Is Fixes line enough?

Is Fixes line enough?

[Larry Finger]

Greg,

A question has come up on the linux-wireless ML. If a patch has a "Fixes" line, 
is that sufficient to get it flagged as a patch in Stable, or is a "Cc: Stable" 
line also needed?

Thanks,

Larry

Re: Is Fixes line enough?

[Greg Kroah-Hartman]

On Tue, Oct 23, 2018 at 09:05:20AM -0500, Larry Finger wrote:
> Greg,
> 
> A question has come up on the linux-wireless ML. If a patch has a "Fixes"
> line, is that sufficient to get it flagged as a patch in Stable, or is a
> "Cc: Stable" line also needed?

No, a "Fixes" line is not sufficient enough, stable@ will get the
attention of my scripts properly and will always be looked at.
Sometimes I look at the fixes: tag, if I have nothing else to do (not
something that's happened in the past few months), and sometimes Sasha's
scripts will pick up on that tag, but not always.

So to be sure, always use cc: stable@ like the stable_kernel_rules.txt
file says to do.

thanks,

greg k-h

Re: Is Fixes line enough?

[Theodore Y. Ts'o]

On Tue, Oct 23, 2018 at 09:05:20AM -0500, Larry Finger wrote:
> Greg,
> 
> A question has come up on the linux-wireless ML. If a patch has a "Fixes"
> line, is that sufficient to get it flagged as a patch in Stable, or is a
> "Cc: Stable" line also needed?

Someone actually asked this question at the Maintainer's Summit,
actually.  The answer was that the Fixes line is not sufficient for
Greg's scripts; you have to have the "Cc: stable" line as well.  Greg
tried using the Fixes line as a trigger, but there were too many cases
where this pulled in commits that weren't really suitable for the
Stable kernels.  Sasha's machine-learning lash up will use the Fixes
line as a signal, but if you want to explicitly request that the patch
should be cherry-picked into Stable, you should have the "Cc: stable"
line.

(Also mentioned that at the Maintainer's Summit was that we probably
do need to take a look at the Stable documentation and see how it
needs to be updated to reflect how things are currently behind
handled.  :-)

Cheers,

						- Ted

Re: Is Fixes line enough?

[Jiri Kosina]

On Tue, 23 Oct 2018, Theodore Y. Ts'o wrote:

> > A question has come up on the linux-wireless ML. If a patch has a "Fixes"
> > line, is that sufficient to get it flagged as a patch in Stable, or is a
> > "Cc: Stable" line also needed?
> 
> Someone actually asked this question at the Maintainer's Summit,
> actually.  The answer was that the Fixes line is not sufficient for
> Greg's scripts; you have to have the "Cc: stable" line as well.  Greg
> tried using the Fixes line as a trigger, but there were too many cases
> where this pulled in commits that weren't really suitable for the
> Stable kernels.  Sasha's machine-learning lash up will use the Fixes
> line as a signal, but if you want to explicitly request that the patch
> should be cherry-picked into Stable, you should have the "Cc: stable"
> line.

FWIW, I brought this up already at KS 2016, see Jon's coverage here:

	https://lwn.net/Articles/705220/

My primary motivation to bring that up back then was to try to reduce the 
number of patches that are taken into -stable while there is no good 
justification for that (by requiring each and every of those having Fixes: 
present as a requirement), but it didn't really lead anywhere.

-- 
Jiri Kosina
SUSE Labs

Re: Is Fixes line enough?

[Theodore Y. Ts'o]

On Tue, Oct 23, 2018 at 06:36:26PM +0200, Jiri Kosina wrote:
> FWIW, I brought this up already at KS 2016, see Jon's coverage here:
> 
> 	https://lwn.net/Articles/705220/
> 
> My primary motivation to bring that up back then was to try to reduce the 
> number of patches that are taken into -stable while there is no good 
> justification for that (by requiring each and every of those having Fixes: 
> present as a requirement), but it didn't really lead anywhere.

Ah, I didn't get that you were trying to suggest that things only go
into stable if it has both Fixes: *and* Cc: Stable.

If that's the problem you were trying to solve, perhaps we could ask
Stephen Rothwell if he would be willing to run a script that sends
nag-o-grams to Maintainers who incluce patches in linux-next that have
Cc: stable but neither Fixes nor a "# 4.x" appended to the end of the
Cc: stable line?

						- Ted

Re: Is Fixes line enough?

[Dmitry Torokhov]

On Tue, Oct 23, 2018 at 10:38 AM Theodore Y. Ts'o <tytso@mit.edu> wrote:
>
> On Tue, Oct 23, 2018 at 06:36:26PM +0200, Jiri Kosina wrote:
> > FWIW, I brought this up already at KS 2016, see Jon's coverage here:
> >
> >       https://lwn.net/Articles/705220/
> >
> > My primary motivation to bring that up back then was to try to reduce the
> > number of patches that are taken into -stable while there is no good
> > justification for that (by requiring each and every of those having Fixes:
> > present as a requirement), but it didn't really lead anywhere.
>
> Ah, I didn't get that you were trying to suggest that things only go
> into stable if it has both Fixes: *and* Cc: Stable.
>
> If that's the problem you were trying to solve, perhaps we could ask
> Stephen Rothwell if he would be willing to run a script that sends
> nag-o-grams to Maintainers who incluce patches in linux-next that have
> Cc: stable but neither Fixes nor a "# 4.x" appended to the end of the
> Cc: stable line?
>

Patches adding new PCI/USB/ACPI IDs or DMI quirks are usually accepted
into stable but normally lack "Fixes" tag.

Thanks.

-- 
Dmitry

Re: Is Fixes line enough?

[Jiri Kosina]

On Tue, 23 Oct 2018, Dmitry Torokhov wrote:

> > Ah, I didn't get that you were trying to suggest that things only go 
> > into stable if it has both Fixes: *and* Cc: Stable.
> >
> > If that's the problem you were trying to solve, perhaps we could ask
> > Stephen Rothwell if he would be willing to run a script that sends
> > nag-o-grams to Maintainers who incluce patches in linux-next that have
> > Cc: stable but neither Fixes nor a "# 4.x" appended to the end of the
> > Cc: stable line?
> >
> 
> Patches adding new PCI/USB/ACPI IDs or DMI quirks are usually accepted
> into stable but normally lack "Fixes" tag.

That could easily be made an easily identifiable exception to the rule 
though.

-- 
Jiri Kosina
SUSE Labs