On Thu, Jun 08, 2017 at 10:16:11AM -0700, Ben Pfaff wrote: > On Wed, Jun 07, 2017 at 05:56:36PM -0700, Joe Stringer wrote: > > On 7 June 2017 at 17:36, Joe Stringer <[email protected]> wrote: > > > On 7 June 2017 at 13:58, Flavio Leitner <[email protected]> wrote: > > >> This change the testsuite macro to release the resources > > >> configured by ovs-vswitchd when exiting as it used to be. > > >> > > >> Fixes: 0f28164be02ac ("netdev-linux: make tap devices persistent") > > >> Fixes: fe13ccdca6a22 ("vswitchd: Add --cleanup option to the 'appctl > > >> exit' command") > > >> > > >> Reported-by: Eric Garver <[email protected]> > > >> Signed-off-by: Flavio Leitner <[email protected]> > > > > > > Thanks for the fix, applied to master. > > > > While this fixes successful test runs with OVS_APP_EXIT_AND_WAIT(), if > > anything fails in the middle of the test then OVS isn't going to > > perform cleanup, which will leave devices hanging around, which will > > cause subsequent tests to fail. > > > > I think we need to amend the _OVS_VSWITCHD_START macro as well, for > > the following line: > > on_exit "kill `cat ovs-vswitchd.pid`" > > > > This should use appctl to do a tidy cleanup as well. > > Be careful about that. There needs to be a fallback to kill vswitchd > even if it doesn't exit gracefully. The test framework waits for all > processes to exit, so the testsuite will ultimately hang if a process > never does.
I sent this patch to fix it which I think addresses both concerns: https://mail.openvswitch.org/pipermail/ovs-dev/2017-June/333693.html -- Flavio _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
