Pass all inputs to contrib/jenkins.sh as environment variables instead
of mixing with commandline arguments. We decided on this method instead
of I3cc01dbc57213b6d6cdc9b394a516510d045f1ae ("use jenkins.sh
--publish"), to make the contrib/jenkins.sh scripts less confusing.
Related: OS#3385
Change-Id: Ifc88af086368ce8eea1ea3a375ba7d5603a5d7e5
Variables for axes are already exported as environment variables.
Change 'GTP="$GTP" ./contrib/jenkins.sh' to './contrib/jenkins.sh' to
avoid confusion when introducing other variables.
Change-Id: I9238da3dd3c4461f82dc7c792f886177399be3dc
Do not build osmo-mgw twice anymore (once with "--enable-mgcp-transcoding" and
once with "--disable-mgcp-transcoding"). configure.ac does not know
these options.
Change-Id: I080cd331caec31c53a43090e6c19d23e31abdec6
At runtime, mount the Jenkins host's ~/.ssh folder inside the Docker
containers, from which manual PDFs will be published.
To my surprise, local testing showed that ssh is looking for the keys
in /home/build/.ssh, although the HOME variable inside the image is set
to /build, not /home/build.
Related: OS#3385
Change-Id: Icfff3b6a620de79d1bab7ff6a35309f828581c09
Add --publish to the contrib/jenkins.sh calls for each Osmocom project,
where PDF manuals should be published. As this parameter is only passed
to jenkins.sh from the master-builds job, not from the
gerrit-verification job, we can add logic to upload the generated PDFs
when getting triggered because of changes in master.
Related: OS#3385
Change-Id: I3cc01dbc57213b6d6cdc9b394a516510d045f1ae
cellmgr-ng is out of maintenance for a long time now (13 months) and some
of its features are going to be served by osmo-stp in the future.
cellmgr-ng is unable to build against recent libosmo-sccp.
Change-Id: I0869ec932ba3b3a9d3d8521b25f677e011334315
Do not publish PDFs for all projects anymore, as we are moving the
manuals to each project's git repository. They will be published from
there instead (see INSTALL.txt in osmo-gsm-manuals.git).
No automatism for building and publishing the documentation of all
projects will be implemented right now, as discussed here:
https://osmocom.org/issues/3385#note-7
Related: OS#3385
Change-Id: I705e20a4f94b308c250580aa1289ee9b2967beef
Kevin would like to upload simtrace firmware to ftp.osmocom.org from master
builds. To facilitate, add '--publish' arg, like in master-osmo-gsm-manuals (to
distinguish and not upload from gerrit verification jobs).
Change-Id: I93435aa6f39e9a072754011ae7696b10315f6092
This seems to resolve the issue where the jenkins job fails due to
missing ssh credentials.
Other jobs can also use this by setting the node property
Change-Id: I23cf2c98f02278a5deb9c4ae8d13438b76538287
Ticket: OS#3338
As rtl-sdr is gaining a ./contrib/jenkins.sh script, we can now also
use it just like our other projects, with a template-generated
"master-rtl-sdr" jenkins job, which can replace the old manually-created
"rtl-sdr" jenkins job.
Change-Id: Ia59f12f31d332bda09242fff1d3e215e8cadc915
When https://gerrit.osmocom.org/#/c/7725/ has been merged, there
is no need to limit this build to OsmocomApi.
Change-Id: Ic5fe39b643e7a15dbfd2efd55dce78dde30cf099
Avoid as many multiple triggers as still ensure all dependent projects are
rebuilt correctly. Keep the full trigger list as comment, and illustrate in a
comment at libosmocore how the chain is intended to work.
Change-Id: Iea2cf25b3872045778f11a985a1c417f37067cd9
proot crashes with current jenkins node setup, which means we cannot use
it to run ARM based axis.
proot bug is already reported upstream in:
https://github.com/proot-me/PRoot/issues/134
Related: OS#3061
Change-Id: I9bc48349c78f395b3842bc5caaf6e948fb4c299e
Introduce more precise labels to allow more flexibility when extending the jenkins setup.
The linux_amd64_debian8 or linux_amd64_debian9 is used across all build jobs which
make it hard to add new nodes which might only support one group of
jobs.
Change-Id: I0fa3d3f81ab01e2488fe07601740f42eb54b6d9c
There were a lot of downstream triggers that we were missing.
I manually reviewed the debian/control files and used that to update
the trigger lists for all jobs in master-builds.yml.
Change-Id: I12057c9bb389041ef3bcabd1c335a0fa8c358092
The sequential parameter was silently skipped because it was absent from
the project template. Fix this for both master- and gerrit- jobs.
Change-Id: I0bc28695f4f270bc7b1cc4bcd5d4d43ede6172f3
Get rid of job name comparison because it depends and exact build server
name and hence is highly fragile. Use dispatcher script the same way we
do in osmo-bts.
N. B: this requires I2955e866bce4f000a53369bd601a346c36c82468 in
libosmocore.
Change-Id: I76dfc11a05007ae5c6e0554fe8132695b67cccaa
Recently we had changes to osmo-ci, and I noticed that although some master
builds were broken by that, the builds were still showing success -- of twenty
days ago.
Run each master build at least once a day to indicate odd side effect failure
sooner.
Change-Id: I126de2bab3db22cb693b0fa665f6579de9238fdf
The vty and ctrl tests are enabled by default and are run on hard-coded
ports. This causes some builds to fail when run in parallel.
Change-Id: I23d5b75825a667e4f043d16a12b841cd8f01af5e
Debian 8 contains quite old qemu and proot packages which have some
issues running the chroot infrastructure set up in osmo-trx's
jenkins.sh.
Change-Id: I24665880fff5a5b918bb6ffaf1e7bb51ae860b0b
The 'refs/heads/origin/master' somehow caused that git polling would not
trigger builds. 'origin/master' was succesful in a manual test, so set all jobs
to that.
Change-Id: Id033d1bfce6cc9e20fbbf9be462842b9e44bde83
As far as I understand, the variable is populated using the name field.
In openggsn build it matches, but it doesn't in osmo-ggsn.
Change-Id: Ifb1a630b77a8c2f442e26dbef8e608882e8f9a71
Firstly, we don't have the downstream-ext plugin installed on our jenkins. We
want to use the 'trigger' publisher instead.
Secondly, since the jobs created here are called master-*, we also want to
trigger master-*.
master-libosmocore also triggers SIMtrace and xgoldmon, which aren't covered by
this jobs config, hence they don't get a master- prefix.
Change-Id: If9e8c4b02fce34fddceb4f07bf024210600f6270
The build-discarder section was silently not working. It needs to be nested
below a 'properties' node.
Also the names need to be dashed and not camelCased.
Change-Id: I9503200a8873e616f9195d4bb1d6163c464b305e
osmo-sgsn uses libosmo-ranap and hence should be triggered from osmo-iuh. This
naturally "includes" a trigger for libosmo-sigtran.
Change-Id: Ia356dc2a8d5120f9d6262bf8eb25c32fe71e76c9
libosmo-sccp, osmo-ggsn: The osmo-gsm-tester builds are downstream builds, yes,
but we configured that with the osmo-gsm-tester builds: instead of telling
libosmo-sccp to build osmo-gsm-tester_build-osmo-stp when done, we configure
the osmo-gsm-tester_osmo-stp to build after libosmo-sccp. So that the master
branch builds don't need to have any knowledge of osmo-gsm-tester.
osmo-msc build triggers should rather be post-build triggers of osmo-iuh and
osmo-mgw, like the others. Then we can also drop the pollscm here, and use the
pollscm that is common to all other builds.
Call these jobs 'master-*'. It more accurately says what they build, and also
we can install the jobs from this file next to the current, old ones, without
overwriting them and thus we'll have an easy rollback path. The new ones can
co-exist with the old ones until the new ones are verified to work, at which
point we can drop the old ones. Line 313:
IIUC the safest git branch is 'refs/remotes/origin/master'.
This is still untested!
Change-Id: If2ad9c90a0986d1304cd53383d3df5b375f23ac8