Our CI pipeline fails 9 times out of 10 on the valgrind tests. The tests
seem to finish in either 35 min or exceed the 60 min timeout limit, with
nothing in between. To avoid this let's split into more groups so we can
a) run those more in parallel and b) are less likely to hit the
timeout when run slowly.
Analysis of recent logs shows the eraser button tests to be the worst
offender, taking 752s (due to the combinatorial explosion) alone. The
various tip and proximity tests together also take some time so let's
group those out.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1274>
vng is much faster than b2c for spinning up the jobs, so better use this
instead.
Bonus point: we don't need the start-in-systemd.sh stunt.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1125>
These take a long time and have a reasonable high chance of failure due
to the timing constraints. Let's split them up so they don't hog the
runners for that long and in case they fail, we only need to re-run a
short test.
Before: one test running approx 21 min, now 3 tests running approx 7 +
11 + 4 min.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>
Instead of extracting the suite name from the test's file name use the
current suite that is being parsed. This way we pave the way for
multiple suites in the same file.
This uses a global because otherwise we'd have to redo all the
litest_add() functions but it does the job here.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>
Actually libinput is one of the last users of harbor.fd.o, because it uses
an outdated version of vm2c.py. Use the new location of the project,
bump its release, and also bump the kernel version we test while at it.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Running "dnf install" during a job can lead to issues when the image is
old - package renames/replacements/etc. may require a dnf upgrade to get
those packages sorted first before our dnf install works.
This hasn't been a problem for us because we had weekly rebuilds of the
images scheduled and were usually on the latest package set but let's do
this properly anyway.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This gets rid of of the following error:
ld-elf.so.1: /lib/libc.so.7: version FBSD_1.7 required by /usr/local/lib/libpython3.9.so.1.0 not found
Too tired to debug what is really going on, so let's pretend the update
is the best way to fix this.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Looks like we are having clock skew issues on qemu, so given that
we just need qemu in the image, we can compile on the host (reliable)
and then only start the tests in qemu.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This will allow us to have the udevadm tool and systemd-udevd available
while running inside qemu
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This allows us to not have to create a specific image, and also
should be more reliable because we don't have to boot a full distribution
each time we just start our test suite.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
the combination of want_qemu and skip_container is not very straight
forward.
What we actually have, is that freebsd is only qemu based, so there is
no point in really having a `_QEMU` tag for it.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
F33 and F34 are both EOL. This also fixes the RPM build job to
automatically use the latest Fedora version and adds
wayland-protocols-devel which is now needed.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Update GTK to version 4 on Fedora, Arch and Alpine Linux.
Not updating Debian and FreeBSD because the package is not available yet
and Ubuntu because it is not available on 20.10.
Signed-off-by: José Expósito <jose.exposito89@gmail.com>
Replace our cross-compilation for FreeBSD with a proper template.
FreeBSD doesn't do normal containers so we need a bunch of if/else to
skip the container builds.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The group names are forced by check (they are called suite names there) but
for our test suite they provide very little benefit. Much easier to just
use the filename a test is in as group name.
This removes the pure substring match for --filter-group, it's now fnmatch
only. group names are short enough that the typing isn't an issue and we don't
want to run tests twice (e.g. 'pad' is also in 'touchpad').
This patch caused #574 until it got fixed in d838e3a3a4
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The string list is getting too confusing.
This gets rid of the separate packageset for qemu. That packageset only
differed by adding valgrind, we can just keep that in the same list.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
coverity doesn't work with gcc 10, it fails with "invalid GNU version
number: 201". F31 is about to be EOL but we can't use to F32 or later.
So let's switch to debian stable instead, that one will stick around for a bit
longer.
Debian packages are the same as the Ubuntu packages
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is not a package intended for, it's a job to fail when we accidentally
change the file list. An rpmbuild job like this was what detected
f15da0f108.
The spec file resembles the Fedora one but has BuildRequires removed (we rely
on the container for that).
The same task could be achieved by keeping a file list and comparing the
installed tree but since I had the rpm spec file already, let's use that for
now.
This requires meson 0.55 which hit F32 yesterday.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This makes the config file simpler, use a variable that is default false but
true where we need it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
