Would it help to encourage users to read/test stable-review patches

Would it help to encourage users to read/test stable-review patches

[Luis R. Rodriguez]

Just thinking out loud here.

We get development review eyes on stable patches during submission
into a subsystem, and perhaps 1-2 user tests then. Typically patches
are pretty easy to review for stable submission but I'm wondering
about cases where things are not so clear and perhaps we want better
testing and are less willing to take a fix into stable than on latest
rc. Also what if a user just wants to test an entire series queues up
for stable-review, is there a single all-in-one file user can just
patch -p1 to help test prior to getting the next stable release made?
Do we want more user exposure to stable-review mailing list or is just
a few developer eyes enough?

  Luis

Re: Would it help to encourage users to read/test stable-review patches

[Chris Wright]

* Luis R. Rodriguez (mcgrof@gmail.com) wrote:
> We get development review eyes on stable patches during submission
> into a subsystem, and perhaps 1-2 user tests then. Typically patches
> are pretty easy to review for stable submission but I'm wondering
> about cases where things are not so clear and perhaps we want better
> testing and are less willing to take a fix into stable than on latest
> rc.

Typically in this case it's either too risky for -stable period, or
we ask for resubmission after it has had more time upstream (through
another -rc cycle, for example).

> Also what if a user just wants to test an entire series queues up
> for stable-review, is there a single all-in-one file user can just
> patch -p1 to help test prior to getting the next stable release made?

There is a queue that can be applied w/ quilt  (or any similar tool that
can handles patches + series file).  This is kept here:

http://git.kernel.org/?p=linux/kernel/git/stable/stable-queue.git

> Do we want more user exposure to stable-review mailing list or is just
> a few developer eyes enough?

More review never hurts, but do you think there's a specific lack or
problem?  Thanks for thinking about -stable ;-)

thanks,
-chris

Re: Would it help to encourage users to read/test stable-review patches

[Luis R. Rodriguez]

On Thu, Aug 27, 2009 at 11:29 AM, Chris Wright<chrisw@sous-sol.org> wrote:
> * Luis R. Rodriguez (mcgrof@gmail.com) wrote:

>> Also what if a user just wants to test an entire series queues up
>> for stable-review, is there a single all-in-one file user can just
>> patch -p1 to help test prior to getting the next stable release made?
>
> There is a queue that can be applied w/ quilt  (or any similar tool that
> can handles patches + series file).  This is kept here:
>
> http://git.kernel.org/?p=linux/kernel/git/stable/stable-queue.git

For us non-quilters and lazy people can we get an all-in-one patch
generated from a script or something? If not I can just try to keep
one up to date myself but figured since you have a tree it may be
easier to refer the one place instead of two.

>> Do we want more user exposure to stable-review mailing list or is just
>> a few developer eyes enough?
>
> More review never hurts, but do you think there's a specific lack or
> problem?

Not really, I was thinking about the fellow Ubunteros users who tend
to not have a clue but sometimes enough good will to want to help. If
the average user understood better the upstream process then perhaps
it would become clearer to them the importance of stable patches and
stable-review. Bringing users closer to stable-review to me indicates
more users would understand the stable process and therefore would
stop bitching to me when a 'non-cirtical' fix never made it into their
ancient stable. Cc'ing one of those.

  Luis

Re: Would it help to encourage users to read/test stable-review patches

[Stefan Richter]

Luis R. Rodriguez wrote:
>> More review never hurts, but do you think there's a specific lack or
>> problem?
> 
> Not really, I was thinking about the fellow Ubunteros users who tend
> to not have a clue but sometimes enough good will to want to help. If
> the average user understood better the upstream process then perhaps
> it would become clearer to them the importance of stable patches and
> stable-review.

Shouldn't these people rather get started with "the other stable
kernel", called 2.6.x-rcZ?  And the next level would be to either test
-next or follow the master branch of a subsystem development tree of
their interest.

I think it could be more effective overall to activate more people who
watch for regressions in the mainline, send qualified and early
regression reports, and test candidate fixes.  A side effect of better
-rc kernels will be a shallower 2.6.x.y branch queue.

Patches which are submitted for stable-review should already have been
tested as part of the upstream kernel on affected and unaffected
systems.  As Chris wrote, patches which obviously need further field
testing when backported to -stable are per definition too risky for
-stable in the first place.  -stable submission should be low-risk
cherry-picking.

> Bringing users closer to stable-review to me indicates
> more users would understand the stable process and therefore would
> stop bitching to me when a 'non-cirtical' fix never made it into their
> ancient stable. Cc'ing one of those.

Ah, you have especially /these/ people in mind.  Well, set up a clever
procmail filter which autoresponds with a pointer to
Documentation/stable_kernel_rules.txt... :-)
-- 
Stefan Richter
-=====-==--= =--- ===--
http://arcgraph.de/sr/

Re: Would it help to encourage users to read/test stable-review patches

[Tilman Schmidt]

[-- Attachment #1: Type: text/plain, Size: 1232 bytes --]

Luis R. Rodriguez schrieb:
> We get development review eyes on stable patches during submission
> into a subsystem, and perhaps 1-2 user tests then. Typically patches
> are pretty easy to review for stable submission but I'm wondering
> about cases where things are not so clear [...]
> Do we want more user exposure to stable-review mailing list or is just
> a few developer eyes enough?

Having been bitten quite recently by a -stable mismerge, I am of two
minds about this. On one hand, it would have been nice to catch that
problem during stable review instead of when it hit an actual
distribution. (Ubuntu, as it happened.) On the other hand, I'm not
sure user exposure would have helped all that much in this case.

Generally speaking, the way I understand stable rules, those cases
where things are not so clear shouldn't make it into the review in
the first place. The problem is with the cases where things appear
to be quite clear but aren't. Those are the ones the review cycle
must catch.

Jm2c
T.

-- 
Tilman Schmidt                    E-Mail: tilman@imap.cc
Bonn, Germany
Diese Nachricht besteht zu 100% aus wiederverwerteten Bits.
Ungeöffnet mindestens haltbar bis: (siehe Rückseite)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 254 bytes --]