GitLab

User stories

Following is the investigation on GitLab based on the user stories collected.

As a package maintainer, I’d like to be able to run my own CI jobs on distgit PRs.
It is possible to run CI jobs (https://docs.gitlab.com/ee/ci/) on GitLab on certain events (like pushes or pull request creation). For specific purposes where a custom environment is required, even custom GitLab runners (https://docs.gitlab.com/runner/) can be associated with a certain project repository or project namespace.
As a package maintainer, I’d like to be able to control the-new-hotness monitoring status from distgit.
For this to work on GitLab, we would need to have metadata stored in repository. This could be for example stored in .metadata file. Anybody with commit rights to repository will be able to change the monitoring settings.

Or we can leverage GitLab badges feature (https://docs.gitlab.com/17.2/ee/user/project/badges.html) which could be accessed by the-new-hotness through API and the maintainer can set the correct monitoring badge for the project.
As a provenpackager, I should have push and merge access to all packages in the Fedora package collection.
GitLab allows for a variety of roles (eg. guest, reporter, developer, maintainer, owner etc.) and binding of user accounts with these groups can ensure that the user accounts are provided with necessary permissions (https://docs.gitlab.com/ee/user/permissions.html).

In this specific case, the proven packager group can be provided with the maintainer-level access so that they can have push and merge access to all packages available in the Fedora Project package repository while the releng-scm-bot can be provided with the owner-level access.
As a package maintainer and Fedora tooling maintainer, I’d like to be able to use Packit in both distgit and upstream repositories.
Packit has an experimental support for GitLab as long as the GitLab instance is publicly available and Packit is made aware about the identity of the instance by manually configuring webhooks at GitLab end using the steps provided in the docs (https://packit.dev/docs/guide#gitlab).

The situation about the experiment support can change as soon as there are more stakeholders from Fedora Project requiring such support from the Packit’s end, thereby making it a primary requirement from not just the CentOS Stream repository workflow but also for Fedora Project.
As a package maintainer, I’d like to have easy access links to Koji, Bodhi, and Koschei from the distgit repo.
GitLab supports the use of badges (https://docs.gitlab.com/ee/user/project/badges.html) that provide a unified way of presenting condensed pieces of information about the project. While these are mostly used for showing status about the pipeline, statistics about the test coverage and information about the latest release - these can also be used for keeping links for Koji, Bodhi and Koschei for the said project.

Take, for instance, these badges can be configured automatically whenever the repository is created for the first time by the releng-scm-bot either in the repository level or as a part of textual links in the README.md file in the project repository. Alternatively, the URL field for the project can be configured to store the Fedora Packages URL (https://packages.fedoraproject.org/).
As a packager sponsor/mentor, I’d like to help users from outside the packager make their first PR without requiring special tooling or authentication schemes.
As mentioned in the answer to the user story #3, GitLab allows for a variety of roles (eg. guest, reporter, developer, maintainer, owner etc.) and it is also possible for the administrators (https://docs.gitlab.com/ee/administration/settings/visibility_and_access_controls.html) the customise the behaviours of these roles - thereby allowing for users who do not necessarily belong from the packagers group to be able to make pull requests to another project repository as long as the project repository is accessible to them.
As a package maintainer and tooling maintainer, I’d like to be able to set Depends-on: and Blocks: relationships between tickets in the issue tracker.
GitLab provides for extensive customization of labels (https://docs.gitlab.com/ee/user/project/labels.html) and interweaving of issue tickets (https://docs.gitlab.com/ee/user/project/issues/related_issues.html) which can be used to denote circumstances where a certain issue ticket depends on another issue ticket and where a certain issue ticket is blocked on another issue ticket.
As a package maintainer, I’d like to continue to be able to use fedpkg to fork repositories and otherwise interact with distgit.
GitLab provides with a comprehensive REST API (https://docs.gitlab.com/ee/api/rest/) that can be used to extend the GitLab instance’s functionality using external applications and services. Fedpkg (https://pagure.io/fedpkg) needs to be reworked to associate the REST API endpoint that allows for interacting with the deployed instance in accordance to the REST API schema of GitLab.
As a package maintainer, I want to be able to control all aspects of the pull request process (read comments, reply to comments, approve PRs, merge PRs, etc) via an REST API to allow creation command line tools to automate work that otherwise requires a pointy-clicky web UI.
As mentioned in the answer to the user story #8, GitLab provides with a comprehensive REST API (https://docs.gitlab.com/ee/api/rest/) that can be used to perform actions that control all aspects of a pull request process (https://docs.gitlab.com/ee/api/merge_requests.html and https://docs.gitlab.com/ee/api/merge_request_approvals.html). This REST API can be included in command line tools for automation.
As a package maintainer, I want to be able to mark a PR to automatically merge if-and-only-if a running CI pipeline has succeeded. This allows reviewing and approving a PR, and moving onto new work, without having to wait around to see if CI passes or not. The PR only needs re-visiting manually if CI failed.
Introducing Mergify! Mergify (https://mergify.com/) allows for automated merging of pull requests whenever a certain set of conditions are satisfied. The conditions for pull requests can be set as all CI jobs to complete successfully. Mergify also supports GitLab (https://docs.mergify.com/integrations/gitlab/).

Alternatively, the inbuilt Auto-merge functionality (https://docs.gitlab.com/ee/user/project/merge_requests/auto_merge.html) of GitLab can be used for merging pull requests for which all the conditions satisfy. Even this has an REST API access for circumstances where tailor-fitting is required.
As a contributor, I want to be able to see who the current maintainer(s) are for a package.
Project members associated with a certain GitLab project repository section can be found in the Project members section along with all necessary information like what level of access they have (eg. guest, reporter, developer, maintainer, owner etc.), the date of expiration of their access to the project (if set), the date of their user account creation, the date of their addition to the project repository or namespace, the date of the last access made by the user account etc. (https://docs.gitlab.com/ee/user/project/members/)
As a contributor, I want to be able to request to join the maintainer(s) for a package, either as a watcher (just notified of all PRs), or as a developer (participate in MR process), or as a maintainer (control all aspects of the package, ie user ACLs). Joining as merely a watcher should not require approval by current maintainers.
As mentioned in the answer to the user story #3, GitLab allows for a variety of roles (eg. guest, reporter, developer, maintainer, owner etc.) and it is possible for external collaborators to request for and for namespace/project collaborators to provide with these accesses as long as they are allowed to perform such actions.

The notification settings can be configured per repository without needing a prior permission of the owner or maintainer of the namespace/project to get informed about the activities taking place in a certain namespace/project repository (https://docs.gitlab.com/ee/user/profile/notifications.html).
As a package maintainer, I want opening a MR to automatically create a scratch-build. Approving the MR, should be able to either automatically promote the scratch-build to non-scratch status, or create a new non-scratch build. The latter should have a way to disable for packages with huge build times where creation of one build for multiple approved MRs is desirable for efficiency.
Creation of a scratch builds based on the branch from which a pull request was created is possible with the use of GitLab CI. Furthermore, the approval of a pull request can be associated with a GitLab webhook (https://docs.gitlab.com/ee/user/project/integrations/webhooks.html) which in turn should trigger the normal builds. Although, conditional cancelling of jobs due to them being overriden by a more recent job is something that needs to be configured using the GitLab CI configuration but it is possible (https://docs.gitlab.com/ee/ci/yaml/).
As a package maintainer, I would like to see the dependencies of this package, and what packages that are required by this package on web UI.
While this is a feature that can be created with the use of some intelligent parsing of the output taken from MDAPI (https://mdapi.fedoraproject.org/), the more elegant version of this implementation is unfortunately behind a paywall (https://docs.gitlab.com/ee/user/application_security/dependency_list/).

Focussing on making lemonades out of whatever lemons we have at hand, the endpoints PKG (eg. https://mdapi.fedoraproject.org/rawhide/pkg/nano) and REQUIRES (eg. https://mdapi.fedoraproject.org/rawhide/requires/nano) can be used to automatically populate the packages required by the project’s package and the packages requiring the project’s package respectively in custom files when the associated package’s repositories are being created for the first time.
As a contributor, I want to be able to see who the current maintainer(s) are for a package.
Please check the answer to the user story #11.
As a contributor, I want to be able to see the contributors that have contributed to this package.
Please check the answer to the user story #11.
As a packager I want to be able to have commit access without being added to the bugzilla watch list.
Please check the answer to the user story #12.
As a community member I want the ability to be added to the bugzilla watch list for a package without having commits or being in the packager group.
Please check the answer to the user story #12.
As a Fedora Infrastructure member, I want to ensure people who sign in to watch a package on bugzilla have a valid bugzilla account associated with their FAS account.
As we plan on deprecating Bugzilla going forward for its use in the packaging workflow and introducing the ticketing mechanism associated with the git forge solution that we choose as a replacement for the same, the notification settings pertaining to certain users can be configured per repository (https://docs.gitlab.com/ee/user/profile/notifications.html) as long as they have an account in the said git forge solution.
As a user of Fedora, not necessarily a member of the Fedora Project, I want to receive information about Fedora packages in the form of readable HTML documents, not in the form of programs (in Javascript or Web Assembly) that I must execute before I get to see the information.
I am afraid Dist Git might not be the correct place for this requirement. Please defer to using Fedora Packages (https://packages.fedoraproject.org/) for obtaining information about packages available in the official Fedora Project repositories. We do not want to duplicate efforts by reproducing features in Dist Git that are already implemented in an associated project in Fedora Infrastructure.
As a security-conscious packager, I want to interact with the web interface without downloading and running programs from a bunch of third parties. I should at most need to whitelist Javascript and Web Assembly from fedoraproject.org only.
The set of JavaScript and WebAssembly tools involved in running a managed GitLab deployment are made available from the source hosting the deployment and not anywhere else. This has been analysed by observing the network behaviour made in a browser environment against a managed GitLab deployment intended for testing.
As a member of the Free Software movement, I want all parts of the system to be Free Software. That includes both server-side software and all Javascript, Web Assembly and/or CSS that my browser will download and execute.
Licensed under the permissive MIT license, GitLab Community Edition is a free and open-source software. The repository of the project codebase can be found at https://gitlab.com/rluna-gitlab/gitlab-ce. There are certain parts of proprietary code involved in the said repository but there is also a pure FOSS variant of it available at https://gitlab.com/gitlab-org/gitlab-foss/ and the actual codebase available at https://gitlab.com/gitlab-org/gitlab. I suspect that there would be certain limitations in the pure FOSS variant but I cannot yet verify the same.
As a package maintainer, I want to be able to use a pull-request-only workflow for specific packages (like packages that could break a lot of things, or packages that are security sensitive).
With the use of Protected branches mechanism (https://docs.gitlab.com/ee/user/project/protected_branches.html) of GitLab, one can ensure that changes to a certain set of branches can only be introduced using a pull request workflow.
As a package maintainer with provenpackager or scm-admin rights, I want to be able to circumvent pull-request-only requirements on packages when pushing changes (i.e. mass rebuilds).
The specifics of the which set of users are restricted by the condition mentioned and the ones that are not can be set using the information found in the answer to the user story #3. According to the answer to the user story #6, this behaviour can be further customized by the administrators to ensure that the ones with the elevated privileges are left unrestricted.
As a package maintainer, I’d like existing STI and TMT CI tests to continue working while also providing a simpler, more familiar YAML-based format.
As mentioned in the answer to the user story #1, as long as the GitLab CI is configured properly to support STI and TMT CI tests, they should work fine on GitLab like they do on Pagure - both in environments provided within the managed GitLab infrastructure and in custom GitLab runners introduced externally.
As a package maintainer, I want to be able to prevent direct push to git, and using merge requests process only.
Please check the answer to the user story #23.
As a package maintainer, I want to be able to deal with bugs directly from the forge, and be able to reassign them between projects.
This is fortunately possible in GitLab in a feature (https://docs.gitlab.com/ee/user/project/issues/managing_issues.html#bulk-move-issues), that is blissfully unrestricted across various tiers of subscription. The transfers work without the restriction of the repositories having to belong under the same namespace.
As a package maintainer, I want to be able to sync a fork with the upstream with a button click.
GitLab user interface provides a nifty button for allowing folks to update their repository forks if they were to fall out of sync (https://docs.gitlab.com/ee/user/project/repository/forking_workflow.html#from-the-ui). Alternatively, people can rely on GitLab CI as mentioned in the answer to the user story #1 to automate the process by periodically updating a certain set of branches in their fork.
As a package maintainer, I wish we had CI compatible with Github Actions.
While GitHub Actions and GitLab CI are not inherently compatible with each other, they do work in a similar manner and there are a bunch of documentation that explain how one can migrate from GitHub Actions to GitLab CI conveniently (https://docs.gitlab.com/ee/ci/migration/github_actions.html).
As a contributor, I’d like to be able to interact with git and distgit via email, namely to reply to issue and PR comments.
As mentioned in the answer to the user story #12, the notification settings can not only be used to for obtaining information via email about the activities associated to a certain namespace/project repository but also to reply to issues or pull requests via email as well - establishing a two-way communication medium.
A contributor, I’d like to be able to create repositories under a personal namespace for work related to Fedora.
Every user associated with the GitLab deployment should have a personal namespace associated where they can create their repositories and store the repositories forked from somewhere else on the same deployment.
As a package maintainer, I’d like to block branch creation outside of official ones, or remove branches outside of official ones.
GitLab provides means to name, manage and protect Git branches according to a certain configuration (https://docs.gitlab.com/ee/user/project/repository/branches/#view-branches-with-configured-protections). This feature can be used to create branch with names following a certain regular expression pattern (eg. f39, f40, rawhide etc.) to protection while other branches might or might not be protected.
As a contributor, I’d like the forge to be accessible (alt text for images, respecting whatwg standards for contrast, color blindness, aria support).
GitLab pays attention to established accessibility standards (https://docs.gitlab.com/ee/development/fe_guide/accessibility/) while the extent of support can be subjective in nature. It might be enough for a certain set of people with special needs while it might not be the same for others and hence, I propose further investigation on this from the Accessibility WG.
As a package maintainer I’d like to be able to delete a wrong file from the look aside cache.
As we are proposing the use of Git LFS (https://git-lfs.com/) for the storage of tarball archives and GitLab inherently supports it (https://docs.gitlab.com/ee/topics/git/lfs/), it should be as easy as pushing a file removal commit in order for deleting a previously added file to the lookaside cache. We understand that a similar set of changes also need to be introduced to fedpkg (https://pagure.io/fedpkg) in order to ensure that the commands associated with interacting with the lookaside cache play well with the Git LFS based solution.
As a package maintainer, I’d wish to be able to see a tree of the Depends-on and Blocks bugs relationship, and export it to JSON or any easily processable document.
As mentioned in the answer to the user story #7, the interweaving of issues is possible (https://docs.gitlab.com/ee/user/project/issues/related_issues.html) and these issue tickets can be exported out easily in the form of CSV (Comma Separated Values) files (https://docs.gitlab.com/ee/user/project/issues/csv_export.html) from the GitLab user interface as long as the user requesting for the export has at least the reporter access to the associated project namespace or repository.
As a package maintainer, I wish to have a Security tab listing known CVE, and dependencies known CVE for the package.
While this is a feature that can be implemented with the use of Renovate (https://www.mend.io/free-developer-tools/renovate/) and GitLab support for Renovate (https://docs.renovatebot.com/modules/platform/gitlab/), the more elegant version of this implementation is unfortunately behind a paywall.

Focussing on making lemonades out of whatever lemons we have at hand, Renovate can be automatically configured to help report CVEs related to the dependencies and remedial pull requests that help with upgrading from the affected version whenever the repositories are created for the first time by releng-scm-bot.
As a package maintainer, I still want to be able to use Grokmirror to mirror the Forge packages locally.
To Grokmirror’s credit (https://github.com/mricon/grokmirror), the project is open enough to allow for mirroring from and mirroring to most Git forges as long as their JSON schema for their REST API is readily available. It should be possible to mirror repositories available on the chosen Git forge replacement to people’s mirrors as long as the destination repos are empty as mentioned in Grokmirror.

Additionally, repository mirroring (https://docs.gitlab.com/ee/user/project/repository/mirror/) is a feature available across multiple tiers of GitLab with the support for various Git forges for not only the contents of the Git repository but also the pull requests and issue tickets associated with it. One can also make use of GitLab CI (https://docs.gitlab.com/ee/ci/) to set a cronjob or event driven interaction.
As an engineer involved in the release process, I want to be able to track blocker bugs and freeze exception bugs in the chosen issue tracker.
GitLab provides issue trackers (https://docs.gitlab.com/ee/user/project/issues/) associated with a certain namespace for purposes like discussing the implementation of an idea, tracking tasks and work progress, accepting feature proposals, questions, requests or bugs, elaborating on code implementations etc.

Additionally, as mentioned in the answer to the user story #7, if the interweaving of the issue tickets across various namespaces is a requirement due to cases like a certain package depending on another package - this can be facilitated by ensuring that the issue tracker associated with the parent namespace of the said packages are used.
As an engineer involved in the release process, I want to be able to properly triage the state of the blocker bugs and freeze exception bugs with Kanban boards.
GitLab provides some cool looking issue boards (https://docs.gitlab.com/ee/user/project/issue_board.html) that are blissfully unrestricted and available across various tiers. These issue boards follow the Kanban or Scrum methodologies for planning, triaging and implementing tasks associated with the namespace.

These can be mixed and matched with the use of epics (https://docs.gitlab.com/ee/user/group/epics/), milestones (https://docs.gitlab.com/ee/user/project/milestones/) and issues (https://docs.gitlab.com/ee/user/project/issues/) to ensure that the progress reporting is verbose and effective in the software engineering process.
As an engineer involved in the release process, I should be able to notify the affected package maintainers and package testers.
Please check the answer to the user story #12.
As an engineer involved in the release process, I want to be able to plan for the effects on release phases (Alpha, Beta, Final) using the Gantt chart.
While the roadmap feature (https://docs.gitlab.com/ee/user/group/roadmap/) is unfortunately behind a paywall, some intelligent use of the time tracking feature (https://docs.gitlab.com/ee/user/project/time_tracking.html) which is available across all the tiers should allow for us to have a similar effect of using a Gantt chart for tracking the various stages of the release process.
As an engineer involved in the release process, I want to be able to control the automated building and testing using comments.
As mentioned in the answer to the user story #13, the events possible with GitLab CI like the creation of scratch builds and actual buiilds can be controlled with the intelligent use of webhooks (https://docs.gitlab.com/ee/user/project/integrations/webhooks.html) that GitLab supports. These webhooks can be triggered internally on the selected git forge as a consequence of activities like commenting under a ticket or a pull request or externally with the use of projects like Webhook To Fedora Messaging (https://github.com/fedora-infra/webhook-to-fedora-messaging).
As an engineer involved in the release process, I want to be able to triage the priorities of the blocker bugs and freeze exception bugs in the chosen issue tracker.
Please check the answer to the user story #7.
As an engineer involved in the release process, I want to be able to speed up the creation of blocker bugs and freeze exception bugs with the use of issue ticket templates.
As mentioned in the answer to the user story #37, GitLab not only provides for issue trackers but also description templates (https://docs.gitlab.com/ee/user/project/description_templates.html) working on both issue ticket (https://docs.gitlab.com/ee/user/project/description_templates.html#create-an-issue-template) and pull requests (https://docs.gitlab.com/ee/user/project/description_templates.html#create-a-merge-request-template) that are available across all tiers for usage.
As an engineer involved in the release process, I want to be able to limit the access of ticket modification to a certain set of people.
Please check the answer to the user story #3.
As an engineer involved in the release process, I want to be able to access the chosen issue tracker with the use of an HTTP API.
Please check the answer to the user story #7.
As an engineer involved in the release process, I want to be able to show related issue tickets to the reporter to avoid repeated rejections.
GitLab shows related or similar issue tickets to the issue reporter based on the textual analysis (https://docs.gitlab.com/ee/user/project/issues/create_issues.html) of the issue content typed so far (both, head and description) in realtime without having to create the issue ticket first. This helps to ensure that the issue reporters are discouraged from reporting similar issue tickets if the same has been created before and instead they can be redirected to one of the existing ones to chime in.
As an engineer involved in the release process, I want to be able to use the already established Bug Status Workflow as closely as possible.
Please check the answer to the user story #7.
As an engineer involved in the release process, I want to be able to automate ticketing workflows according to the phases of a release.
With the use of the milestones feature (https://docs.gitlab.com/ee/user/project/milestones/) of GitLab which is available across various tiers, it is possible to control the behaviour of the namespace based on an event associated with a milestone. Actions like creating, editing, closing and deleting a milestone have effect on the namespace and can be tailored to ensure that they meet the requirements intended for the namespace.
As an engineer involved in the release process, I want to be able to track accepted previous release blocker bugs in the chosen issue tracker.
Please check the answer to the user story #7.
As a consumer of the fedora messaging bus, I want to be able to continue to receive messages in a JSON format on events.
With the introduction of the Webhook To Fedora Messaging (https://github.com/fedora-infra/webhook-to-fedora-messaging/) project, webhooks (https://docs.gitlab.com/ee/user/project/integrations/webhooks.html) on the platform can be integrated with the service to stay posted about the activities at the forge end.
As a developer/release engineer/QA/contributor, we need the new dist-git service to have hooks to attach new and existing CI pipelines.
GitLab supports webhooks (https://docs.gitlab.com/ee/user/project/integrations/webhooks.html) so as long as we are able to leverage the features with our existing CI pipelines - the service should be able to function similarly to the existing Pagure based git forge.
As a contributor, I need to be able to login using my Fedora account.
GitLab supports SAML (https://docs.gitlab.com/ee/integration/saml.html) and hence, plugging in support for authentication using a Fedora Account System account should be possible. This has been implemented to various degrees of success in the fedora namespace of the GitLab.com offering and hence, the same knowledge can be made use of in this implementation. This extends to group based access controls as well which means that Fedora Account System groups can be used to map access levels at the GitLab end.
As a developer, I would like there to be a separate staging and production instance of dist-git so I can test my changes before pushing them.
Being a free and open source software, we should be able to deploy multiple instances of GitLab with each instance having its own purpose and authentication backend without any issues.
As a maintainer of dist-git(s), I would like to use industry-standard tools to maintain large files.
Please check the answer to the user story #34.
As a maintainer, I would like an easy-to-use interface to make suggestions on individual code blocks and suggest changes that the submitter can simply apply.
While I suggest rebasing and applying changes based on the review comments for a cleaner Git history, it is technically possible (https://docs.gitlab.com/ee/user/project/merge_requests/reviews/) for collaborators to apply the suggest changes with a click of a button on GitLab.
As a package maintainer, I would like to be able to automatically transfer issues between packages (for example from ansible -> ansible-core).
Please check the answer to the user story #27.
A package maintainer, I would like a push hook to block pushes to retired or EOL branches
GitLab supports Git Server hooks (https://docs.gitlab.com/ee/administration/server_hooks.html) which can be used to conditionalize the rejection of a pushed commit whenever they are made against a branch that is slated to be retired or EOLed. The presence of a documentation should make the process relatively convenient.
As a package maintainer, I want to be able to grant access to specific branches, or branches matching a pattern, to individual users. In pagure this is referred to as collaborator access.
It is possible to configure the access levels per branch dependent on certain branch names or branch names matching a certain pattern in GitLab. The feature is called “Protected Branches” (https://docs.gitlab.com/ee/user/project/repository/branches/protected.html) and is blissfully unrestricted across various tiers of subscription.
As a policy enforcer, I want to be able to “orphan” packages that are not mine.
The homepage of the said package (consisting of README.md file) can be updated to reflect the changed status of the package and the name of the package can be added to a list where potential packagers can pick it up from. While there is no 1:1 support in the feature parity, we can have an issue tracker where people can request the orphanment of the packages that they own and an automated script can be run that removes their access (https://docs.gitlab.com/ee/user/permissions.html) from the said package repository and makes the status reflecting changes.
As a developer, I want to be query packages by their maintainers and vice versa.
Packages (or in this case, their representation as project repositories) have an ability of managing members (https://docs.gitlab.com/ee/user/project/members/) associated with the project. Checking the packages associated with maintainers and checking maintainers associated with packages should be possible.
As a packager, I need the new forge to support the orphaned packages process: a maintainer should be able to automatically orphan a package they own and another packager should be able to take ownership through a self-service interface.
Extending the answer to the user story #60, a workflow can be established around the action of requesting a package orphanment to an issue tracker - that in turn should kick off some automation that ends up changing the project’s homepage to reflect the status and enlisting the package for potential packagers to “adopt”.
As a package maintainer, I need the forge to allow me to push branches that do not share history with other branches.
This is a Git technicality and as GitLab is based on Git, this should be possible.
As a driveby contributor, who is not a provenpackager (or does not want to use their provenpackager rights), I need to be able to open a Pull Request/Merge Request which introduces zero code changes (it has only empty commit(s) to bump the release for packages using %autorelease).
If by empty commits, minimal changes to the specfile is inferred - this should be possible by contributors as long as they have minimal access of being able to READ the said project repository. For more access, it is strongly advised for the contributor to be onboarded onto the process so that they can conveniently keep contributing.
As a proven packager, I want to be able to do basic repo operations from the command line. For example, I have the following in my history fedpkg fork && git remote rename zbyszek my && git push my rawhide:bin-sbin-merge, which I used maybe a hundred times when working on packages for the bin-sbin merge. I would then click on the link printed by the push command and review the diff before submitting a pull request and possibly adjust the commit message. Generalizing from this, all operations like forking, merging and closing of pull requests, should be possible via a command line tool. The tool must also be packaged in Fedora.
Please check the answer to the user story #8.
As a documentation contributor, I would love to try out new Git forge test systems. I’m agnostic to tools, so I believe I can spend time on UX and docs workflow (build, preview, issue tracking, GitHUb Action-like features and so on).
GitLab’s UX feels a lot more polished as compared to Forgejo for those who are agnostic to tools. It is crucial to note that GitLab provides an accessible (https://docs.gitlab.com/ee/development/fe_guide/accessibility/) UX for the convenience of users using screen readers and keyboard-only functionality.
As a member of the CDT team I need to be able to get notified automatically when someone opens a design request. (An email will be sent to a mailing list).
Please check the answer to the user story #3 and user story #12.
As a member of release engineering, I need to be able to enforce that all repos in dist-git disallows the rewriting of history. (There’s nuance here; it might be okay to allow rewriting on non-standard branches, but we’d need a way to confirm that none of the affected commits were ever used to perform an official build. This is probably more effort than it’s worth vs simply disallowing rewriting on any branch)
Please check the answer to the user story #23 and user story #32.
As a packager/user, I’d like to be able to easily determine which commits were used in the creation of particular Fedora builds. A preferred mechanism for this would be to have hooks in place that can automatically commit git tags (signed with an official GPG signature) indicating associated NVRs.
Certain continuous integration workflows can be associated to check for NVR changes associated with a certain commit and following that the GitLab REST API (https://docs.gitlab.com/ee/api/rest/) can be made use of to push tags to the source repository. This can help with conveniently determining which commits were used to create particular builds. This operation is involved more with the Git workflow so all kinds of git forges should be able to do this.
As a packager, I’d like to be able to use “draft builds” in merge requests that are then promoted to official builds when the MR is merged. This is to ensure that the build that was tested and approved in the MR is the same one that gets submitted as an erratum.
Using GitLab CI, the replacement git forge can be configured to kickoff mock builds whenever pull requests are created and actual builds whenever pushes are made against the primary branch of the repository. This should facilitate packagers to immediately head over to Bodhi to schedule a new release after a pull request was merged.
As a packager, I’d like the option to have packages submitted as an erratum automatically when a merge request is merged. Ideally, this would be supplemental to the above “draft builds”, just submitting that exact build to errata.
Please check the answer to the user story #70.
As a packager, I’d like to be able to provide my own test pipelines for merge requests on individual projects (supplemental to any standard tests provided and/or required by Fedora).
One can declare custom workflow configuration on GitLab CI (https://docs.gitlab.com/ee/ci/) in addition to the standard tests and builds (draft and actual) facilitated by Fedora Project. Furthermore, self proposed GitLab Runners (https://docs.gitlab.com/runner/) can also be registered to a certain project to ensure that any testing that has sensitive information or requires specialized resources are addressed suitably.
As a member of Fedora Quality, I need to be able to transparently move any ticket opened by any user for one package to another package and preserve history.
Please check the answer to the user story #27.
As a member of Fedora Quality, I would like to be able to define ticket submission form templates with custom fields. And create input validations and query all opened tickets based on those field definitions.
Extending the answer to the user story #44 and user story #47, it is possible for GitLab to implement templates for ticket submissions and show already opened issue tickets with the similar verbiage to validate if the potential submission has already been proposed. The proposer needs not search for issue ticket by themselves before opening one.
As a member of Fedora Quality, I need to be able to specify “need_info” on a specific ticket. This function should regularly ping (email/matrix or any other channels) the specified user that I am requesting information from.
It is recommended to use the assignee section of a certain issue ticket and/or to tag the people involved in the action using their username. While this might not facilitate pings to be made to a certain user at equal intervals of time, it can very well be used to keep them posted about the updates made to the issue ticket and to inform them about their action items.
As a policy enforcer, I need to be able to get all the information about a ticket/issue/bug from a Python script in a structured way.
Please check the answer to the user story #8 and user story #9.
As a policy enforcer, I need to be able to open/modify/comment on a ticket/issue/bug from a Python script (including all the possible metadata).
Please check the answer to the user story #8 and user story #9.
As a policy designer, I need to be able to add custom and structured metadata to a ticket/issue/bug, in the form of key=value pairs. I need to be able to set rules for individual keys (such as which values are valid).
While GitLab provides a custom set of fields to be set for the issue tickets, those fields are not as customizable as they are in Bugzilla. As a result, if this is a requirement people will have to rely on the issue templates (https://docs.gitlab.com/ee/user/project/description_templates.html#create-an-issue-template) to fill in custom information.
As a package maintainer, I want to have syntax highlighting for RPM spec files in the git forge web interface. Currently Pagure and Forgejo have this, but GitLab does not.
Good point! We can ask the GitLab folks nicely to add that I suppose.
As a Fedora Badges junkie, I want to be able to receive badges for activities related to dist-git. In technical terms, the dist-git activities must be broadcast to the Fedora Message Bus (or however it’s called nowadays).
Please check the answer to the user story #51.
Scalability - As a SaaS administrator, I want the system architecture to support scalability, manually or automatically based on demand, so that we can maintain optimal performance during traffic spikes and efficiently handle user growth.
Gitlab provides reference architectures which you can use when determining what level of users and interactions we require a deployment to be able to service comfortably. This is best done prior to installation as it is potentially not easy to change after the fact. If we choose to deploy gitlab within a cloud provider such as AWS we can take advantage of auto scaling features. See the reference architectures here: https://docs.gitlab.com/ee/administration/reference_architectures/5k_users.html
Reliability and High Availability - User Story: As a SaaS administrator, I want the service architecture to be highly available, to ensure 24/7 operation with minimal downtime, so that end users may access and use the application whenever needed without interruptions.
Gitlab architecture offers the ability to run in a highly available fashion. Ensure a suitable configuration is deployed at installation time. It is not easily changed later. See the reference architecture here: https://docs.gitlab.com/ee/administration/reference_architectures/5k_users.html there are a number of maintenance tasks which must be carried out regularly to maintain the instance see: https://docs.gitlab.com/ee/administration/operations/
Security - User Story: As a SaaS administrator, I want robust security measures implemented across our infrastructure, including encryption, access controls, and regular security audits, so that we can protect our end users data and maintain their trust.
Gitlab has robust security features, if installed, configured and maintained correctly, the system can be secure enough for our needs in a self managed installation. We need to ensure that we follow the best practices during the installation phase. https://docs.gitlab.com/ee/security/ for notes related to security breach see https://docs.gitlab.com/ee/security/responding_to_security_incidents.html and for recommendations to hardening the installation https://docs.gitlab.com/ee/security/hardening.html
Security - User Story: As a SaaS administrator, I want to quickly identify when the system is affected by CVEs, so that steps can be taken to plan upgrades to mitigate vulnerabilities.
Gitlab notifies when the system is affected by known CVEs, popups suggesting the system be updated show to administrators on logging into the system. See information about patching and maintenance with regards to upgrading and mitigating vulnerabilities https://docs.gitlab.com/ee/policy/maintenance.html
Monitoring and Observability - User Story: As a SaaS administrator, I want a centralised monitoring and logging system that provides real-time insights into application performance, resource utilisation, and user experience, so that we can quickly identify and resolve issues before they impact users.
Gitlab has integrations with prometheus and grafana to provide monitoring. See https://docs.gitlab.com/ee/administration/monitoring/ This can be scraped by Zabbix later as we move towards using Zabbix as the main monitoring system within Fedora Infra.
Infrastructure as Code (IaC) - User Story: As a SaaS administrator, I want to manage our the system infrastructure using code that we can version control, easily replicate environments, and automate provisioning and configuration.
Configuration will be possible to keep under source control. Depending on the method of installation we choose, there are Linux packages, Helm, Docker, Operator, source, or scripts. It will likely require that we develop configurations to add custom functionality such as integrations with FAS. https://docs.gitlab.com/ee/install/
Multi-tenancy - User Story: As a SaaS administrator, I want the application to be a secure multi-tenant system that efficiently shares resources among end users while ensuring complete data isolation, so that we can serve multiple clients cost-effectively without compromising security.
By default, Gitlab is configured for multi-tenancy. The reference architectures detail the requirements to guarantee a service level capable of supporting the desired number of users.
Disaster Recovery and Backup - User Story: As a SaaS administrator, I want automated backup systems and a comprehensive disaster recovery plan in place, so that we can quickly recover from any unforeseen events and minimise data loss and downtime.
The process for backing up and restoration of Gitlab instances is well documented see: https://docs.gitlab.com/ee/administration/backup_restore/index.html and restoring see https://docs.gitlab.com/ee/administration/backup_restore/restore_gitlab.html
Upgradability - User Story: As a SaaS administrator, I want to apply automated system upgrades, preferrably without causing system downtime, so we can continue to provide service to end users.
Gitlab can be upgraded, but not automated, and probably not without downtime or degraded performace. See the guide depending on the installation method we go with: https://docs.gitlab.com/ee/update/?tab=Helm+chart+%28Kubernetes%29#upgrade-based-on-installation-method
Upgradability - User Story: As a SaaS administrator, I want to safely apply database schema upgrades without causing outages or data loss.
The Gitlab system can be upgraded safely, but not automatically, see the upgrade guide for more information depending on the method of installation we choose https://docs.gitlab.com/ee/update/?tab=Helm+chart+%28Kubernetes%29#upgrade-based-on-installation-method and migrations also likely required to be carried out see https://docs.gitlab.com/ee/update/background_migrations.html
As a maintainer of packages in Fedora, I would like the ability to move a ticket assigned to my project to a new project. As an example, . I get a few tickets every month against “fedora-release” because people just use that as a stand-in for “I don’t know what package to report against”
Please check the answer to the user story #27.
As a Fedora CI maintainer, I would like to have the ability to enable mandatory checks for each dist-git pull request to ensure Fedora distribution stability.
Please check the answer to the user story #10.
As a Fedora forge administrator, I would like to have legally-enforceable assurances that features available in the open-source (or “community edition”) of the forge will remain that way in the future in order to prevent platform decay from degrading the functionality.
GitLab has versioned documentation, the archives https://docs.gitlab.com/archives/ of which go as far back as GitLab 10.3. These documentation draw a line between what is provided in the free-of-cost community edition and what can be made available using the ultimate edition. As these documents can be made available to be downloaded locally to as a container, we can be very sure that the if there are any platform decay that takes place. Although, it is very likely not legally enforceable and should they do something like this, this might end up causing a backlash in the community but with not much power on our hand to be able to reverse such a change.