/network-monitor/create-in-thread fails in (LXC) containers on glib-2-56
Like this:
GTest: random seed: R02S8509dd6d1e533434790e1d9241555716
GTest: run: /network-monitor/create-in-thread
(MSG: Bug Reference: https://bugzilla.gnome.org/show_bug.cgi?id=793727)
GTest: subprocess: /network-monitor/create-in-thread
(MSG: child process (/network-monitor/create-in-thread [20670]) exit status: 1 (error))
(MSG: child process (/network-monitor/create-in-thread [20670]) stdout: "")
(MSG: child process (/network-monitor/create-in-thread [20670]) stderr: "\n(./gio/tests/.libs/network-monitor-race:20670): GLib-GIO-WARNING **: 18:26:12.430: Error on netlink socket: Error receiving message: Bad file descriptor\n")
**
GLib-GIO:ERROR:/home/laney/dev/gnome/glib/gio/tests/network-monitor-race.c:78:test_network_monitor: child process (/network-monitor/create-in-thread [20670]) failed unexpectedly
(ERROR: GLib-GIO:ERROR:/home/laney/dev/gnome/glib/gio/tests/network-monitor-race.c:78:test_network_monitor: child process (/network-monitor/create-in-thread [20670]) failed unexpectedly)
[1] 20669 abort (core dumped) G_MESSAGES_DEBUG=all ./gio/tests/.libs/network-monitor-race --verbose
this does not happen on master, so I bisected to find out where it was fixed. Quite interestingly it's e0b120cc. Cherry picking that to glib-2-56
was enough to get the test to pass. I'm only not proposing that in an MR because I don't have much of a clue as to why that would work. I would suspect I messed something up, but I tried both reverting that commit on master and cherry-picking it to glib-2-56
and the results were as expected.
We discovered this in Ubuntu's armhf CI, which runs in such containers.
Edited by Iain Lane