Add the variable to projects that don't use docker at all or overwrite
docker_run instead of using the global variable that has
OSMO_GSM_MANUALS_DIR defined.
Related: OS#4912
Change-Id: I65102655b4b3e907a0b0de8026f8f9af76281f90
Don't mount $HOME/bin to /build_bin, and don't overwrite PATH to include
/build_bin anymore. This isn't needed because the scripts are installed
to /usr/local/bin inside the docker images now.
Depends: docker-playground I0820f13cd740f7d4a8999d279cef6ae293a67d81
Related: OS#4911
Change-Id: I45bd7ea56849dab839a98cfd52806f7cd288094e
Save time by only cloning the repository, and not running autoreconf
-fi, ./configure, make and make install. Especially the tests during
make took up significant time, that slowed every project depending on
osmo-gsm-manuals down while being built through master-builds or
gerrit-verifications jobs. Set OSMO_GSM_MANUALS_DIR to the clone
location.
This is an interim measure, I'll submit more patches soon that remove
the 'osmo-build-dep.sh osmo-gsm-manuals' calls from all projects.
Change-Id: I5238cf3f93ded97ed2b44d27868123a646e122dc
Related: OS#4912
Gerrit verification jobs sometimes hang (usually during the VTY tests).
At some point we should just time out, default is 30 minutes
Change-Id: I6fc6ee976f2d88efdab1582a418d1adcd384c850
This will rune once per week and make sure that registry.osmocom.org
has a fpga-build container that the buld slaves can use.
Change-Id: Ie700d7c6e3c5e794edd81c3acc4b689e05d6864d
We use centos8 instead of centos in all docker-playground.git setups and
the infrastructure there expects that kind of naming.
Related: OS#4888
Change-Id: Idfbb2c4fc1ca10741406c8ab8930dabe8ce632ee
It's really annoying that OBS uses Debian_8.0 and Debian_9.0 (for 8.x
and 9.x) and then Debian_10 for 10.x. Let's work around this by some
conditional vars_file trickery, hopefully once and for all.
Change-Id: I45c4752774f44332b3f9aaf875f49043c03ec5ca
Install TTCN3 before installing build-deps, as build-deps
requires libulfius, which is only provided by network:osmocom:latest
feed, which in turn is only activated as part of the TTCN3 / TITAN
installation.
Change-Id: I66f1d5f7479a6d028abdac0cc6f4000faa78b135
podman has one particularly interesting feature which docker
rejected years ago: The ability to run a container in a pre-existing
namespace. This is quite useful for some of our test setups,
particularly those that involve hdlc net-devices
Change-Id: I4b14599f39c07ec83fa4222f9d040d443caae3e6
As of osmo-gsm-tester.git Change-Id
I938602ee890712fda82fd3f812d8edb1bcd05e08, python3-watchdog is required.
Change-Id: I35026e82b255cd4e06b4211c4829cd12be59976e
we cannot even create new slaves as lxc-create doesn't support debian
jessie anymore. Let's focus on Debian 9 + Debian 10
Change-Id: I2fddd460f7d1db755cc9b732b2be80494d26e091
The docker company official builds are only present for amd64, but not
for other architectures. Let's use the Debian docker.io package in such
situations.
Change-Id: Id85137fc0eb7fa5e2b31aa928c78bfdaadc281ac
We actually do the verification inside a (debian 9) docker container anyway,
so there's no need to run those on the old debian8 slaves.
Change-Id: I2168bd2f0e6221518ece111e29dc604c33eaca6a
If that package is not installed, adding the GPG key fails with SSL
certificate verification, at least on a debian10 i386 lxc:
TASK [add docker gpg key to apt keyring] *************************************************************
fatal: [gtp0-deb10build32-ansible]: FAILED! => {"changed": false, "msg": "Failed to download key at https://download.docker.com/linux/debian/gpg: Request failed: <urlopen error [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1056)>"}
Change-Id: I58259a99258c9c7471e1fcf03a3a7b62fa9c6a48
Since Idba84164b90e3d183a20b5eb69cbfe15745e447c, we started to
generate separate PDFs for different osmo-bts-variants, so
let's enable them all. Otherwise only osmo-bts-{virtual,sysmo}
PDFs are generated and uploaded to the server.
Change-Id: I517be2d475f57026dab7f001d2298cf1a1aa2c7c
Related: SYS#4937
This will make the Dockerfiles from docker-playground.git
use registry.osmocom.org rather than the public docker.io for base
distribution images like debian:stretch or the like.
This will avoid any strange timeout/EOF problems while doing
docker build --pull in our test jobs.
Change-Id: If4275cf3c8a5b70a67d742950a35ae8ead59a52e
Closes: OS#4839
Before this patch, only the debian+master jobs were spread around
several hours, while debian+latest and centos+nightly were all triggered
at midnight.
Let's instead run all flavours of teh same test (debian+nightly/latest,
centos) around same time but not concurrently by marking blocking one
each other so that they are run one after the other. Then, each group is
separated by a given time spreading the load over the night + morning.
This also means we no longer start tests at 00:00, which means docker
images will use software from the day after, since the nightly
repositories are built around 01:00.
Related: SYS#5121
Change-Id: Ib1f374974444cdd5d829c4e57153137ecd9149f1
Some unit tests in those repositories started failing due to a fix in
libosmocore logging system where extra coloring tags were printed when not
needed.
Unit tests in current master of those repositories are fixed to work with
both old and new versions of libosmocore since they got coloring disabled
in their output, and new patch releases have been done containing the
fix, which means last releases already work with both libosmocore
version. However, older releases are expected to fail when built against
libosmocore master.
Change-Id: I03ca926b903a4dcc9967ab5fe455d715cdb9ed45
The osmo-python-tests will be used to test vty commands.
The old python2 package are only required to test old images because
the current master is only supporting python3.
Installing both doesn't conflict because the python site-packages are
different for the two version.
Change-Id: Ica279e83b763389ababb526336ebd95d45beee7a
This reverts commit e17a4d66d0, i.e.
we are back to building the latest tag, now that v20.07.1 has been
released, which actually builds again.
However, the package name in debian/changelog has not been updated to
reflect that version change, resulting in v20.07.1 being packaged
in a package called v20.07.0.
Change-Id: I01b77f03924a0b303103fb737dfee15b9c4b0c9c
overlayfs2 should be the default anyway. Enable it on all hosts,
since it was enabled on older hosts.
Change-Id: I20728d6017204c3978e23376baa89de6e91fed1e
See https://github.com/myriadrf/LimeSuite/issues/313
This also reverts the previous commits that removed the work-around for
building LimeSuite on Debian10, as that one is still required for
v20.01.0 (and not for v20.07.0).
Change-Id: Ib70418f0b8a4c6aafa3098b6fa3e240f89112b59
As of yesterday (upstream commit fdfe3d340506a7de4f41d74296e1f32a906700f2),
LimeSuite upstream finally builds without patches on Debian 10
Change-Id: Ibbed53f57ac750f39ea20ac73c28c4773c244a6f