5. Contributor’s Guide¶
5.1. Getting Started¶
If you plan to contribute a major piece of work, it is usually a good idea to start a discussion around it on the mailing list. This gives everyone visibility of what is coming up, you might learn that somebody else is already working on something similar or the community might be able to provide some early input to help shaping the design of the feature.
If you intend to include Third Party IP in your contribution, please mention it explicitly in the email thread and ensure that the changes that include Third Party IP are made in a separate patch (or patch series).
Create a local topic branch based on the Trusted Firmware-A
5.2. Making Changes¶
Make commits of logical units. See these general Git guidelines for contributing to a project.
Ensure your commit messages comply with the Conventional Commits specification:
<type>[optional scope]: <description> [optional body] [optional footer(s)]
You can use the tooling installed by the optional steps in the prerequisites guide to validate this locally.
Keep the commits on topic. If you need to fix another bug or make another enhancement, please address it on a separate topic branch.
Split the patch in manageable units. Small patches are usually easier to review so this will speed up the review process.
Avoid long commit series. If you do have a long series, consider whether some commits should be squashed together or addressed in a separate topic.
Ensure that each commit in the series has at least one
Signed-off-by:line, using your real name and email address. The names in the
Commit:lines must match. By adding this line the contributor certifies the contribution is made under the terms of the
Developer Certificate of Origin.
There might be multiple
Signed-off-by:lines, depending on the history of the patch.
More details may be found in the Gerrit Signed-off-by Lines guidelines.
Ensure that each commit also has a unique
Change-Id:line. If you have cloned the repository with the “Clone with commit-msg hook” clone method (following the Prerequisites document), this should already be the case.
More details may be found in the Gerrit Change-Ids documentation.
Write informative and comprehensive commit messages. A good commit message provides all the background information needed for reviewers to understand the intent and rationale of the patch. This information is also useful for future reference.
What does the patch do?
What motivated it?
What impact does it have?
How was it tested?
Have alternatives been considered? Why did you choose this approach over another one?
If it fixes an issue, include a reference.
Use the checkpatch.pl script provided with the Linux source tree. A Makefile target is provided for convenience, see this section for more details.
Where appropriate, please update the documentation.
If you are submitting new files that you intend to be the code owner for (for example, a new platform port), then also update the Code owners file.
For topics with multiple commits, you should make all documentation changes (and nothing else) in the last commit of the series. Otherwise, include the documentation changes within the single commit.
Ensure that each changed file has the correct copyright and license information. Files that entirely consist of contributions to this project should have a copyright notice and BSD-3-Clause SPDX license identifier of the form as shown in License. Files that contain changes to imported Third Party IP files should retain their original copyright and license notices.
For significant contributions you may add your own copyright notice in the following format:
Portions copyright (c) [XXXX-]YYYY, <OWNER>. All rights reserved.
where XXXX is the year of first contribution (if different to YYYY) and YYYY is the year of most recent contribution. <OWNER> is your name or your company name.
Ensure that each patch in the patch series compiles in all supported configurations. Patches which do not compile will not be merged.
Please test your changes. As a minimum, ensure that Linux boots on the Foundation FVP. See Arm Fixed Virtual Platforms (FVP) for more information. For more extensive testing, consider running the TF-A Tests against your patches.
Ensure that all CI automated tests pass. Failures should be fixed. They might block a patch, depending on how critical they are.
5.3. Submitting Changes¶
Submit your changes for review at https://review.trustedfirmware.org targeting the
Add reviewers for your patch:
At least one code owner for each module modified by the patch. See the list of modules and their Code owners.
At least one maintainer. See the list of Maintainers.
If some module has no code owner, try to identify a suitable (non-code owner) reviewer. Running
git blameon the module’s source code can help, as it shows who has been working the most recently on this area of the code.
Alternatively, if it is impractical to identify such a reviewer, you might send an email to the TF-A mailing list to broadcast your review request to the community.
Note that self-reviewing a patch is prohibited, even if the patch author is the only code owner of a module modified by the patch. Getting a second pair of eyes on the code is essential to keep up with the quality standards the project aspires to.
The changes will then undergo further review by the designated people. Any review comments will be made directly on your patch. This may require you to do some rework. For controversial changes, the discussion might be moved to the TF-A mailing list to involve more of the community.
Refer to the Gerrit Uploading Changes documentation for more details.
The patch submission rules are the following. For a patch to be approved and merged in the tree, it must get:
Code-Owner-Review+1for each of the modules modified by the patch.
In the case where a code owner could not be found for a given module,
Code-Owner-Review+1is substituted by
In addition to these various code review labels, the patch must also get a
Verified+1. This is usually set by the Continuous Integration (CI) bot when all automated tests passed on the patch. Sometimes, some of these automated tests may fail for reasons unrelated to the patch. In this case, the maintainers might (after analysis of the failures) override the CI bot score to certify that the patch has been correctly tested.
In the event where the CI system lacks proper tests for a patch, the patch author or a reviewer might agree to perform additional manual tests in their review and the reviewer incorporates the review of the additional testing in the
Code-Owner-Review+1as applicable to attest that the patch works as expected. Where possible additional tests should be added to the CI system as a follow up task. For example, for a platform-dependent patch where the said platform is not available in the CI system’s board farm.
When the changes are accepted, the Maintainers will integrate them.
Typically, the Maintainers will merge the changes into the
If the changes are not based on a sufficiently-recent commit, or if they cannot be automatically rebased, then the Maintainers may rebase it on the
integrationbranch or ask you to do so.
After final integration testing, the changes will make their way into the
masterbranch. If a problem is found during integration, the Maintainers will request your help to solve the issue. They may revert your patches and ask you to resubmit a reworked version of them or they may ask you to provide a fix-up patch.
5.4. Binary Components¶
Platforms may depend on binary components submitted to the Trusted Firmware binary repository if they require code that the contributor is unable or unwilling to open-source. This should be used as a rare exception.
All binary components must follow the contribution guidelines (in particular licensing rules) outlined in the readme.rst file of the binary repository.
Binary components must be restricted to only the specific functionality that cannot be open-sourced and must be linked into a larger open-source platform port. The majority of the platform port must still be implemented in open source. Platform ports that are merely a thin wrapper around a binary component that contains all the actual code will not be accepted.
Only platform port code (i.e. in the
plat/<vendor>directory) may rely on binary components. Generic code must always be fully open-source.
Copyright (c) 2013-2020, Arm Limited and Contributors. All rights reserved.