Update azure-pipelines.yml to run builds for release branches following
the new naming convention. Release branches now follow the format
`xlnx/release/linux-v<kernel_version>-<release>`.
Update prepare_artifacts.sh scripts to upload artifacts under
`linux-v<kernel_version>-<release>`.
Signed-off-by: bia1708 <bianca-maria.popu@analog.com>
Move the Publish_to_Cloudsmith job from self-hosted agents to
cloud-hosted infrastructure to avoid disk space issues on the
self-hosted linux_default agent.
Signed-off-by: Liviu Tomoiaga <Liviu.Tomoiaga@analog.com>
Publish_to_Cloudsmith jobs run on:
- PRs (non-fork) targeting main, next_stable, or 202* branches
- Direct pushes to main, next_stable, or 202* branches
Signed-off-by: Liviu Tomoiaga <Liviu.Tomoiaga@analog.com>
Update DTBs build location to match the changes introduced by 3b1f15d
(zynq and socfpga DTS location change). Used for later upload to
artifacts server.
Signed-off-by: bia1708 <bianca-maria.popu@analog.com>
Use the checkout action resolved environment, detect merge commits and
choose parent, drop mirror main branch 'xcomm_zynq'
Signed-off-by: Jorge Marques <jorge.marques@analog.com>
Pipelines triggered by PRs coming from forks don't have access to the
secret variables used by the PushToArtifactory and TestHarness stages.
From a security standpoint, it makes sense to limit those pipelines to
not have access to our internal servers.
Add condition so that pipelines of PRs coming from forks skip the
PushToArtifactory and TestHarness stages.
Signed-off-by: bia1708 <bianca-maria.popu@analog.com>
Use artifacts provided by the analogdevicesinc/buildroot pipeline (the
rootfs for Microblaze) and a standalone cross compiler to run the specific
method for Microblaze from `run-build.sh` script.
Signed-off-by: bia1708 <bianca-maria.popu@analog.com>
Add stages:
- PushArtifacts: Run on a local agent, this stage uploads the artifacts to
the Artifactory server
- TestHarness: Triggers the Jenkins job for the Test Harness run
Signed-off-by: bia1708 <bianca-maria.popu@analog.com>
This brings Linux 6.6 support. Main note on the merge is to revert to
use xilinx version of timer-cadence-ttc. This driver used to give
problems with some zynq platforms. Let's see if this is still an issue
and if so fix it __and__ upstream the fix to xilinx (so we get it in a
future merge). Other changes:
* I2c .probe_new is now gone;
* Small fixes in IIO_ENUM where the proper enum symbol was not being
used (just explicit literal ir true/false booleans);
* Completely removed usage of IIO mlock;
* Legacy spi-engine driver using spi-controller (instead of master).
This one can bring some bugs (we need to double check). We should
anyways move to the new spi-engine driver in this cycle.
Signed-off-by: Nuno Sa <nuno.sa@analog.com>
Remove trigger for master branch because we switched to main and also
add next_stable branch as a trigger for azure build.
Signed-off-by: Andreea Andrisan <Andreea.Andrisan@analog.com>
- Update the help and examples from upload_to_artifactory.py script
to use 'main' for uploading to artifactory server.
- Update the default branch from which to cherry-picks commits to
rpi specific branches.
- Update lib.sh script to use 'main' as libiio default branch.
Signed-off-by: Stefan Raus <Stefan.Raus@analog.com>
As we now want to enforce new files to *not* be dual licensed, name the
job as expected by the CI script "check_new_file_license".
Signed-off-by: Nuno Sa <nuno.sa@analog.com>
Note that for now, only mxfe is supported in versal so we pretty much
ignore all of the other versal based devicetrees.
Signed-off-by: Nuno Sa <nuno.sa@analog.com>
dt_binding_check needs to be run before upstreaming any patch,
and since we intend on upstreaming ALL (or at least most) of our
changes, it makes sense to run the same validation locally.
Our tree is too old to properly run validation, and even if it
wasn't, schema changed drastically, and it doesn't make sense
to build a schema for our tree and then build it differently for
upstream, so just run the validation against upstream.
Signed-off-by: Cosmin Tanislav <cosmin.tanislav@analog.com>
This commit fixes a long standing issue with the checkpatch job. The
problem is that whenever a not so old commit is referenced in a new
commit (eg: on Fixes tag), checkpatch will report that the commit is not
known even though the commit exists. The problem is that a shallow clone
with depth=50 (to speed up `git fetch`) is being done in the job and
hence checkpatch won't know about the commit because we don't have the
whole history.
This all means that we need to fetch the whole git history. To do that
while minimizing fetch times, we'll do a partial clone. What this means is
that only the tree and blobs respective to HEAD will be fetched instead of
all trees and blobs for all existing commits. This considerably speeds up
the fetch time when compared to a full fetch (more than 50% is some tests).
As treeless clones are not supported in azure, we need to tell it to not
do any checkout and manually handle things before calling run-build.sh.
Signed-off-by: Nuno Sá <nuno.sa@analog.com>
For 4.19 we needed to apply some patches to be able to run the
dtb_build_test job properly.
We no longer need to do that, as the patches are in the mainline kernel.
So, remove the mechanism to apply patches.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
This was not omitted by accident during migration to Azure, and wasn't
working because the TRAVIS_PULL_REQUEST check was still around.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
These would probably be interesting later for some build statistics, if we
ever get interested in these.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
It reduces some duplication with regard to steps, as they need to be
defined only once.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
When this is defined, the CSE CI user is configured as commiter for
cherry-picking.
Otherwise git complains it needs a user.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
When we want to push back to Github, we need to have the credentials
available.
Luckily in Azure Pipelines, there is a flag that does that.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Azure doesn't support regex matching. It does support wildcard matching,
which isn't too bad. The odds of branches prefixed with '20' being matched
is pretty low.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
