| Age | Commit message (Collapse) | Author |
|---|
|
* Move the script for installing required packages into the main repo.
Based on discussion with Costa.
Most of what this script is actually used for directly is in the main
repo anyway, so it makes more sense to have it there so that any changes
get reviewed properly.
* Fix typo in RPM package build setup.
|
|
|
|
This decouples the PR labeling process from the main CI process, allowing
them to run completely independenly of each other.
The new labeling configuration is based on copying, extending, and
correcting the existing labeling rules in the `labeler.sh` script which
the commit is removing.
Labeling is configured through `.github/labeler.yml` file. Check there
for info on the exact format.
The exact configuration being used will update labels on PR's whenever
one is opened, updated, reopened, or marked ready for review. It will
not remove existing labels from the PR, only add new ones.
|
|
* Remove Fedora 28 from the lifecycle checks.
It's been EOL since 2019-05-28, and anybody continuing to use it is
recieving no support from upstream _or_ the Fedora community.
* Remove Ubuntu 18.10 from package builds.
It's been EOL since 2019-07-18, and anybody continuing to use it is
recieving no support from upstream or the Ubuntu community.
* Update docs.
|
|
* Add Ubuntu 19.10 to packaging and lifecycle checks.
It's been out since 2019-10-07.
* Update docs.
|
|
This adds a check to prevent the 'Support activities on main branch'
stage from running on forked repos. In such repos, it will automatically
fail without special effort from the user, requiring them to carry local
patches on their master branch.
Without this, CI builds of the 'master' branch will fail for users who
do not go through the effort of setting up the custom patches to disable
this stage in their repo.
|
|
* packaging: Set default release channel to stable for gh releases
* Leave kickstart files alone
* Update .travis/create_artifacts.sh
Co-Authored-By: Konstantinos Natsakis <5933427+knatsakis@users.noreply.github.com>
Co-authored-by: Mansour Behabadi <mansour@oxplot.com>
Co-authored-by: Konstantinos Natsakis <5933427+knatsakis@users.noreply.github.com>
|
|
* Skip unit testing during CI when it's not needed.
Our current unit testing takes almost 20 minutes on average during the
CI process, which is almost twice as long as any other step, and it's
only going to get longer as we get better test coverage.
This commit slightly alters how we handle unit testing during CI so that
only those unit tests which are actually needed get run.
This is achieved by using the `$TRAVIS_COMMIT_RANGE` environment
variable provided by Travis to determine which commits we're actually
testing, and then using that information to query Git for what files
actually changed.
As of right now, the only unit testing we're doing is in the dbengine
code, so we check to see if any of those files or the C unit testing
files are modified by the PR and only runs the unit tests for those if
they were modified. The checks are a bit more wide-ranging than they
probably need to be so that we make sure to avoid any false negatives.
This should speed up CI checks for PR's (as well as not monopolizing the
master branch after every PR that gets merged).
* Restructure to better handle PR's.
The previous code did not work correctly with PR's because it ignored
the fact that Travis does not appear to report the full list of commits
for a PR in `TRAVIS_COMMIT_RANGE`, instead only reporting the most
recent set of commits that were pushed.
* Add listing of commits being checked.
This will allow for verification of what is being tested, making it
easier to spot issues with the detection logic.
* Don't assume PR's are targeting the master branch.
|
|
The `matrix` key is officially deprecated and gets auto-merged with the
`jobs` key during config parsing.
If this auto-merging occurs, then the config will fail config
validation, which will in turn cause odd failues during the build
because of how Travis handles config validation (namely, it falls back
to a default config for the detected language).
This change allows us to enable the Config Validation feature in Travis,
which will in turn cause the build to fail very quickly if the
configuration is invalid in some way.
|
|
* Add a job to build and publish Fedora 31 RPMs
Fedora 31 was released on 2019-10-29.
* Also enable Fedora 31 lifecycle checks.
Previously disabled because Fedora 31 was not yet released at the time.
|
|
* .travis.yml: Do not define docker service
It is already there for the bionic image
* .travis.yml: Do not attempt to upgrade docker-ce package
It is already in the latest version for bionic image
* .travis.yml: Do not run docker info
Travis runs docker version by default
|
|
* .travis.yml: Add timestamps to output
* .travis.yml: Use '%H:%M:%.S' intead of '%.T'
Because %.T is not availabe in moreutils 0.60-1
* .travis.yml: Add timestamps to stderr as well
* .travis.yml: Add extra space to timestamp format for better readability
|
|
* ci: remove ubuntu trusty 14.04 from build
* Update packaging matrix with ubuntu 14.04 changes
|
|
|
|
Added a linting pass. This is non-blocking but will output a measure of how much each .c and .h file deviates from our current .clang format. Changed the standard build to include all of the warning that we are using in dev. Added the dependencies for dbengine and confirmed that the standard build enables dbengine. Fixed the original warnings.
|
|
* .travis.yml: Add tick() and retry() functions
* .travis.yml: Remove superfluous 'docker info' commands
The docker info command is ran during the install phase. There is no
need to run it again
* .travis.yml: Use the tick() function instead of travis_wait()
|
|
|
|
|
|
Build was failing, because where the function is called, we have only cloned the last 50 commits.
|
|
* netdata/packaging: bring el/6 first, because this appears on all cases. Otherwise travis craps out the steps
* netdata/ci: Attempt to sanitise workflow for master branch commits
1) Rename Packaging for release, now named: Support activities on main branch
2) Adjust instruction messages to reflect more realistically the activities
3) Put labeler first on the stage, since this will always run
4) Make changelog and tag for release conditionally run, so that it does not confuse people
* netdata/ci: Sanitise workflow for release more (explained below)
We need to clean up tagging logic more and in order to do this we need to bring tagger logic to travis.yml
Then we have the complete control of tagging and triggering for changelog generation in travis.
To mitigate this, we need to:
1) make tagger script a lirbary. Keep the tagging logic in a method and remove the things needed when it was an executable (checks, executions etc)
2) make travis handle GIT_TAG checking and setting at the beginning. We need this outside of any conditionals so that we have GIT_TAG result available inside the steps
3) COMMIT_TAG_POINTS_AT can now become GIT_TAG, since this is what we actually wanted. We just named differently before to avoid any weird conflicts
4) make changelog generation for release to NOT source tagger any more. Also make it fail in case GIT_TAG is not there, we wont run the script if git tag not in place so that should be error if it happens
5) Rename .travis/generate_changelog_for_release.sh -> .travis/create_changelog_for_release.sh
6) Rename .travis/generate_changelog_and_tag_release.sh -> .travis/generate_changelog_for_release.sh
* netdata/ci: reinstate filename here. we basically do the git tag operation here still
* netdata/ci: simplify more the changelog generation process
1) Rename create_changelog_for_release.sh -> create_changelog.sh
2) Add comments and the standard path checks in create_changelog.sh
3) Do some minor changes to help diff with nightlies changelog (no functional change though, primarily messages and move flags lower)
4) Inside generate_changelog_for_nightlies.sh, use create_changelog.sh and remove code related to create changelog process
5) in nightlies.sh, use standard shebang adn fix some messages
Note: might follow up with more changes, this is just a first batch
* netdata/packaging: reinstate accidentally removed OPTS from other iterationns (different engineer)
|
|
|
|
|
|
(#6724)
* netdata/packaging: baby steps, create the new package and introduce the install file
* netdata/packaging: [ci skip] change branch to be able to test
* netdata/packaging: we do define usrlib path to be /usr/lib
* netdata/packaging: silence sign errors for now
* netdata/packaging: Add rules for plugin-cups sub package
* netdata/packaging: [ci skip] That is libexec, not lib, duh
* netdata/packaging: [ci skip] explicitly arrange the other plugins also
* netdata/packaging: [ci skip] fix bug
* netdata/packaging: copy, dont move
* netdata/packaging: move all plugins around, also remove a slash
* netdata/packaging: [ci skip] align travis.yml changes with the ones we used on rpm cups pkg split
* netdata/packaging:[ci skip] add netdata dependency
* netdata/packaging:[ci skip] Update control files
* netdata/packaging: Add separate control file for buster, dependencies deviate (libprotoc17 instead of libprotoc10....)
* netdata/packaging: [ci skip] whoops
* netdata/packaging: [ci skip] Copy over files from temp dir, if destination non existent
* netdata/packaging:[ci skip] -d, not -f
* netdata/packaging: [ci skip] long shot
* netdata/packaging: [ci skip] revert this - need to retest
* netdata/packaging: [ci skip] remove the .install, as per an article input (https://askubuntu.com/questions/660373/dh-install-fails-with-missing-files) these files are for handling items not managed by the upstream build system, so bye bye
* netdata/packaging: Ok, lets ditch dh_auto_configure
* Revert "netdata/packaging: Ok, lets ditch dh_auto_configure"
This reverts commit 3b497cf5f6e157ce5429ffce38452515670a934a.
Well, that didn't affect anything
* netdata/packaging: [ci skip] here's a crazy thought
Okay, assuming that packager decided to use debian/tmp because we introduced multiple packages.
If that is the case, then move the custom stuff out first and then just move all debian/tmp to debian/netdata.
In theory, that should get me back on track for the rest of the package processing.
Lets see if i was right
* netdata/packaging: [ci skip] mv complains, forgot about that, fix
forgot about that, just copy for now to verify the use case and we will consider rsync approach for faster processing afterwards
* netdata/packaging: nope, that didn't work. rollback and remove also a suspicious backslash
* netdata/packaging: ok, trying this
* netdata/packaging: really?
* netdata/packaging: [ci skip] reinstate original setup, the bug probably was too obvious to spot
* netdata/packaging:[ci skip] now that you finally noted the www bug, try to copy over the rest of the stuff in
* netdata/packaging: [ci skip] long shot, etc is not properly created, lets see if it was already there and we messed it up
* netdata/packaging: [ci skip] fix perms, add placeholder folders in etc
* netdata/packaging: [ci skip] preserve ownership details when copying over
* netdata/packaging:[ci skip] reinstate branch condition
* netdata/packaging: redudant
* netdata/packaging: xenial - add cups subpackage
* netdata/packaging: remove cups dependency from those distros, missed to update the control file
* netdata/packaging: dont use personal account info
|
|
triggered only on specific keyword, that should be enough (#6834)
|
|
* netdata/ci: fine grain coverity scan toolkit
1) Deprecate coverity-install.sh
2) add set -e to raise errors more aggressively when something is wrong
3) refactor some variable definitions for temp and install paths, use mktemp and clean up temp dir when all ok
* netdata/ci: reduce the scope
* netdata/ci: require curl
* netdata/ci: go soft on functions loading
* netdata/ci: allow override of build version without touching the script
* netdata/ci: handle shellcheck errors, some were silenced because we dont handle them
* netdata/ci: coverity scan, parse any extra options other than --with-install and pass them to configure
1) add default configure arguments if no extras are given
2) parse all possible arguments individually, handle --with-install and pass the rest to scanit
* netdata/ci: focus, argument parsing in one place a bit more consistently
* netdata/ci: prepare for using
* netdata/ci: coverity - add more flags, adjust deps list
1) Add with math and with zlib flags on the script
2) add xenstat dependencies
3) run in coverity with custom flags, so that we add xenstat (no need to be by default for the users of the script)
* netdata/packaging: add xen-utils also
* test the flow
* netdata/ci: dont forget to go back to the original directory
* netdata/ci: please, travis, stop it
* netdata/ci: how about this then
* netdata/ci: silence that, we cant call it otherwise
* netdata/packaging: update documentation around xenstat on centos
* netdata/packaging: dont enable xenstat at this stage
* netdata/ci: add += instead of variable reusage
|
|
|
|
(#6808)
|
|
(#6766)
enable auto-generation of packages for RPM and DEB, during stable release
|
|
different version (#6753)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
coverity scanning
|
|
1) Add slack notification when things fail, we dont want to silently fail it
2) Change the tarball version, seems we now downloading a different version
|
|
|
|
Make packaging/docker/check_login.sh serve it's purpose and have the
ability to cancel the build promptly.
|
|
* netdata/packaging: Add purging logic for package cloud repositories
1) Introduce scriptlet that is executing the cleanup on a designated repository, with a designated retention period
2) Add .travis.yml configuration to execute purging logic with the right config on the nightly operations
* netdata/packaging: uncomment the temp dir clean up
* netdata/packaging: fix codacy
* netdata/packaging: [ci skip] Purge only -edge and -devel packages
|
|
|
|
|
|
* packaging/docker/{build,publish}.sh: Ability to externally defined ARCHS
* .travis.yml: Run docker build and release jobs in parallel
|
|
|
|
|
|
false for full repo retrieval
|
|
|
|
force pushing before approvals, i will be closely monitoring the effect by triggering nightlies on demand and evaluate the result
|
|
Travis CI deprecated the sudo tag as explained here:
https://blog.travis-ci.com/2018-11-19-required-linux-infrastructure-migration
and here:
https://changelog.travis-ci.com/linux-builds-run-on-vms-by-default-77106
|
|
* netdata/packaging: complete the list of bare OS validations, to be extra sure all cases are covered
* netdata/packaging: centos 6 has a different release file
* netdata/ci: [ci skip] no centos 8 yet
* netdata/ci: [ci skip] dont run for Debian 7 - its dead
* trigger travis
* netdata/ci: remove Debian 8 as bats not available with the standard repos - we will revisit our strategy on how we execute tests anyway
|
