Git reorganisation

classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Git reorganisation

David Seifert
Dear users of the sci overlay,
we've recently rearranged the git setup. The current sci setup is now
exactly like the main tree setup, namely:

1. The authoritative repo is the one hosted by infra
(git://anongit.gentoo.org/proj/sci.git)
2. All commits to the sci repo will be synced over to Github
automatically, in ONE DIRECTION only. This means all the dual HEAD
merging is obsolete now.
3. The Github repo is now meant as a (friendly) interface to potential
contributors.
4. As a new QA policy, merge commits in the overlay are banned now. The
sci overlay has much lower contention than the main repository, such
that you can realistically always avoid merge commits, even for large
batches of commits. This will require you to rebase your commits on top
of remote:

  git pull --rebase=preserve

I will likely further tighten the QA standards of the repository, due
to a history of poor COMMITMSGs and other QA violations. This is
supposed to be a testing ground for the main repo, where plans are to
also introduce such QA measures.

Furthermore, I am considering requiring full GPG-signed commits for the
 overlay, and for this I would like to get some input. I believe this
prepares contributors for eventually joining Gentoo. For low-volume
contributors not wanting to join, we can always merge pull requests
from Github. Ideas? Are you opposed to this?

Regards
David

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

François Bissey-4

> On 23/05/2017, at 19:11, David Seifert <[hidden email]> wrote:
>
> Furthermore, I am considering requiring full GPG-signed commits for the
> overlay, and for this I would like to get some input.

Weren’t we already doing this?

François
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

Guilherme Amadio-2
In reply to this post by David Seifert
Hi David,

On Tue, May 23, 2017 at 09:11:03AM +0200, David Seifert wrote:

> Dear users of the sci overlay,
> we've recently rearranged the git setup. The current sci setup is now
> exactly like the main tree setup, namely:
>
> 1. The authoritative repo is the one hosted by infra
> (git://anongit.gentoo.org/proj/sci.git)
> 2. All commits to the sci repo will be synced over to Github
> automatically, in ONE DIRECTION only. This means all the dual HEAD
> merging is obsolete now.
> 3. The Github repo is now meant as a (friendly) interface to potential
> contributors.
> 4. As a new QA policy, merge commits in the overlay are banned now. The
> sci overlay has much lower contention than the main repository, such
> that you can realistically always avoid merge commits, even for large
> batches of commits. This will require you to rebase your commits on top
> of remote:
>
>   git pull --rebase=preserve
>
> I will likely further tighten the QA standards of the repository, due
> to a history of poor COMMITMSGs and other QA violations. This is
> supposed to be a testing ground for the main repo, where plans are to
> also introduce such QA measures.
>
> Furthermore, I am considering requiring full GPG-signed commits for the
>  overlay, and for this I would like to get some input. I believe this
> prepares contributors for eventually joining Gentoo. For low-volume
> contributors not wanting to join, we can always merge pull requests
> from Github. Ideas? Are you opposed to this?

I welcome all these changes. If we can help in educating people on the
more tricky things, like signing with a GPG key, even better. I have
some ebuilds I use personally now that I will try to add in the next
few days to the overlay.

That said, once we reach good enough quality of ebuilds in the overlay,
we should start just moving them to the main tree. Gentoo is used by
quite a few physicists (myself included) and other scientists, so
eliminating the need for an extra overlay would be nice. I remember
having problems with things like blas/atlas and eselect due to
divergences with the main tree in not so distant past. Also, using
overlays with prefix is not always a seamless experience.

I'm not saying the overlay should go away, but just be a staging area
for scientific packages before they land on the main tree. What are your
thoughts on this?

Cheers,
—Guilherme

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

David Seifert
On Wed, 2017-05-24 at 09:17 +0200, Guilherme Amadio wrote:

> Hi David,
>
> On Tue, May 23, 2017 at 09:11:03AM +0200, David Seifert wrote:
> > Dear users of the sci overlay,
> > we've recently rearranged the git setup. The current sci setup is
> > now
> > exactly like the main tree setup, namely:
> >
> > 1. The authoritative repo is the one hosted by infra
> > (git://anongit.gentoo.org/proj/sci.git)
> > 2. All commits to the sci repo will be synced over to Github
> > automatically, in ONE DIRECTION only. This means all the dual HEAD
> > merging is obsolete now.
> > 3. The Github repo is now meant as a (friendly) interface to
> > potential
> > contributors.
> > 4. As a new QA policy, merge commits in the overlay are banned now.
> > The
> > sci overlay has much lower contention than the main repository,
> > such
> > that you can realistically always avoid merge commits, even for
> > large
> > batches of commits. This will require you to rebase your commits on
> > top
> > of remote:
> >
> >   git pull --rebase=preserve
> >
> > I will likely further tighten the QA standards of the repository,
> > due
> > to a history of poor COMMITMSGs and other QA violations. This is
> > supposed to be a testing ground for the main repo, where plans are
> > to
> > also introduce such QA measures.
> >
> > Furthermore, I am considering requiring full GPG-signed commits for
> > the
> >  overlay, and for this I would like to get some input. I believe
> > this
> > prepares contributors for eventually joining Gentoo. For low-volume
> > contributors not wanting to join, we can always merge pull requests
> > from Github. Ideas? Are you opposed to this?
>
> I welcome all these changes. If we can help in educating people on
> the
> more tricky things, like signing with a GPG key, even better. I have
> some ebuilds I use personally now that I will try to add in the next
> few days to the overlay.
>
> That said, once we reach good enough quality of ebuilds in the
> overlay,
> we should start just moving them to the main tree. Gentoo is used by
> quite a few physicists (myself included) and other scientists, so
> eliminating the need for an extra overlay would be nice. I remember
> having problems with things like blas/atlas and eselect due to
> divergences with the main tree in not so distant past. Also, using
> overlays with prefix is not always a seamless experience.
>
> I'm not saying the overlay should go away, but just be a staging area
> for scientific packages before they land on the main tree. What are
> your
> thoughts on this?
>
> Cheers,
> —Guilherme
>

You're putting it much too lightly - in its current state, the overlay
is a disaster. More often than not it contains awful ebuilds, awfully
broken, and noone feels a responsibility to fix the mess. People keep
adding broken ebuilds to it. So yes, people should stop adding broken
stuff to it, and more importantly, should rather send in high-quality
stuff directly to the main tree instead of using the overlay as a
quasi-permanent dumping ground.

David

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

David Seifert
In reply to this post by Guilherme Amadio-2
On Mon, 2017-05-29 at 16:08 +0200, Marius Brehler wrote:

> Hi everybody,
>
> On 05/24/2017 09:17 AM, Guilherme Amadio wrote:
> > Hi David,
> >
> > On Tue, May 23, 2017 at 09:11:03AM +0200, David Seifert wrote:
> > > Dear users of the sci overlay,
> > > we've recently rearranged the git setup. The current sci setup is
> > > now
> > > exactly like the main tree setup, namely:
> > >
> > > 1. The authoritative repo is the one hosted by infra
> > > (git://anongit.gentoo.org/proj/sci.git)
> > > 2. All commits to the sci repo will be synced over to Github
> > > automatically, in ONE DIRECTION only. This means all the dual
> > > HEAD
> > > merging is obsolete now.
> > > 3. The Github repo is now meant as a (friendly) interface to
> > > potential
> > > contributors.
> > > 4. As a new QA policy, merge commits in the overlay are banned
> > > now. The
> > > sci overlay has much lower contention than the main repository,
> > > such
> > > that you can realistically always avoid merge commits, even for
> > > large
> > > batches of commits. This will require you to rebase your commits
> > > on top
> > > of remote:
> > >
> > >   git pull --rebase=preserve
> > >
> > > I will likely further tighten the QA standards of the repository,
> > > due
> > > to a history of poor COMMITMSGs and other QA violations. This is
> > > supposed to be a testing ground for the main repo, where plans
> > > are to
> > > also introduce such QA measures.
> > >
> > > Furthermore, I am considering requiring full GPG-signed commits
> > > for the
> > >  overlay, and for this I would like to get some input. I believe
> > > this
> > > prepares contributors for eventually joining Gentoo. For low-
> > > volume
> > > contributors not wanting to join, we can always merge pull
> > > requests
> > > from Github. Ideas? Are you opposed to this?
> >
> > I welcome all these changes. If we can help in educating people on
> > the
> > more tricky things, like signing with a GPG key, even better. I
> > have
> > some ebuilds I use personally now that I will try to add in the
> > next
> > few days to the overlay.
> >
> > That said, once we reach good enough quality of ebuilds in the
> > overlay,
> > we should start just moving them to the main tree. Gentoo is used
> > by
> > quite a few physicists (myself included) and other scientists, so
> > eliminating the need for an extra overlay would be nice. I remember
> > having problems with things like blas/atlas and eselect due to
> > divergences with the main tree in not so distant past. Also, using
> > overlays with prefix is not always a seamless experience.
> >
> > I'm not saying the overlay should go away, but just be a staging
> > area
> > for scientific packages before they land on the main tree. What are
> > your
> > thoughts on this?
> >
> > Cheers,
> > —Guilherme
> >
>
> I actually agree and support the idea. However, I recently noticed
> that
> I no longer have the permission to close issues in github. Is it
> possible to get those for sci again?
> Before the announcement, I was using git.gentoo.org:proj/sci.git, for
> which I have sufficient rights. Is there any difference, or rather
> should I switch over to anongit.gentoo.org/proj/sci.git?
> Best Regards
>
> Marius

Which brings me to the next point - that thing shouldn't even be there.
The Github issue tracker was used in ancient days, and I'd much rather
people use bugzie than having 3 divergent issue trackers. I will add
you in the mean time.

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

David Seifert
In reply to this post by François Bissey-4
On Tue, 2017-05-23 at 19:26 +1200, François Bissey wrote:
> > On 23/05/2017, at 19:11, David Seifert <[hidden email]> wrote:
> >
> > Furthermore, I am considering requiring full GPG-signed commits for
> > the
> > overlay, and for this I would like to get some input.
>
> Weren’t we already doing this?
>
> François

Nope, not required. I'll talk on infra and likely start enabling this
soon, right after the COMMITMSG filter.

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

François Bissey-4

> On 3/06/2017, at 22:03, David Seifert <[hidden email]> wrote:
>
> On Tue, 2017-05-23 at 19:26 +1200, François Bissey wrote:
>>> On 23/05/2017, at 19:11, David Seifert <[hidden email]> wrote:
>>>
>>> Furthermore, I am considering requiring full GPG-signed commits for
>>> the
>>> overlay, and for this I would like to get some input.
>>
>> Weren’t we already doing this?
>>
>> François
>
> Nope, not required. I'll talk on infra and likely start enabling this
> soon, right after the COMMITMSG filter.
>

OK, now I remember. repoman forced you to do it, but just doing "git commit"
wasn’t requiring it. I changed my default settings after noticing that I didn’t
have to sign to send changes to an eclass.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Merging alternative.eclass ecosystem into main repo(Was: Git reorganisation)

Benda Xu
In reply to this post by David Seifert
Hi,

David Seifert <[hidden email]> writes:

> You're putting it much too lightly - in its current state, the overlay
> is a disaster. More often than not it contains awful ebuilds, awfully
> broken, and noone feels a responsibility to fix the mess. People keep
> adding broken ebuilds to it. So yes, people should stop adding broken
> stuff to it, and more importantly, should rather send in high-quality
> stuff directly to the main tree instead of using the overlay as a
> quasi-permanent dumping ground.

One pain in migrating ebuild to the main repo is that the
alternative.eclass ecosystem.  Is there a plan to merge it?  I am
wondering what is the blocking issue remaining.

Yours,
Benda

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Merging alternative.eclass ecosystem into main repo(Was: Git reorganisation)

François Bissey-4

> On 5/06/2017, at 18:53, Benda Xu <[hidden email]> wrote:
>
> I am
> wondering what is the blocking issue remaining.

Personally I’d like to discuss https://github.com/gentoo/sci/issues/709
before it is merged.

François
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Git reorganisation

David Seifert
In reply to this post by François Bissey-4
On Sat, 2017-06-03 at 22:06 +1200, François Bissey wrote:

> > On 3/06/2017, at 22:03, David Seifert <[hidden email]> wrote:
> >
> > On Tue, 2017-05-23 at 19:26 +1200, François Bissey wrote:
> > > > On 23/05/2017, at 19:11, David Seifert <[hidden email]> wrote:
> > > >
> > > > Furthermore, I am considering requiring full GPG-signed commits
> > > > for
> > > > the
> > > > overlay, and for this I would like to get some input.
> > >
> > > Weren’t we already doing this?
> > >
> > > François
> >
> > Nope, not required. I'll talk on infra and likely start enabling
> > this
> > soon, right after the COMMITMSG filter.
> >
>
> OK, now I remember. repoman forced you to do it, but just doing "git
> commit"
> wasn’t requiring it. I changed my default settings after noticing
> that I didn’t
> have to sign to send changes to an eclass.

I have now made signed commits and signed pushes mandatory on the
repository. To have your commits signed in the repo run

        git config commit.gpgsign true

You will also need to sign your pushes to the repo.

        git config push.gpgsign true

I will also start with adding a QA COMMITMSG filter that rejects bad
COMMITMSGs lacking the prelude: part.

Regards
David

Loading...