[llvm-dev] [cfe-dev] GitHub Migration Schedule and Plans
Mehdi AMINI via llvm-dev
llvm-dev at lists.llvm.org
Thu Oct 10 17:48:48 PDT 2019
On Thu, Oct 10, 2019 at 4:22 PM Tom Stellard <tstellar at redhat.com> wrote:
> On 10/10/2019 03:50 PM, Mehdi AMINI wrote:
> >
> >
> > On Thu, Oct 10, 2019 at 1:14 PM Jordan Rupprecht <rupprecht at google.com
> <mailto:rupprecht at google.com>> wrote:
> >
> >
> >
> > On Thu, Oct 10, 2019 at 12:29 PM Tom Stellard via llvm-dev <
> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote:
> >
> > On 10/10/2019 11:40 AM, Mehdi AMINI wrote:
> > >
> > >
> > > On Thu, Oct 10, 2019 at 10:59 AM Tom Stellard <
> tstellar at redhat.com <mailto:tstellar at redhat.com> <mailto:
> tstellar at redhat.com <mailto:tstellar at redhat.com>>> wrote:
> > >
> > > On 10/09/2019 11:05 PM, Mehdi AMINI wrote:
> > > >
> > > >
> > > > On Wed, Oct 9, 2019 at 10:16 PM Tom Stellard via cfe-dev
> <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> <mailto:
> cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org>> <mailto:
> cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> <mailto:
> cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org>>>> wrote:
> > > >
> > > > Hi,
> > > >
> > > > We're less than 2 weeks away from the developer
> meeting, so I wanted to
> > > > give an update on the GitHub migration and what's
> (hopefully) going to
> > > > happen during the developer meeting.
> > > >
> > > > Everyone who has added their information to the
> github-usernames.txt
> > > > file in SVN before today should have received an
> invite to become a collaborator
> > > > on the llvm-project repository. If you did not
> receive an invite and think
> > > > you should have, please contact me off-list. I will
> continue to monitor the
> > > > file for new updates and periodically send out new
> batches of invites.
> > > >
> > > > There is still some ongoing work to get the
> buildbots ready and the mailing lists
> > > > ready, but we are optimistic that the work will be
> done in time.
> > > >
> > > > The team at GitHub has finished implementing the
> "Require Linear History"
> > > > branch protection that we requested. The feature is
> in beta and currently
> > > > enabled in the llvm-project repository. This means
> that we will have the
> > > > option to commit directly via git, in addition to
> using the git-llvm script.
> > > > A patch that updates git-llvm to push to git instead
> of svn can be found here:
> > > > https://reviews.llvm.org/D67772. You should be
> able to test it out on your
> > > > own fork of the llvm-project repository.
> > > >
> > > > The current plan is to begin the final migration
> steps on the evening (PDT)
> > > > of October 21. Here is what will happen:
> > > >
> > > > 1. Make SVN read-only.
> > > > 2. Turn-off the SVN->git update process.
> > > > 3. Commit the new git-llvm script directly to github.
> > > > 4. Grant all contributors write access to the
> repository.
> > > >
> > > >
> > > > Is the repo configured to forbid contributors to create
> new branches? I'm worried about the "jungle" it can become quickly if we
> leave open the possibility to create branches "at will" in the repo, I
> rather leave this to maintainers.
> > > >
> > >
> > > I haven't been able to find a way to restrict branch
> creation for committers,
> > > I'm not sure if this is even possible.
> > >
> > >
> > > I think you can just go to the branch settings, add a new
> branch protection rule, match on everything but master, and check "Restrict
> who can push to matching branches".
> > >
> >
> > I tried this, and the branch protections only come into effect
> after a branch
> > has been creating, so this doesn't prevent new branches. It's
> actually worse
> > than doing nothing, because once the branch is created the
> branch protection
> > prevents you from deleting it.
> >
> >
> > Ah, interesting :)
> > Since for now `git-llvm` is necessary to push to the repo, I guess
> creating a branch won't even really be possible? So when adding the support
> for branch creation to `git-llvm` we could print a disclaimer about it or
> something?
> >
>
> With my latest changes in https://reviews.llvm.org/D67772 it is actually
> possible
> to create a new branch with git-llvm, so we could print a disclaimer or
> block this,
> however we still can't stop people from creating a new branch using the
> normal git command.
>
Oh, of course, because the same branch protection we use for master can't
apply to branches before they are created in the repo.
Thanks,
--
Mehdi
>
> -Tom
>
> >
> >
> > -Tom
> >
> >
> > FWIW, we're interested in periodically (weekly) tagging
> well-tested/stable revisions, but via a branch instead of just a tag so we
> can include which cherrypicks (e.g. reverts or fixes) are needed. We do
> this with the current svn repo so we'd just be porting existing
> functionality to github.
> >
> >
> > I would likely include this in the general category of "project
> maintainer" created branch?
> > The main difference with SVN, is that on GitHub a fork is "cheap" and
> everyone can have their own. Also compared to SVN you don't really know
> ahead of time when you type `git clone` or `git pull` that you may pull a
> lot of noise.
> > I'm not against branches for the sake of it, I just have some bad
> memories from a repo where every developer was pushing their work in
> progress branches all the time and they never got deleted: to the point
> where some git CLI auto-completion was unusable (there were >10k branches
> and probably as many tags).
> >
> >
> >
> > (Also, I'm not sure this announcement thread with all the *-dev
> lists is the best place to discuss branching policy, but I wanted to get
> this bit in since y'all brought it up :) )
> >
> >
> > (we discussed branches in past GitHub migration thread, we just never
> had a dedicated email thread to user-branches, as it didn't seem like there
> was a strong need for it)
> >
> >
> >
> >
> >
> > >
> > >
> > > We could try to enforce this rule in the git-llvm script,
> but this would
> > > mean making use of the script mandatory, which was our
> original plan, but
> > > that was based on the assumption that the "Require Linear
> History"
> > > protection would not be ready in time.
> > >
> > > Generally, would it be better if we kept use of the script
> mandatory so that
> > > we can handle this and other potential restrictions in the
> future?
> > >
> > > - Tom
> > > >
> > > >
> > > > 5. Email lists announcing that the migration is
> complete.
> > > >
> > > > Once the migration is complete, if you run into any
> issues, please file
> > > > a bug, and mark it as a blocker for the github
> metabug PR39393.
> > > >
> > > > If you have any questions or think I am missing
> something, please
> > > > let me know.
> > > >
> > > >
> > > > This is fantastic! Thank you so much for doing this work
> Tom :)
> > > >
> > > > --
> > > > Mehdi
> > > >
> > > >
> > >
> >
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20191010/2b1ca841/attachment-0001.html>
More information about the llvm-dev
mailing list