Announcement: "stable" branches for multipath-tools

Announcement: "stable" branches for multipath-tools

[Martin Wilck]

All,

the multipath-tools maintainers have decided to start providing stable
branches that accumulate bug fix patches only. We currently provide two
branches, stable-0.10.y [1] and stable-0.11.y [2]. 

The stable branches are meant as a guideline for distribution
maintainers to pick bug fixes without major regression risk. Only small
patches that fix actual problems are eligible for inclusion in the
stable trees.

This is a best-effort endeavor that doesn't come with any guarantees.
The maintainers will provide maximum two stable branches at any point
in time. They will usually be based on the previous two .0 releases,
but this isn't set in stone.

The introduction of the stable branches goes along with a modified
versioning scheme for the "master" branch. The development tree is
still hosted at [3]. As before, this tree will be "released" to the
opensvc "master" branch [4] in roughly quaterly intervals, to become
the next 0.x.0 release (0.12.0 will be next).

Best regards
Martin

[1] https://github.com/opensvc/multipath-tools/tree/stable-0.10.y
[2] https://github.com/opensvc/multipath-tools/tree/stable-0.11.y
[3] https://github.com/openSUSE/multipath-tools/tree/queue
[4] https://github.com/opensvc/multipath-tools/

Re: Announcement: "stable" branches for multipath-tools

[Xose Vazquez Perez]

On 2/10/25 10:25 PM, Martin Wilck wrote:

> the multipath-tools maintainers have decided to start providing stable
> branches that accumulate bug fix patches only. We currently provide two
> branches, stable-0.10.y [1] and stable-0.11.y [2].
> 
> The stable branches are meant as a guideline for distribution
> maintainers to pick bug fixes without major regression risk. Only small
> patches that fix actual problems are eligible for inclusion in the
> stable trees.

hwtable patches ( https://github.com/opensvc/multipath-tools/commits/master/libmultipath/hwtable.c )
could be backported, as they are low risk.

Thanks.

Re: Announcement: "stable" branches for multipath-tools

[Martin Wilck]

On Sat, 2025-04-26 at 14:08 +0200, Xose Vazquez Perez wrote:
> On 2/10/25 10:25 PM, Martin Wilck wrote:
> 
> > the multipath-tools maintainers have decided to start providing
> > stable
> > branches that accumulate bug fix patches only. We currently provide
> > two
> > branches, stable-0.10.y [1] and stable-0.11.y [2].
> > 
> > The stable branches are meant as a guideline for distribution
> > maintainers to pick bug fixes without major regression risk. Only
> > small
> > patches that fix actual problems are eligible for inclusion in the
> > stable trees.
> 
> hwtable patches (
> https://github.com/opensvc/multipath-tools/commits/master/libmultipath/hwtable.c
>  )
> could be backported, as they are low risk.

Yes, but they aren't actual, bug fixes, either, and they change default
behavior, which I'd prefer to avoid in the stable branches.

Ben, what's your take?

Martin

Re: Announcement: "stable" branches for multipath-tools

[Benjamin Marzinski]

On Mon, Apr 28, 2025 at 03:56:30PM +0200, Martin Wilck wrote:
> On Sat, 2025-04-26 at 14:08 +0200, Xose Vazquez Perez wrote:
> > On 2/10/25 10:25 PM, Martin Wilck wrote:
> > 
> > > the multipath-tools maintainers have decided to start providing
> > > stable
> > > branches that accumulate bug fix patches only. We currently provide
> > > two
> > > branches, stable-0.10.y [1] and stable-0.11.y [2].
> > > 
> > > The stable branches are meant as a guideline for distribution
> > > maintainers to pick bug fixes without major regression risk. Only
> > > small
> > > patches that fix actual problems are eligible for inclusion in the
> > > stable trees.
> > 
> > hwtable patches (
> > https://github.com/opensvc/multipath-tools/commits/master/libmultipath/hwtable.c
> >  )
> > could be backported, as they are low risk.
> 
> Yes, but they aren't actual, bug fixes, either, and they change default
> behavior, which I'd prefer to avoid in the stable branches.
> 
> Ben, what's your take?

I think you're right. Even if we're just adding a new built-in config,
that could change how someone's multipath setup is configured, and while
the defaults are the recommeneded way to configure the device, it still
makes sense to not change them in what would otherwise be just a bugfix
branch, which is designed to safe to pull into a distribution at any
time.

On the other hand, if there are cases where an existing built-in config
is actually broken for a device, it would make sense to pull those fixes
into the stable branch.

-Ben

> 
> Martin