After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 795234 - Fix Jenkins CI build failures for network tests
Fix Jenkins CI build failures for network tests
Status: RESOLVED FIXED
Product: glib
Classification: Platform
Component: gio
2.56.x
Other Linux
: Normal normal
: ---
Assigned To: gtkdev
gtkdev
Depends on:
Blocks:
 
 
Reported: 2018-04-13 15:09 UTC by Philip Withnall
Modified: 2018-04-13 17:51 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
tests: Add more debug output to the network-address test (1.51 KB, patch)
2018-04-13 15:14 UTC, Philip Withnall
committed Details | Review
tests: Skip GSocket tests if setting up a server fails (7.77 KB, patch)
2018-04-13 15:15 UTC, Philip Withnall
committed Details | Review
tests: Drop a slightly suspect GResolver test (1.67 KB, patch)
2018-04-13 16:43 UTC, Philip Withnall
committed Details | Review

Description Philip Withnall 2018-04-13 15:09:35 UTC
Currently some of the networking tests (network-address.c and socket.c) are partially failing when running on GitLab CI on jenkins.gnome.org (provided by AWS) but not when run on progress.gnome.org (hosted by us). This is probably due to differences in how they have their networking configured — AWS networking is locked down a bit more.

I’ve got some patches which might fix this by skipping tests where appropriate, but they need testing.
Comment 1 Philip Withnall 2018-04-13 15:14:53 UTC
Created attachment 370895 [details] [review]
tests: Add more debug output to the network-address test

Trying to diagnose why some of the resolver tests fail on AWS.

Signed-off-by: Philip Withnall <withnall@endlessm.com>
Comment 2 Philip Withnall 2018-04-13 15:15:00 UTC
Created attachment 370896 [details] [review]
tests: Skip GSocket tests if setting up a server fails

There are various reasons why setting up a server might fail; it
reliably fails on AWS with IPv6 addresses (are we binding to the right
address?). Since we’re trying to test GSocket as a client, skip tests
where that happens.

Signed-off-by: Philip Withnall <withnall@endlessm.com>
Comment 3 Emmanuele Bassi (:ebassi) 2018-04-13 15:21:37 UTC
Review of attachment 370895 [details] [review]:

Okay
Comment 4 Emmanuele Bassi (:ebassi) 2018-04-13 15:22:48 UTC
Review of attachment 370896 [details] [review]:

Makes sense.
Comment 5 Philip Withnall 2018-04-13 15:36:20 UTC
Both pushed; let's see what the CI does now.

Attachment 370895 [details] pushed as d990f86 - tests: Add more debug output to the network-address test
Attachment 370896 [details] pushed as 6751424 - tests: Skip GSocket tests if setting up a server fails
Comment 6 Philip Withnall 2018-04-13 15:56:36 UTC
https://gitlab.gnome.org/GNOME/glib/-/jobs/23660

That fixed the socket test. Just the network-address test to fix now. Unfortunately those g_test_message() calls aren’t resulting in any additional output to the test logs.

I’ll for the GLib repository and push some test commits in a throwaway branch to debug further.
Comment 7 Philip Withnall 2018-04-13 16:12:21 UTC
The network-address failure is this, for address `fe80::42%1`:

Error resolving ?fe80::42%1?: Address family for hostname not supported (g-resolver-error-quark, 2)
Comment 8 Philip Withnall 2018-04-13 16:43:09 UTC
Created attachment 370898 [details] [review]
tests: Drop a slightly suspect GResolver test

This test will only work on machines which have IPv6 enabled and have a
local IPv6 interface with ID 1. On machines which don’t (such as AWS
servers, which we run CI tests on), the GResolver tests will fail with
G_RESOLVER_ERROR_INVALID. We can’t differentiate this kind of failure
(where we’d want to skip the test) from an actual failure (where we’d
want to fail the test), so the only other option is to drop this
particular test vector. I don’t think it’s a significant loss.

This is the last fix needed to get our CI tests working reliably on
jenkins.gnome.org.

Signed-off-by: Philip Withnall <withnall@endlessm.com>
Comment 9 Emmanuele Bassi (:ebassi) 2018-04-13 17:45:07 UTC
Review of attachment 370898 [details] [review]:

:+1:
Comment 10 Philip Withnall 2018-04-13 17:51:13 UTC
Pushed to master, thanks. This should fix the CI failures; I will reopen this bug if we get the same failure show up again.

Attachment 370898 [details] pushed as ad50fdb - tests: Drop a slightly suspect GResolver test