.
The Build-and-Test workflow on series_503840 reports a single failure
on the "linux gcc test -300" matrix entry, at:
194: ovn.at:13500 - vlan traffic for external network with
distributed router gateway port
I don't believe this is caused by the patch. Some context, in case
it helps a maintainer triage:
* Same v5 run, same -300 test range: "linux gcc test -ljemalloc
-300" and "linux clang test sanitizers -300" both pass.
Identical source tree, only the compiler / options differ.
All 301-600, all 601- (unstable), and the ovn-kubernetes
workflow all pass.
https://github.com/ovsrobot/ovn/actions/runs/25712639237
* v4 (same northd.c, only diff to v5 is "check" -> "check_uuid"
on one Load_Balancer_Health_Check creation in the *new* unit
test at tests/ovn-northd.at:1755) passed "linux gcc test -300"
on its run. v5's only fail-mode change relative to v4 is in
the new unit tests #687/#688, which now pass.
* Test 194 does not configure any Load_Balancer health_check, so
build_lb_health_check_response_lflows() in northd/northd.c is
never reached for that test -- the per-backend short-circuit
at northd.c:8891 ('if (!backend_nb->health_check ||
!backend_nb->svc_mon_lrp) continue;') skips before any lflow
this patch modifies is emitted.
* Local 'make check' on a clean Ubuntu 24.04 build passes all
227 ovn-northd.at unit tests, including the new #687/#688,
and test 194/195 (run in isolation) also passes.
Test 194 is a 4-HV pcap-based packet-flow scenario with RA/ARP/L3
timing dependencies, so an intermittent failure under CI container
parallelism seems plausible. Happy to dig further if anything
looks off; otherwise a retrigger should clear it.
Thanks,
JayGue Lee
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
