Skip to content

Latest commit

 

History

History
439 lines (305 loc) · 39.4 KB

RELEASING.md

File metadata and controls

439 lines (305 loc) · 39.4 KB
Raw

Temurin Release Guide

Don't be scared off by this document! If you already understand the stuff in the glossary section and are only working on a HotSpot release, then skip to Steps for every version later on.

Release glossary and background information

OpenJDK Vulnerability Group

  • A private group of trusted people who take reports of vulnerabilities in the openjdk codebase and work to resolve them and get them into releases in a timely manner

Non-Oracle (usually Red Hat) maintained OpenJDK Quarterly Patch Set Update (PSU)

  • Maintainers work on quarterly update fixes for the next update, in the head stream of the updates repository e.g. https://hg.openjdk.java.net/jdk-updates/jdk11u
  • Fixes for the subsequent update are developed in the dev stream e.g. https://hg.openjdk.java.net/jdk-updates/jdk11u-dev/
  • Regular builds are tagged every week or so, e.g. jdk-11.0.5+6
  • Eventually after final testing it comes to general availability (GA) day for the update, at this point any fixes from the Vulnerability Group are merged with the final GA build, and tagged, e.g. jdk-11.0.5+10, jdk-11.0.5-ga

Oracle Managed OpenJDK Quarterly PSU

  • The most recent JDK release updates are managed by Oracle, and there will only be two of them, e.g. for jdk-13 Oracle produced jdk-13.0.1 and jdk-13.0.2.
  • Oracle work on the quarterly updates internally and these update source branches and Vulnerability Group fixes are not public until GA date.
  • On GA day, Oracle merges the internal branch and Vulnerability Group fixes to produce the final GA build, and this is tagged, e.g. jdk-13.0.1+9, jdk-13.0.1-ga
  • If the release is a short term support release there are no more releases after the two Oracle-led updates, but if it is a long term support (LTS) release the OpenJDK community picks up subsequent release maintenance, and all work continues in public as described above.

OpenJDK Quarterly/New Release Process

Lockdown period

During the week before release, the Release Champion makes changes in preparation for the release:

Branching message for build related repositories

Post the below message to the #build & #release channels in Slack:

In Preparation for next weeks release, I'm proposing to branch the following repositories, in order that this branch becomes the baseline for the release. Shout now if you need to get any PRs merged for the release, as today is the last day, the following repositories will be branched: ( temurin-build, ci-jenkins-pipelines, jenkins-helper).

Create release branch on below repositories:

Create release branch in the format vYYYY.MM.NN on each of the following repositories:

These branches should be named according to the following format (vYYYY.MM+NN) ,e.g v2023.03+01 , whereby the final element is an incremental counter appended to the year and month of the release.

If anything needs to be merged into the new branch, it should typically be merged into master, then a git cherry-pick operation should be done to create a new PR against the release branch. This can typically be merged without further approval.

Code Freeze message

Paste the below message into the #release channel in Slack:

With under a week to go until releases, we are entering a lockdown period for the following repositories: temurin-build, ci-jenkins-pipelines, github-release-scripts, containers, installer, and mirror-scripts.

If you need to submit a pr for any of these repos during this period, you should:

  • Add a comment saying “Approval to merge during the lockdown cycle please” and post in the appropriate slack channel for awareness. This can be done before the PR is finalised.
  • Add a note into this channel saying you are requesting the approval with a link to the comment in the first bullet point.
  • The comment should have approval from at least one build committer and one PMC member to indicate that they agree it is critical that it goes in.
  • The PR can be merged after 2 hours of the post going into the build channel (to give people time to object). This delay may be skipped in cases where the delay will result in something breaking within that time.

Only include "critical" fixes (i.e. those which will otherwise cause a build break or other problem which will prevent shipping the release builds). This stops last minute changes going in, which may destabilise things. "installer" repo might have exception due to the fact it requires new version of build. If a change has to go in during this "lockdown" period it should be done by posting a comment saying "Requesting approval to merge during the lockdown period. Please thumbs up the comment to approve" in Slack release channel. If two committers into the repository express approval then the change can be merged during the lockdown period.

Enable code-freeze on main branches of below repositories

In order to enable the code freeze GitHub you need to change the line if: github.repository_owner == 'adoptium' && false to be if: github.repository_owner == 'adoptium' && true in the code-freeze.yml GitHub workflow. Please contact the PMC if you need help merging this change.

Affected repositories:

Release pipelines and jobs need to be re-generated with new tags by the Release Champion:

  • run release-pipeline-generator with correct value:
    1. releaseTag is the branch on ci-jenkins-pipeline and temurin-build git repo.
    2. helperTag is the branch on jenkins-helper repo.
    3. aqaTag is the branch on aqa-tests repo, in form of vX.Y.Z-release and usually the latest stable release
  • ensure release-openjdkXX-pipeline in https://ci.adoptium.net/job/build-scripts are properly generated for the first time or updated:
    1. jdk8 aarch32Linux shares the same pipeline release-openjdk8-pipeline with other jdk8 targets, therefore, only one release pipeline for jdk8
    2. targetConfigurations should only include what we officially release for temurin.
    3. buildReference and ciReference should have the value of releaseTag used when we generate pipeline.
    4. helperReference should have the same value of helperTag used when we generate pipeline.
    5. additionalConfigureArgs has correct value, especially in release-openjdk8-pipeline it is different than other jdk versions. This requires certain modification manually
  • ensure downstream build jobs in https://ci.adoptium.net/job/build-scripts/job/jobs/job/release/job/jobs/jdkXXu/ are created or updated
    1. BUILD_CONFIGURATION.USE_ADOPT_SHELL_SCRIPTS is set to true
    2. DEFAULTS_JSON.repository.build_branch, ADOPT_DEFAULTS_JSON.repository.build_branch, DEFAULTS_JSON.repository.pipeline_branch and ADOPT_DEFAULTS_JSON.repository.pipeline_branch should get correct release branch name as releaseTag
    3. DEFAULTS_JSON.repository.helper_ref and ADOPT_DEFAULTS_JSON.repository.helpe_ref should get correct release branch name as helperTag
FLOW CHART OF THE PIPELINE GENERATOR PROCESS 
flowchart TD

1["ReleaseChampion"] -- manual run --> 2["release-build-pipeline-generator"] 

2 --generate--> 3.1["release-openjdk8-pipeline"]
2 --call--> 3.2["release_pipeline_jobs_generator_jdk8u"]
2 --generate--> 3.3["release-openjdk11-pipeline"]
2 --call--> 3.4["release_pipeline_jobs_generator_jdk11u"]
2 --generate--> 3.5["release-openjdk17-pipeline"]
2 --call--> 3.6["release_pipeline_jobs_generator_jdk17u"]
2 --generate--> 3.7["release-openjdkXX-pipeline"]
2 --call--> 3.8["release_pipeline_jobs_generator_jdkXXu"]

3.2-- create --> 3.2.1["jdk8u-release-mac-x64-temurin"]
3.2-- create --> 3.2.2["jdk8u-release-linux-x64-temurin"]
3.2-- create --> 3.2.3["jdk8u-release-<os>-<arch>-temurin"]

3.4-- create --> 3.4.1["jdk11u-release-mac-x64-temurin"]
3.4-- create --> 3.4.2["jdk11u-release-linux-x64-temurin"]
3.4-- create --> 3.4.3["jdk11u-release-<os>-<arch>-temurin"]

3.6-- create --> 3.6.1["jdk17u-release-mac-x64-temurin"]
3.6-- create --> 3.6.2["jdk17u-release-linux-x64-temurin"]
3.6-- create --> 3.6.3["jdk17u-release-<os>-<arch>-temurin"]

3.8-- create --> 3.8.1["jdkXXu-release-mac-x64-temurin"]
3.8-- create --> 3.8.2["jdkXXu-release-linux-x64-temurin"]
3.8-- create --> 3.8.3["jdkXXu-release-<os>-<arch>-temurin"]

Disable nightly testing so the release builds aren't delayed by any nightly test runs (set enableTests : false in defaults.json). Ensure the build pipeline generator job runs successfully (https://ci.adoptium.net/job/build-scripts/job/utils/job/build-pipeline-generator/), and the flag is disabled by bringing up the Build pipeline job and check the enableTests checkbox is unticked.

Add a banner to the website to indicate that the releases are coming in the near future (Example Changes).

Steps for every version

Automatic trigger of GA pipeline jobs

In order to reduce time to GA, we have automated triggers to ensure that the GA build pipelines are triggered as soon as the GA tags come out so it does not rely on one of the Adoptium team members watching it and kicking them off manually. These examples use JDK17 - adjust for the version you're interested in.

  1. Jenkins "release trigger" job (e.g https://ci.adoptium.net/job/build-scripts/job/utils/job/releaseTrigger_jdk17u/) runs every hour in the release week to check if new GA tag has been detected in the adoptium's source code repo - the script run from the checks for the new release every 10 minutes five times (e.g https://github.com/adoptium/jdk17u) This excludes https://github.com/adoptium/aarch32-jdk8u
  2. If it finds new GA tag matches expected tag set in mirror-script repo, job triggers release-openjdk19-pipeline (e.g https://ci.adoptium.net/job/build-scripts/job/release-openjdk19-pipeline/) with parameters: scmReference.
  3. If it couldn't find the correct "_adopt" tag but GA tag has been applied in the upstream Skara source code repo. Several things can check:

"release-openjdkXX-pipeline" can be manually run by Release Champion: set scmReference with correct "_adopt" tag.

  1. For jdk8 aarch32Linux, the Release Champion needs to manually trigger https://ci.adoptium.net/job/build-scripts/job/release-openjdk8-pipeline/
FLOW CHART OF THE RELEASE TRIGGER PROCESS 
flowchart TD

start["Every hour"] --trigger--> step0["releaseTrigger job"] --> step1["Check GA tag available\nCheck _adopt tag available"] ----> |yes| step2["Trigger release-openjdkXX-pipeline\nwith scmReference"] --> step3["Trigger all jdkXXu-release-OS-ARCH-temurin job\nExcept jdk8u-release-linux-arm-temurin when XX is 8"] --> step4["Trigger remote Temurin compliance CI jck test jobs: jck.sanity,jck.extended,jck.special"]
step1 ---->|No| step1.1["sleep 10 minutes"] -->step1

jdk8armStep1["ReleaseChampion check once GA tag on jdk8 aarch32Linux is ready"] -->jdk8armStep2["Get _adopt tag on jdk8 aarch32Linux"] -->jdk8armStep3["Manual run release-openjdk8u-pipeline with\n1: scmReference\n2: targetConfiguration\n3: overridePublishName"]

Dry run tests: Do this at least 2 weeks before release in the same calendar month

It is recommended that we perform an auto trigger test on a chosen version (suggest jdk8 and one other) to validate the trigger and build processes and the release pipeline. jdk-17 example:

  1. Update releasePlan.cfg with the correct version numbers for the new release.
  2. Choose the second latest openjdk tag without the _adopt suffix (unless it is the same as the latest in which case keep going backwards..) in the adoptium/jdkNNu repository that you are using for the dry run.
  3. Ensure that the branch of aqa_tests has been created for this release.
  4. Update testenv/testenv.properties in the branch of aqa-tests to point to the tag as the JDKnn_BRANCH e.g. jdk-17.0.x+y (i.e. not dev)
  5. Get an Adoptium administrator to create the -dryrun tag to build in the adoptium mirror, as in the following example:

IMPORTANT: trial tag MUST be something that is sorted before -ga. Recommended format: "-dryrun-ga"

git clone git@github.com:adoptium/jdk17u.git

cd jdk17u

git tag -a "jdk-17.0.6-dryrun-ga" jdk-17.0.6+8^{} -m"YYYY.MM release dry run test"

git push --tags origin master

  1. Wait for the release trigger job to detect the tag (wait up to 10mins), e.g. releaseTrigger_jdk17u (Note that the schedule for that job is only run on the release months, so may not work if you are keen and try to do this in the month before)
  2. The trial release pipeline job should now be running, eg: https://ci.adoptium.net/job/build-scripts/job/release-openjdk17-pipeline/
  3. Once you have verified that everything looks good, testenv.properties should be adjusted to have the expected GA tag before the final release appears.
Manual execution of the build pipelines (without using trigger jobs - now mostly obsolete other than jdk8u/arm32)

Here are the old manual steps:

  1. Ensure that the appropriate mirror job has completed and that the corresponding repository at https://github.com/adoptium/jdkXX has successfully received the tag for the level you are about to build. If there are any conflicts they can be resolved on the machine where it failed if you have access to the private adoptium_temurin_bot_ssh_key.gpg key, or ask someone( e.g @gdams or @andrew-m-leonard) with push access to the repositories to manually run the mirror job and resolve the conflicts.
  2. Run OpenJDK pipeline build and test release:
  • Job: https://ci.adoptium.net/job/build-scripts/job/openjdk8-pipeline/build (Switch openjdk8 for your version number)
  • targetConfigurations:
    • only keep temurin variants
    • make sure windows aarch64 is not presented (as this is written, it is not officially supported yet)
  • releaseType: Release
  • adoptBuildNumber: Leave blank unless you are doing a point release in which case it should be a number starting at 1 for the first point release.
  • additionalConfigureArgs:
    • For JDK8, no need change default value. It automatically adds--with-milestone=fcs in build.sh.
    • For JDK11+ use --without-version-pre --without-version-opt (for EA releases use: --with-version-pre=ea --without-version-opt)
  • scmReference: One of the following:
    • For jdk8 aarch32Linux, the tag usually takes the form jdk8u322-b04-aarch32-xxxxxxxx
    • For the rest, it's the same tag suffixed with _adopt e.g. jdk-17.0.2+9_adopt
  • buildReference: tag or the ongoing release on temurin-build repo
  • ciReference: tag or the ongoing release on ci-jenkins-pipeline repo
  • helperReference: tag for the ongoing release on jenkins-helper repo
  • overridePublishName: only for jdk8 aarch32Linux, to set to the actual OpenJDK tag (jdk8u322-b04)
  • aqaReference should be set to the appropriate branch of the aqa-tests repository which is appropriate for this release. Generally of the form vX.Y.Z-release
  • enableTests: tick
  • Click "Build" button !!!

After build pipeline finished

Once the openjdk pipeline has completed:

  1. Triage TRSS result:
  • Follow triage the results. Go to TRSS which will guide you through creating an aqa test triage issue for the release
  • Find the section of each jdk build, e.g openjdk8-pipeline in server https://ci.adoptium.net/job/build-scripts for JDK8
  • Click "Grid" link on the correct Build row
  • Check if not all are "Green", create new "release triage" issue in aqa-tests repository, set description to "Release Summary Report" content and follow the Jenkins link to triage error and failure.
  • Raise issues either at:
  • Discuss failing tests with Shelley Lambert or post on testing-aqavit Slack channel
  • Once all AQA tests on all platforms and all JDK versions have been signed off, then nightly tests can be re-enabled. See the notes on "Disable nightly testing".
  1. Publish build results:

  • If "good to publish", get permission to publish the release from the Adoptium PMC members, discussion is via the Adoptium #release Slack channel.

  • Once permission has been obtained, run the openjdk_release_tool to publish the releases to GitHub (restricted access - if you can't see this link, you don't have access). It is strongly recommended that you run first with the DRY_RUN checkbox enabled and check the output to verify that the correct list of files you expected are picked up.

    -- TAG: (GitHub binaries published name)  e.g. jdk-11.0.5+9. If doing a point release, add that into the name e.g. for a .3 release use something like this: jdk8u232-b09.3 -- VERSION: (select version e.g. jdk11) -- UPSTREAM_JOB_NAME: e.g "build-scripts/release-openjdkXX-pipeline" for new way and "build-scripts/openjdkXX-pipeline" for old way -- UPSTREAM_JOB_NUMBER: the build number of above upstream job, e.g. 86 -- RELEASE: "ticked" -- If you need to restrict the platforms or only ship jdks or jres, either use ARTIFACTS_TO_COPY e.g. **/*jdk*mac* or add an explicit exclusion in ARTIFACTS_TO_SKIP e.g. **/*mac*. These may be required if you had to re-run some of the platforms under a different pipeline earlier in the process. If you're unsure what the possible names are, look at the artifacts of the appropriate openjdkNN-pipeline job. If you are shipping x64_linux ensure that you include the sources tar.gz files with the corresponding checksum and json file. -- ARTIFACTS_TO_SKIP: **/*testimage* -- If you need to restrict the platforms, fill in ARTIFACTS_TO_COPY and if needed add to ARTIFACTS_TO_SKIP. This may also be required if you had to re-run some of the platforms under a different pipeline earlier in the process. I personally tend to find it cleaner to release Linux in one pipeline, Windows+Mac in another, then the others together to keep the patterns simpler. Sample values for ARTIFACTS_TO_COPY are as follows (use e.g. _x64_linux_ to restrict by architecture if required): --- **/*_linux_*.tar.gz,**/*_linux_*.sha256.txt,**/*_linux_*.json,**/*_linux_*.sig (Exclude **/*alpine_linux* if you don't really want that to be picked up too) --- Alternative that wouldn't pick up Alpine: target/linux/x64/hotspot/**.tar.gz,target/linux/x64/hotspot/target/linux/x64/hotspot/*.sha256.txt --- **/*_mac_*.tar.gz,**/*_mac_*.sha256.txt,**/*_mac_*.json,**/*_mac_*.pkg,**/*_mac_*.sig --- **/*_windows_*.zip,**/*_windows_*.sha256.txt,**/*_windows_*.json,**/*_windows_*.msi,**/*_windows_*.sig --- **/*_aix_*.tar.gz,**/*_aix_*.sha256.txt,**/*_aix_*.json,**/*_aix_*.sig --- **/*_solaris_*.tar.gz,**/*_solaris_*.sha256.txt,**/*_solaris_*.json,**/*_solaris_*.sig -- Click "Build" button !!!

  • Once the job completes successfully, check the binaries have uploaded to GitHub at somewhere like https://github.com/adoptium/temurin8-binaries/releases/tag/jdk8u302-b08

  • Within 15 minutes the binaries should be available on the website too. e.g. https://adoptium.net/?variant=openjdk11&jvmVariant=hotspot (NOTE: If it doesn't show up, check whether the API is returning the right thing (e.g. with a link such as this, and that the .json metadata files are uploaded correctly)

  • During the waiting time, good to update:

    -- https://github.com/adoptium/website-v2/blob/main/src/asciidoc-pages/support.adoc which is the source of https://adoptium.net/support (Sample change) -- (if required) the supported platforms table at https://github.com/adoptium/website-v2/edit/main/src/asciidoc-pages/supported-platforms.adoc which is the source of https://adoptium.net/supported-platforms

  1. Publish packages for different OS

3.1. [Mac only] Once the binaries are available on the website you need to update the Homebrew casks. There are 4 casks in total and all but the first one is in the hombrew-cask-versions repository. If you're doing a point release, the format of the version string is 11.0.20.1,1 so the version is always the same as "our" one but with the + replaced with a ,

An example PR can be found here. The required SHA sums can be updated by brew bump-cask-pr temurinXX --version 11.0.XX,Y command if you're on a macos system, or manually if not . The separate pull request is required for each version you update. If in doubt reach out to @gdams as he's a maintainer.

3.2. [Linux only] Once the binaries are available on the website you can begin updating the specfiles for the RPM/DEB/APK files. There are 4 different types of linux installer

  • debian

  • Red Hat

  • SuSE

  • Alpine

    All need to be updated:

    • Debian you need to locate the rules file for each version. This file contains the URLs and Checksums for each package (example). The changelog file should also be updated for the new version.
    • Red Hat/SuSE you need to modify the temurin-<version>-jdk.spec file for each version. The links/checksum links are all defined as Source<number> variables (example).
    • Alpine you need to modify pkgver and checksum

Once the PRs to change those files have been merged, the adoptium-packages-linux-pipeline job needs to be kicked off. It is recommended to run it without the UPLOAD checkbox to begin with as a 'dry-run' before re-running with the UPLOAD checkbox ticked to publish to our JFrog artifactory instance.

  1. [Docker Hub] The information on updating the Adoptium official dockerhub repository is at https://github.com/adoptium/containers#maintenance-of-dockerfiles at the moment you cannot do this until all Linux architectures and windows64 are published for the appropriate version

  2. Once everything has been published to GitHub, use the EclipseMirror job to mirror the artifacts to our Eclipse server for backup purposes. Note that this will need to be done by a team member in the temurin-compliance project (Run once for each of the releases)

  3. Publicise the Temurin release:

  • Via slack on the Adoptium #release channel
  • Find someone with the appropriate authority (Carmen, George, Martijn, Shelley, Stewart, Tim) to post a tweet about the new release from the Adoptium twitter account

Post Release Tasks

Once all the release binaries have been published the following tasks should be completed: If the latest version just released has come to the end of its non-LTS lifecycle (2 CPU updates, eg.jdk-15.0.2)

For the api.adoptium.net repository:

  • Update the LATEST_JAVA_VERSION_PROPERTY in Versions.kt to be n+1, where n is the highest major version you've just released.
  • Ensure that LATEST_JAVA_VERSION is not set on the api.adoptium.net server, as it will override the LATEST_JAVA_VERSION_PROPERTY.

OpenJDK "New" Major Release process

  • The refers to a "new" major (Short or Long Term) OpenJDK Release (e.g. jdk13, jdk14, ...)

  • Oracle and contributors work on releases in the "head" OpenJDK stream: https://hg.openjdk.java.net/jdk/jdk

  • 3 months prior to the GA date, the head stream is branched into a new release stream for development rampdown e.g. https://hg.openjdk.java.net/jdk/jdk13

  • Regular builds are tagged every week or so in a format such as jdk-13+21

  • Eventually after rampdown and final phase testing the GA build is tagged and released, e.g. the jdk-13-ga code level is tagged along side the actual release build tag.

  • When a new release occurs, we must also update one of our job generators to match the new jdk versions and remove old STR that are no longer needed. The full details on what these are in the regeneration README.md but for a quick run down on how to update them when we want to build a new release, follow the steps below:

    1. Update the Job Folder - https://ci.adoptium.net/job/build-scripts/job/utils/: The jobs themselves you are looking for are called pipeline_jobs_generator_jdkxx (pipeline_jobs_generator_jdk for HEAD). Firstly, ensure that the job description of each generator (and it's parameter's descriptions) are up to date. Then, follow these steps:
    • If you are ADDING a JDK version:
      • Ensure that JDK N-1 is available as build JDK on the builders. For example in order to build JDK 15, JDK 14 needs to be installed on the build machines. As a temporary measure, code so as to download the JDK to the builder via the API has been added. NOTE: For the transition period shortly after a new JDK has been branched, there might not yet exist a generally available release of JDK N-1.
      • Ensure that JDK sources are being mirrored. Example infrastructure request
      • Ensure that a repository which contains the binary releases exists. Example temurin15-binaries
      • Add build scripts for the new JDK release. Example for JDK 14
      • Regenerate build jobs:
        • Create a New Item in the folder linked above that copies the pipeline_jobs_generator_jdk job. Call it pipeline_jobs_generator_jdk<new-version-number>.
        • Change the Script Path setting of the new job to pipelines/build/regeneration/jdk<new-version-number>_regeneration_pipeline.groovy. Don't worry if this currently doesn't exist in this repository, you'll add it in step 3.
        • Update the Script Path setting of the JDK-HEAD job (pipeline_jobs_generator_jdk) to whatever the new JDK HEAD is. I.e. if the new head is JDK16, change Script Path to pipelines/build/regeneration/jdk16_regeneration_pipeline.groovy
    • If you are REMOVING a JDK version:
      • Delete the job pipeline_jobs_generator_jdk<version-you-want-to-delete>
    1. Create the new build configurations for the release - https://github.com/adoptium/ci-jenkins-pipelines/tree/master/pipelines/jobs/configurations:
    • Create a new jdk<new-version-number>_pipeline_config.groovy file with the desired buildConfigurations for the new pipeline. 99% of the time, copy and pasting the configs from the previous version is acceptable. Ensure that the classname and instance of it is changed to Config<new-version-number>. Don't remove any old version configs.
    • Furthermore, you will also need to create another config file to state what jobs will be run with any new versions. If it doesn't currently exist, add a jdkxx.groovy file to configurations/. Example on how to do this. Note, some files will need to be named jdkxxu.groovy depending on whether the version is maintained in an update repository or not. These will be the ONLY os/archs/variants that are regenerated using the job regenerators as described in the regeneration readme.
    1. Create a new Regeneration Pipeline for the downstream jobs - https://github.com/adoptium/ci-jenkins-pipelines/blob/master/pipelines/build/regeneration:

    Create a new jdk<new-version-number>_regeneration_pipeline.groovy. Ensure that the javaVersion, targetConfigurations and buildConfigurations variables are what they should be for the new version. Don't remove any old version configs. While you're here, make sure all of the current releases have a regeneration_pipeline.groovy file (including head). If they don't, create one using the same technique as above.

    1. Build the pipeline_jobs_generator that you just made. Ensure the equivalent openjdkxx_pipeline to the generator exists or this will fail. If the job fails or is unstable, search the console log for WARNING or ERROR messages for why. Once it has completed successfully, the pipeline is ready to go!

    2. Update the view for the build and test pipeline calendar to include the new version.

Update Repository (jdkXXu)

At some point in a java version's lifecycle, the JDK version will be maintained in an update repository. The first notification of this will be via mail list in one of two places:

New Adoptium mirror repository creation, by an Adoptium GitHub Admin:

  1. Create a new empty repository adoptium/openjdk-jdkNNu
  2. Rename mirror job from https://ci.adoptium.net/view/git-mirrors/job/git-mirrors/job/git-skara-jdkNN to https://ci.adoptium.net/view/git-mirrors/job/git-mirrors/job/git-skara-jdkNNu
  3. Update mirror job "Execute shell" to pass jdkNNu as parameter to bash ./skaraMirror.sh jdkNNu
  4. Run the renamed job twice, first one will fail due to empty repository, 2nd run should succeed.
  5. Add the Adoptium.md "marker" text file to both branches "dev" and "release".

When the repository has been created, a few changes to the codebase will be necessary where the code references a jdk version but not it's new update version. I.e. jdk11 became jdk11u when it was moved to an update repository.

If a product is to be moved to an update repository, follow these steps in chronological order to ensure our builds continue to function:

  1. ci-jenkins-pipelines: Update the configurations Rename the nightly build targets file (it will be named jdkxx.groovy, example here) to be jdkxxu.groovy. Do the same for the pipeline config file (named jdkxx_pipeline_config.groovy, example here).

  2. ci-jenkins-pipelines: Update version from jdkxx to jdkxxu inside docs/generateBuildMatrix.sh

  3. ci-jenkins-pipelines: Run on "linux" platform your updated docs/generateBuildMatrix.sh script to generate the updated README.md table to be updated at the end of README.md

  4. openjdk-build: Update the JDKXX_VERSION from jdkxx to jdkxxu inside the build script constants that is being shifted to an update repository.

  5. openjdk-build: Update the version from jdkxx to jdkxxu inside .github/workflows/build.yml

  6. Merge both ci-jenkins-pipelines and openjdk-build Pull Requests.

  7. Cancel jdkxx job regenerator that will have just been triggered: job regenerator

  8. Rename the jenkins pipeline jobs regenerator job regenerator from pipeline_jobs_generator_jdkxx to pipeline_jobs_generator_jdkxxu. Then manually re-build.

  9. Check the regenerator has created all the new jdkxxu build jobs successfully: build jobs

  10. Delete the old jdkxx build jobs folder: https://ci.adoptium.net/job/build-scripts/job/jobs/jdkxx: build jobs

  11. Submit a test pipeline build

Summary on point releases

Occasionally we may have to do an out-of-band release that does not align with a quarterly release from the upstream OpenJDK project. This may occur if there has been a problem with our build process that we missed at GA time, to fix a critical issue, or when a project outside OpenJDK :

  1. When triggering the pipeline, set AdoptBuildNumber to a unique number for the point release (the original will be "blank" so any subsequent point release required should start at "1")
  2. If you used a custom entry in overridePublishName when kicking off the GA pipeline, keep it the same as for the GA release - we DO NOT want the filenames changed to include the point number
  3. When running the publish job, you need to use a custom TAG in order to publish it to the website with a separate name from what you had initially e.g. jdk-11.0.5+10.1_openj9-0.17.1 (Note the position of the .1 for OpenJ9 releases in that example - it's after the openj9 version but before the OpenJ9 version.

If you need to create a point release with a one-off patch (Usually a cherry pick of something already in the codebase which will go into the next release) use the following process:

  1. Clone the repository you need to patch (e.g. git clone git@github.com:adoptium/jdk21u)
  2. Check out the tag you wish to base it on (e.g. git checkout jdk-21.0.1+12_adopt)
  3. Create a branch for the new release (e.g. git checkout -b jdk-21.0.1+12.1)
  4. Cherry pick your patches to apply them onto the source
  5. Create a new _adopt tag for the point release (e.g. git tag -a jdk-21.0.1+12.1_adopt)
  6. Push the branch and tag: git push origin jdk-21.0.1+12.1 && git push origin jdk-21.0.1+12.1_adopt
  7. If the patch does not affect tests, create a PR to update testenv.properties in the aqa-tests release branch with the point release version number (Sample PR against v0.9.9-release)
  8. If the patch does change tests, a new ".1" branch should be created in aqa-tests, based off the release branch, which has the updatre to testenv.properties (If this was needed in the above example, you would create a v0.9.9.1-release branch)
  9. Run the release-openjdkXX-pipeline with:
    • the new _adopt tag as the scmReference
    • additionalConfigureArgs of --with-version-build=12 (replace 12 with the number after + but before new new .1 in the version string)
    • the aqaReference updated if step 8 was followd
    • the desired set of platforms defined in the targetConfigurations parameter