Community
Participate
Working Groups
I'd like to propose that the EF runs its own GitLab server farm. GitLab is a great and widely spread open source tool which mostly resembles what GitHub does. It contains a lot of services, and all the services are greatly working together: * Issue Tracker (could replace Bugzilla) * Agile project management * Git + LFS * CI / DI included (Gitlab CI Runners could replace Jenkins) * Can integrate with existing services, in particular can mirror GitHub. In the future, it could be a great replacement for many other tools currently in place. And for the present it would be a great tool for those projects not wanting to use GitHub (as it is external to the EF).
We likely qualify for free, on-prem Entreprise edition. https://about.gitlab.com/solutions/open-source/program/
*** Bug 502323 has been marked as a duplicate of this bug. ***
Thanks to Denis Roy for pointing me here. Is GitLab used for issue tracking by any project of size comparable to bugs.eclipse.org? Does GitLab issues allow tracking importance? At first sight, they don't. When I searched for a ticket about that, I found one about priority instead: https://gitlab.com/gitlab-org/gitlab-ce/issues/25508 I only checked quickly, but this doesn't look good. It looks like a ticket could have multiple priority labels at the same time. Labels can be relevant, but trying to do so much with them suggests the product is still far from mature. As reported in https://gitlab.com/gitlab-org/gitlab-ce/issues/23530 there is currently no migration path from Bugzilla. So I'm afraid this request cannot be fulfilled without a colossal effort involving important development at this time. That being said, I very much agree that Bugzilla should be dropped as soon as possible. The other candidate I see is Redmine, which is implemented in Ruby and mature, but I don't know it enough to recommend more than an evaluation.
(In reply to Filipus Klutiero from comment #3) > Thanks to Denis Roy for pointing me here. > > Is GitLab used for issue tracking by any project of size comparable to > bugs.eclipse.org? > > Does GitLab issues allow tracking importance? At first sight, they don't. > When I searched for a ticket about that, I found one about priority instead: > https://gitlab.com/gitlab-org/gitlab-ce/issues/25508 > I only checked quickly, but this doesn't look good. It looks like a ticket > could have multiple priority labels at the same time. Labels can be > relevant, but trying to do so much with them suggests the product is still > far from mature. > > As reported in https://gitlab.com/gitlab-org/gitlab-ce/issues/23530 there is > currently no migration path from Bugzilla. So I'm afraid this request cannot > be fulfilled without a colossal effort involving important development at > this time. > > > That being said, I very much agree that Bugzilla should be dropped as soon > as possible. The other candidate I see is Redmine, which is implemented in > Ruby and mature, but I don't know it enough to recommend more than an > evaluation. See https://about.gitlab.com/customers/. Are Cloud Native Foundation, CERN, ESA, SIEMANS and Goldman Sachs big enough to be comparable? The software itself is also available as cloud-hosted offering, with thousands of projects, so I doubt it will fail at the EF. If you have ever used Gitlab and Redmine, you won't touch Redmine anymore. It is just crap and way obsoleted by Gitlab. BTW, there *are* migration paths from Bugzilla, and the needed development is *not* "colossal".
(In reply to Markus Karg from comment #4) > (In reply to Filipus Klutiero from comment #3) > > Thanks to Denis Roy for pointing me here. > > > > Is GitLab used for issue tracking by any project of size comparable to > > bugs.eclipse.org? > > > > Does GitLab issues allow tracking importance? At first sight, they don't. > > When I searched for a ticket about that, I found one about priority instead: > > https://gitlab.com/gitlab-org/gitlab-ce/issues/25508 > > I only checked quickly, but this doesn't look good. It looks like a ticket > > could have multiple priority labels at the same time. Labels can be > > relevant, but trying to do so much with them suggests the product is still > > far from mature. > > > > As reported in https://gitlab.com/gitlab-org/gitlab-ce/issues/23530 there is > > currently no migration path from Bugzilla. So I'm afraid this request cannot > > be fulfilled without a colossal effort involving important development at > > this time. > > > > > > That being said, I very much agree that Bugzilla should be dropped as soon > > as possible. The other candidate I see is Redmine, which is implemented in > > Ruby and mature, but I don't know it enough to recommend more than an > > evaluation. > > See https://about.gitlab.com/customers/. Are Cloud Native Foundation, CERN, > ESA, SIEMANS and Goldman Sachs big enough to be comparable? I don't know them all, but assuming you mean SIEMENS, these are organizations, not projects. Even then, I looked at 2 case studies to get an idea, and I don't see a word about issue tracking in ESA's or CERN's. For all I've seen on these pages, these organizations may not have any GitLab ticket. > The software > itself is also available as cloud-hosted offering, with thousands of > projects, so I doubt it will fail at the EF. But how many tickets does it have? > If you have ever used Gitlab and Redmine, you won't touch Redmine anymore. > It is just crap and way obsoleted by Gitlab. I've used both a bit. > BTW, there *are* migration paths from Bugzilla, and the needed development > is *not* "colossal". No one claimed the development needed was colossal, but thanks. If you are referring to bugzilla2gitlab, it would help to see a demonstration. The ticket documents that it features "Mappings to set how Bugzilla components are translated into GitLab labels". In other words, GitLab Issues seems to be so labelist it doesn't have any equivalent to components other than labels. To make this manageable, Diego Giovane Pasqualin requested the ability to group labels of similar purposes: https://gitlab.com/gitlab-org/gitlab-ce/issues/21902 There is both good and bad in this request's handling. It was recently fulfilled with what GitLab calls "scoped labels": https://docs.gitlab.com/ee/user/project/labels.html#scoped-labels-premium But only in GitLab Enterprise Edition. Nevertheless, the ticket - which was requesting the feature in Community Edition - was unceremoniously closed, which indicates frail QA from GitLab. Labels are something good, but stretching their use to replace proper fields makes the system inconvenient. bugs.eclipse.org would be much less convenient if components had to be replaced with labels (although they do have one major advantage since they do not require reporters to define the components). And what about products? What would create all of Eclipse's products? Could tickets mistakenly reported to a certain product be moved to another? The fate of the above ticket suggests no. In summary, from what I see, GitLab's Issues can be good for small projects, which have a small number of issues. They can be setup quickly, and at some point in growth, the development team can add labels to manage the number of tickets. But they're not designed to handle hundreds of thousands of issues, when statuses, importance, products and components are necessary to manage the volume.
https://gitlab-test.eclipse.org/ is up and running, and you can log in using your Eclipse Foundation account. One noteworthy thing - although we see it as a replacement for Bugzilla and Git (and others), unlike our tightly controlled Bugzilla/Gerrit, GitLab follows the fork model, similar to GitHub. This means anyone can host a fork'ed repo. In fact, anyone can create their own personal public repos, just like GitHub. I've restricted everything to "public" since we don't want any private repos. My intent is to create one GitLab Group per Eclipse Project. The GL group is like a GH organization. I am following a feature request on GitLab, that could prevent users from creating projects outside the Group scope: https://gitlab.com/gitlab-org/gitlab/issues/1112 Please kick the tires and follow-up with your comments.
(In reply to Denis Roy from comment #6) > This means anyone can host a fork'ed repo. In fact, anyone can create their > own personal public repos, just like GitHub. Did you see the workaround mentioned on the GitLab issue? I believe this would better serve the Foundation's needs, i.e. only Webmaster will be able to create repos. This is similar to how GitHub integration work, isn't it? I think it's ok to open tickets for requesting new repos - unless you see it as a burden for Webmaster and would like to delegate to projects. :)
(In reply to Gunnar Wagenknecht from comment #7) > (In reply to Denis Roy from comment #6) > > This means anyone can host a fork'ed repo. In fact, anyone can create their > > own personal public repos, just like GitHub. > > Did you see the workaround mentioned on the GitLab issue? Can you be more specific? I browsed through the comments but didn't see anything. > I believe this > would better serve the Foundation's needs, i.e. only Webmaster will be able > to create repos. This is similar to how GitHub integration work, isn't it? I > think it's ok to open tickets for requesting new repos - unless you see it > as a burden for Webmaster and would like to delegate to projects. :) I'm not sure if allowing just anyone to create a repo at Eclipse is a bad thing. One thing -- if we do entertain the idea of generic repos and official project ones, we'd definitely need a visual way of identifying them.
(In reply to Denis Roy from comment #8) > (In reply to Gunnar Wagenknecht from comment #7) > > (In reply to Denis Roy from comment #6) > > > This means anyone can host a fork'ed repo. In fact, anyone can create their > > > own personal public repos, just like GitHub. > > > > Did you see the workaround mentioned on the GitLab issue? > > Can you be more specific? I browsed through the comments but didn't see > anything. https://gitlab.com/gitlab-org/gitlab/issues/1112#note_16828012 > Right now, you can set the project limit to 0 and users won't be able to > create any projects whatsoever. As a current workaround they could use > that and have only admins create projects in groups [...] > I'm not sure if allowing just anyone to create a repo at Eclipse is a bad > thing. One thing -- if we do entertain the idea of generic repos and > official project ones, we'd definitely need a visual way of identifying them. Exactly: it is scope creep. There is a ton of additional needs and responsibilities once you start hosting arbitrary user repositories. I don't think the EF should be in the position of competing with GitHub and other free hosting services for hosting free Git repositories. Just projects. That's it. (IMO)
> > > > This means anyone can host a fork'ed repo. In fact, anyone can create their > > > > own personal public repos, just like GitHub. > > > > > > Did you see the workaround mentioned on the GitLab issue? > > > > Can you be more specific? I browsed through the comments but didn't see > > anything. > > https://gitlab.com/gitlab-org/gitlab/issues/1112#note_16828012 > > > Right now, you can set the project limit to 0 and users won't be able to > > create any projects whatsoever. As a current workaround they could use > > that and have only admins create projects in groups [...] I had that initially set to 0, but it breaks the easy fork-edit-submit model that makes Git[Hub|Lab] so easy to use, and compelling for fly-by contributions.
I do believe that this proposal is in-line with what we would want: https://gitlab.com/gitlab-org/gitlab/issues/1112#note_72515013 "For open-source projects, it would be useful for arbitrary members of the public to be able to post their changes and submit MRs against extant projects, but no more than that."
(In reply to Denis Roy from comment #10) > I had that initially set to 0, but it breaks the easy fork-edit-submit model > that makes Git[Hub|Lab] so easy to use, and compelling for fly-by > contributions. Sight. I hadn't realized that forking requires the same permissions as creating an empty/new.
Will you consider enabling login through Github? If users need an Eclipse account just to file an issue, it might be a non-started for some projects that want to keep the barrier as low as possible.
Nice, I love GitLab. We are looking to move WindowBuilder to Github but I would like to explore this route first.
Denis can you add group creation permission to the sandbox?
(In reply to Andrii Berezovskyi from comment #13) > Will you consider enabling login through Github? If users need an Eclipse > account just to file an issue, it might be a non-started for some projects > that want to keep the barrier as low as possible. Makes sense, I'll add it to the radar. Not sure what that entails but the logic is sound. > Denis can you add group creation permission to the sandbox? That is next on the list for the sandbox. In the meanwhile, I've created a repo (project) for WindowBuilder inside a group in which you are the maintainer: https://gitlab-test.eclipse.org/eclipse-windowbuilder/eclipse-windowbuilder Please feel free to kick the tires!
Overall goals and timeline are on GitLab.com, as the GL team have offered some assistance. https://gitlab.com/gitlab-org/gitlab/issues/195865
Just a quick status update. We've set up a production-ready infra in a hosted location in Switzerland. Gitlab is being set up as we speak. We are targetting the end of the month for new (to Eclipse) projects that do not expect to import/migrate issues & bugs.
https://gitlab.eclipse.org/ is up and running. I will work with the webdev team to get the ECA validation and team sync mechanisms in place. We'll soon be ready for early adopter projects who are new to the Eclipse Foundation and do not have issues/bugs trackers to import (unless done manually by the project).
Is the plan to also support GitLab CI as a build environment eventually? Our in-house experience with both (we're currently in the process of migrating everything over to GitLab CI, after using both in parallel for an extensive time) suggests that it's not just easier for projects to declare readable and clean builds, but the K8s integration is also far better and overall it seems more stable and less resource-hungry than Jenkins. Not sure about scaling it up to EclipseFdn size though...
(In reply to Carsten Reckord from comment #20) > Is the plan to also support GitLab CI as a build environment eventually? > > Our in-house experience with both (we're currently in the process of > migrating everything over to GitLab CI, after using both in parallel for an > extensive time) suggests that it's not just easier for projects to declare > readable and clean builds, but the K8s integration is also far better and > overall it seems more stable and less resource-hungry than Jenkins. Not sure > about scaling it up to EclipseFdn size though... We plan to evaluate GitLab CI by the end of the year.
(In reply to Carsten Reckord from comment #20) > Is the plan to also support GitLab CI as a build environment eventually? > > Our in-house experience with both (we're currently in the process of > migrating everything over to GitLab CI, after using both in parallel for an > extensive time) suggests that it's not just easier for projects to declare > readable and clean builds, but the K8s integration is also far better and > overall it seems more stable and less resource-hungry than Jenkins. Not sure > about scaling it up to EclipseFdn size though... Are you comparing Gitlab's internal build system with Jenkins? What does Gitlab CI use as a build engine?
(In reply to Mikaël Barbero from comment #21) > We plan to evaluate GitLab CI by the end of the year. Okay, let me know if I can help or serve as a lab rat when you do :) (In reply to Wim Jongman from comment #22) > Are you comparing Gitlab's internal build system with Jenkins? Yes > What does Gitlab CI use as a build engine? It's their own engine, in some aspects not unlike TravisCI, but overall I think easier and more intuitive, yet very powerful. It's also very tightly and very neatly integrated into Gitlab workflows, i.e. "just works". See https://docs.gitlab.com/ee/ci/ for the full documentation, and in particular the .gitlab-ci.yaml reference at https://docs.gitlab.com/ee/ci/yaml/README.html for a good impression of what your typical build might look like. I can also post an example pipeline if that's of interest.
(In reply to Denis Roy from comment #19) > We'll soon be ready for early adopter projects who are > new to the Eclipse Foundation and do not have issues/bugs trackers to import > (unless done manually by the project). I'd like to volunteer moving MPC over to GitLab. We do have existing issues in Bugzilla, but I'm fine with leaving them there for now. (If there's a decent path for me to migrate manually, I could look into that as well.) Just one question: What would the contribution workflow for non-committers look like then? Is ECA validation for merge requests already in place, or should we validate manually for now, or stay with Gerrit?
I'd like to migrate the repository of Eclipse CHESS [1] to GitLab. [1] https://projects.eclipse.org/projects/polarsys.chess
> Just one question: What would the contribution workflow for non-committers > look like then? Is ECA validation for merge requests already in place, or > should we validate manually for now, or stay with Gerrit? Great question! The Webdev team has implemented the same ECA and commit check mechanism that is availble today on Gerrit and Github. The ECA check will be installed today. No manual checks for you to do.
MPC and Chess should be on GitLab by tomorrow. I'll notify via this bug. Any other early adopters?
I'd like to migrate the repositories of Eclipse Hawk too: https://projects.eclipse.org/projects/modeling.hawk If possible, both the website and regular source code repositories.
We would also like to migrate our project to the eclipse gitlab instance, if possible: https://projects.eclipse.org/projects/technology.set Our project is in the phase of the initial code contribution, therefore we don't have any existing files in our repo and also no bugtracker whatsoever. Would be good to start our work at gitlab!
(In reply to Denis Roy from comment #27) > MPC and Chess should be on GitLab by tomorrow. I'll notify via this bug. > > Any other early adopters? Thanks Denis. Is the migration of CHESS on GitLab completed?
I'll likely be doing early adopter migrations in about 5 hours!
The CHESS project is migrated (https://gitlab.eclipse.org/eclipse/chess), but please do not use it yet. We have a small snag with the Project Management Infra (PMI) @ projects.eclipse.org and Gitlab. I'll ping back once it's resolved, then migrate the others in the queue (MPC, Hawk, set).
Thanks for everyone's patience -- we'll have this sorted out very soon.
Hawk is migrated! https://gitlab.eclipse.org/eclipse/hawk/hawk Hawk team: please file any Gitlab or migration issues directly on Gitlab: https://gitlab.eclipse.org/eclipse-foundation/gitlab/-/issues
MPC is migrated! https://gitlab.eclipse.org/eclipse/mpc/org.eclipse.epp.mpc MPC team: please file any Gitlab or migration issues directly on Gitlab: https://gitlab.eclipse.org/eclipse-foundation/gitlab/-/issues
Chess is migrated! https://gitlab.eclipse.org/eclipse/chess/chess Chess team: please file any Gitlab or migration issues directly on Gitlab: https://gitlab.eclipse.org/eclipse-foundation/gitlab/-/issues
(In reply to Joachim Bleidiessel from comment #29) > We would also like to migrate our project to the eclipse gitlab instance, if > possible: > https://projects.eclipse.org/projects/technology.set > > Our project is in the phase of the initial code contribution, therefore we > don't have any existing files in our repo and also no bugtracker whatsoever. > > Would be good to start our work at gitlab! Your repo is here: https://gitlab.eclipse.org/eclipse/set/set I will update the set project provisioning bug
I'm marking this as FIXED. All Gitlab-related issues or enhancements should be filed here: https://gitlab.eclipse.org/eclipsefdn/gitlab/-/issues
I've started a Gitlab quick-start guide, for those of us who are familiar with Bugzilla: https://wiki.eclipse.org/Gitlab
Hi! I've read below comments and FAQ: https://www.eclipse.org/europe/faq.php. We have 1 project candidate that we would like to transfer to https://gitlab.eclipse.org/: https://github.com/eclipse/iceoryx (and after that we would propose some more autonomous driving projects). I have reviewed projects in https://gitlab.eclipse.org/eclipse and they do not seem to be much alive. Are the plans wrt the Eclipse own Gitlab instance more concrete? We find gitlab the best development platform but we would need these features (: 1. CI/CD 2. Container registry (which I see is disabled on gitlab.eclipse.org: https://bugs.eclipse.org/bugs/show_bug.cgi?id=565468) 3. Pages 4. Code quality, code testing and coverage 5. Merge trains --- If the plans with the self-hosted https://gitlab.eclipse.org aren't ready yet, is it possible to host an eclipse project on gitlab.com?
> We have 1 project candidate that we would like to transfer to > https://gitlab.eclipse.org/: https://github.com/eclipse/iceoryx (and after > that we would propose some more autonomous driving projects). That is great! > > I have reviewed projects in https://gitlab.eclipse.org/eclipse and they do > not seem to be much alive. We're still in early stages and have not publicized the availability much. > Are the plans wrt the Eclipse own Gitlab instance more concrete? Everything is always a Work In Progress and subject to improvement. As it is, the instance is concrete. > We find > gitlab the best development platform but we would need these features (: > 1. CI/CD We will not be opening up CI/CD at this time until we fully understand how it differentiates from our in-house Jenkins instance (Jiro) at https://ci.eclipse.org/ Unless I'm not understanding correctly, our Jiro is more expandable and flexible than many other offers. > 2. Container registry (which I see is disabled on gitlab.eclipse.org: > https://bugs.eclipse.org/bugs/show_bug.cgi?id=565468) The best way to get that functionality enabled would be to reopen the bug and state your needs :) > 3. Pages We are tracking that here: https://gitlab.eclipse.org/eclipsefdn/gitlab/-/issues/2 > 4. Code quality, code testing and coverage Jenkins CI at https://ci.eclipse.org can integrate with SonarCloud and possibly other tools. > 5. Merge trains I'm not sure I know what that is. > If the plans with the self-hosted https://gitlab.eclipse.org aren't ready > yet, is it possible to host an eclipse project on gitlab.com? That is not a possibility. We think the best options are to work with you, the project owners, in getting gitlab.eclipse.org as functional as possible.