core-updates invites to an ungrafting party

core-updates invites to an ungrafting party

[Maxim Cournoyer]

Hello Guix!

The core-updates branch is still alive, and has accumulated (or plans
to) a few changes that cause world-rebuilds, such as fixes to
git-minimal (bug#65924) as well as docbook improvements (bug#65479) and
fixes to the build systems so that deep input rewriting works as
intended (bug#65665).

I think we could also batch ungrafting of all grafted packages, to make
the most out of this complete rebuild.

To recall, the policy surrounding what goes to core-updates is still
unchanged (per the Contributing section of our manual), except for areas
covered by teams (which is still patchy at best -- have you considered
joining teams?)

What do you think?  If you are interested in participating in the
effort, you can send your ungrafting patches for review with the
--subject-prefix='PATCH core-updates' prefix or if you are a committer
you could simple version bumps to core tools that have been posted to
guix-patches, if any.

-- 
Thanks,
Maxim

Re: core-updates invites to an ungrafting party

[John Kehayias]

Hi Maxim et al,

On Sun, Oct 08, 2023 at 11:12 AM, Maxim Cournoyer wrote:

> Hello Guix!
>
> The core-updates branch is still alive, and has accumulated (or plans
> to) a few changes that cause world-rebuilds, such as fixes to
> git-minimal (bug#65924) as well as docbook improvements (bug#65479) and
> fixes to the build systems so that deep input rewriting works as
> intended (bug#65665).
>
> I think we could also batch ungrafting of all grafted packages, to make
> the most out of this complete rebuild.
>

That sounds good, we have suddenly got a bunch of grafts deep in the
dependency tree.

Speaking of which, I was planning to at least ungraft libx11 and
libxpm, recipients of recent grafts for security reasons, on a
forthcoming mesa-updates branch. I'm just waiting for the next point
release of mesa, since 23.2.1 is actually the first release where
typically a first .1 release is considered the start of the stable
series. (Though 23.2 has had a long release candidate time.)

So, what are we thinking of the time to build/merge core-updates? I
was hoping to do some ungrafting and updating in the mesa-related
ecosystem this week, depending on upstream.

I'll start a separate thread soon to ask for what patches to include
there that I don't already know about, but I'm happy to include
similar scope ungrafting if that makes sense before core-updates.

What does everyone think? I think it is more a question of
timing/resources, either doing some ungrafting earlier but then having
more builds again soon after (e.g. glibc ungraft), or knocking some of
it out earlier with a smaller scope.

> To recall, the policy surrounding what goes to core-updates is still
> unchanged (per the Contributing section of our manual), except for areas
> covered by teams (which is still patchy at best -- have you considered
> joining teams?)
>

And thanks for pointing this out. I do hope we continue building teams
and scopes for them so core-updates doesn't end up getting too
unwieldy. I'm optimistic of a quicker merge timeline here as well, the
ungrafting being a nice immediate reason to do this.

> What do you think?  If you are interested in participating in the
> effort, you can send your ungrafting patches for review with the
> --subject-prefix='PATCH core-updates' prefix or if you are a committer
> you could simple version bumps to core tools that have been posted to
> guix-patches, if any.

Thanks Maxim for getting things rolling here!

John

Re: core-updates invites to an ungrafting party

[Maxim Cournoyer]

Hi John,

John Kehayias <john.kehayias@protonmail.com> writes:

> Hi Maxim et al,
>
> On Sun, Oct 08, 2023 at 11:12 AM, Maxim Cournoyer wrote:
>
>> Hello Guix!
>>
>> The core-updates branch is still alive, and has accumulated (or plans
>> to) a few changes that cause world-rebuilds, such as fixes to
>> git-minimal (bug#65924) as well as docbook improvements (bug#65479) and
>> fixes to the build systems so that deep input rewriting works as
>> intended (bug#65665).
>>
>> I think we could also batch ungrafting of all grafted packages, to make
>> the most out of this complete rebuild.
>>
>
> That sounds good, we have suddenly got a bunch of grafts deep in the
> dependency tree.
>
> Speaking of which, I was planning to at least ungraft libx11 and
> libxpm, recipients of recent grafts for security reasons, on a
> forthcoming mesa-updates branch. I'm just waiting for the next point
> release of mesa, since 23.2.1 is actually the first release where
> typically a first .1 release is considered the start of the stable
> series. (Though 23.2 has had a long release candidate time.)
>
> So, what are we thinking of the time to build/merge core-updates? I
> was hoping to do some ungrafting and updating in the mesa-related
> ecosystem this week, depending on upstream.

I should be able to drive this merge in a speedy manner, since I have a
lot of time at the moment.  I'm hopeful it could be merged into master
in a month time (so, approximately mid-November).

> I'll start a separate thread soon to ask for what patches to include
> there that I don't already know about, but I'm happy to include
> similar scope ungrafting if that makes sense before core-updates.

To keep things easy to follow and avoid duplicating efforts, I'd keep
most ungrafting to core-updates, unless those pertaining to other teams
such as mesa.

-- 
Thanks,
Maxim

Re: core-updates invites to an ungrafting party

[Simon Tournier]

Hi,

On Sun, 08 Oct 2023 at 11:12, Maxim Cournoyer <maxim.cournoyer@gmail.com> wrote:


>                                              such as fixes to
> git-minimal (bug#65924)

Why not also update git-minimal/pinned?

Well, the idea with git-minimal/pinned (as well as with all /pinned
packages) is to have a variant that is barely updated – barely means a
core-updates round, I guess. :-) It avoids a world rebuild and allows to
update non-pinned variants more frequently.

Somehow, the update of git-minimal should not be a world rebuild. :-)

--8<---------------cut here---------------start------------->8---
$ guix refresh -l -e '(@@ (gnu packages version-control) git-minimal)' | cut -f1 -d':'
Building the following 152 packages would ensure 370 dependent packages are rebuilt

$ guix refresh -l -e '(@@ (gnu packages version-control) git-minimal/pinned)' | cut -f1 -d':'
Building the following 2200 packages would ensure 5224 dependent packages are rebuilt
--8<---------------cut here---------------end--------------->8---

Although, the change in bug#65924 also impacts git-minimal/pinned, hence
maybe the world rebuild you are talking about. :-)

Anyway.  Update git-minimal/pinned too?

Cheers,
simon

Re: core-updates invites to an ungrafting party

[Maxim Cournoyer]

Hi,

Simon Tournier <zimon.toutoune@gmail.com> writes:

> Hi,
>
> On Sun, 08 Oct 2023 at 11:12, Maxim Cournoyer <maxim.cournoyer@gmail.com> wrote:
>
>
>>                                              such as fixes to
>> git-minimal (bug#65924)
>
> Why not also update git-minimal/pinned?
>
> Well, the idea with git-minimal/pinned (as well as with all /pinned
> packages) is to have a variant that is barely updated – barely means a
> core-updates round, I guess. :-) It avoids a world rebuild and allows to
> update non-pinned variants more frequently.
>
> Somehow, the update of git-minimal should not be a world rebuild. :-)
>
> $ guix refresh -l -e '(@@ (gnu packages version-control) git-minimal)' | cut -f1 -d':'
> Building the following 152 packages would ensure 370 dependent packages are rebuilt
>
> $ guix refresh -l -e '(@@ (gnu packages version-control) git-minimal/pinned)' | cut -f1 -d':'
> Building the following 2200 packages would ensure 5224 dependent packages are rebuilt
>
> Although, the change in bug#65924 also impacts git-minimal/pinned, hence
> maybe the world rebuild you are talking about. :-)
>
> Anyway.  Update git-minimal/pinned too?

It's simply been overlooked.  Would you mind sending a patch?  :-)

-- 
Thanks,
Maxim