Skip to contentSkip to conversation

Suggestion About Kernel Releases

DISCUSSION Tarkan Erimer 8 participants 16 messages May 21, 2008 – May 24, 2008

lore
Share & tools
lore
#0 lore raw

Hi,

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

- 2 weeks of merge window for each "rc" releases should remain the same .
- When decided the 2.6.xx-rcx is ready to become 2.6.xx, it should be 2.6.xx-test1 instead of 2.6.xx
- Chris Wright (or whoever will handle these "testX" maintaining) can maintain these "testX" releases as like 2.6.x.y maintaining.
- After releasing the "testX" releases, the new kernel release process should start.
- When Chris decided it is stable enough, it should release as 2.6.xx (stable).
- Of course, 2.6.x.y releases follows then as usually.

Please share your opinions. Thanks :-)

Tarkan

#1 lore raw
Quoted 5 lines · On Wed, 21 May 2008 16:08:37 +0300 Tarkan Erimer <[email protected]> wrote:

Hi,

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

what would you want to accomplish with your idea?

#2 lore raw
Quoted 2 lines · Tarkan Erimer
Quoted 2 lines

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

Quoted 1 line · Arjan van de Ven

what would you want to accomplish with your idea?

And how is it different from what we do now, modulo the -test1 postfix to the name?

#3 lore raw
Quoted 6 lines · Pekka Enberg
Quoted 3 lines · Tarkan Erimer
Quoted 3 lines

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

Quoted 8 lines
Quoted 2 lines · Arjan van de Ven

what would you want to accomplish with your idea?

And how is it different from what we do now, modulo the -test1 postfix to the name?

Please see my reply to Arjan's post.

#4 lore raw
Quoted 21 lines · Tarkan Erimer
Quoted 1 line · Pekka Enberg

On Wed, 21 May 2008 16:08:37 +0300, Tarkan Erimer <[email protected]>

Quoted 10 lines · wrote:
Quoted 8 lines
Quoted 7 lines

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

On Wed, May 21, 2008 at 4:38 PM, Arjan van de Ven <[email protected]>

Quoted 6 lines · wrote:
Quoted 5 lines

what would you want to accomplish with your idea?

Quoted 6 lines
Quoted 4 lines

And how is it different from what we do now, modulo the -test1 postfix to the name?

Please see my reply to Arjan's post.

I think you're missing Arjan and Pekka's point: your proposal doesn't _really_ offer any change to the current procedure. It might make you feel more warm and fuzzy but in practice:
1) Linus would still release a kernel (be it to -testX or "final") when he and others feel it is time
2) The stable team will track fixes and release stable kernels as needed

The only thing that is different in your approach is the "final" release would theoretically be more well tested. Unfortunately that is not a valid assumption because the wider Linux audience likely won't embrace the latest kernel until it is "final" anyway. This delayed uptake can/will result in early stable fixes.

Again, no real change... we know that the "final" release that Linus makes _could_ have some minor oversight that will be fixed fairly quickly by the stable team.

Mike

#5 lore raw
Quoted 9 lines · Mike Snitzer

I think you're missing Arjan and Pekka's point: your proposal doesn't _really_ offer any change to the current procedure. It might make you feel more warm and fuzzy but in practice:
1) Linus would still release a kernel (be it to -testX or "final") when he and others feel it is time
2) The stable team will track fixes and release stable kernels as needed

The only thing that is different in your approach is the "final"

Yes, that's what I'm talking about. The change will be the "testX" thing. And as you already stated it will help more stable releases.

Quoted 6 lines

release would theoretically be more well tested. Unfortunately that is not a valid assumption because the wider Linux audience likely won't embrace the latest kernel until it is "final" anyway. This delayed uptake can/will result in early stable fixes.

Of course as didn't embrace like the "rc" releases. I don't think that it will cause early stable fixes. Contrarily, it will produce more tested/stable releases. You may think it as a kind of "rc" series.

Quoted 6 lines

Again, no real change... we know that the "final" release that Linus makes _could_ have some minor oversight that will be fixed fairly quickly by the stable team.

Mike

It is. As I described above : the "testX" series will change the things as producing more tested /stable releases. BTW, it should be much better to release less, trouble free kernels instead of releasing fast but fixing soon. Also, it will take the load of deep kernel testing. Linus should only do the applying the patches for the next kernel release via "rc" releases. Then when Linus decided all the patches, for the next kernel, applied he will pass it to the "testX" series maintainer as Chris did for 2.6.x.y.

#6 lore raw
Quoted 4 lines · Tarkan Erimer

It is. As I described above : the "testX" series will change the things as producing more tested /stable releases. BTW, it should be much better to release less, trouble free kernels instead of releasing fast but fixing soon.

You keep saying that but it simply is not true. The more people we have _waiting_ for a "stable" kernel to graduate from the "testing" series, the less people we have actually _testing_ the kernel.

So it's probable that your suggestion actually makes the current situation worse, not better.

#7 lore raw
Quoted 3 lines · Pekka Enberg

You keep saying that but it simply is not true. The more people we have _waiting_ for a "stable" kernel to graduate from the "testing" series, the less people we have actually _testing_ the kernel.

ACK.

I, personally use the newest releases on mostly unimportant systems, where a kernel bug is ugly but not that bad. For production systems I use the stableized/well-tested distro kernels, eg. unmasked on Gentoo). If you're using such an distro, you already have an more tested kernel, less chance of bugs (than with vanilla).

My suggestion istead is bringing these individual efforts from distros to some central point, let's call this "mature kernel". This (IMHO) wouldn't affect the current development/release process, mission-critical systems can simply use that mature tree, and perhaps some pressure is taken from vanilla.

cu

Signature · 8 lines
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
#8 lore raw
In reply to Arjan van de Ven
Quoted 2 lines · Arjan van de Ven

what would you want to accomplish with your idea?

To make kernel releases more stable/tested via "testX" series, as recently discussed by David Miller's "Slow DOWN, please!!!" thread. These "testX" releases will give more opportunity and time to test more these new releases. Also, with the aim of the "testX" series, the new kernel release schedules/merge windows will be not slow down. Even, Linus can releases new kernels more quickly. Because, testing purpose should be handled by someone else like Chris did for 2.6.x.y.

#9 lore raw
Quoted 11 lines · On Wed, 21 May 2008 16:59:56 +0300 Tarkan Erimer <[email protected]> wrote:
Quoted 2 lines · Arjan van de Ven

what would you want to accomplish with your idea?

To make kernel releases more stable/tested via "testX" series, as recently discussed by David Miller's "Slow DOWN, please!!!" thread. These "testX" releases will give more opportunity and time to test more these new releases. Also, with the aim of the "testX" series, the new kernel release schedules/merge windows will be not slow down. Even, Linus can releases new kernels more quickly. Because, testing purpose should be handled by someone else like Chris did for 2.6.x.y.

sorry but you're wrong ;(
calling something -test won't change anything. It just means it'll get tested less, not more.

Also the "Slow DOWN please" had NOTHING to do with releases, but only with what happens during the actual merge window.

#10 lore raw
Quoted 5 lines · Arjan van de Ven

sorry but you're wrong ;(
calling something -test won't change anything. It just means it'll get tested less, not more.

It's just a suggestion. Of course, I may be wrong ;-)
Why do you think that calling something test means that less testing ?

Quoted 3 lines

Also the "Slow DOWN please" had NOTHING to do with releases, but only with what happens during the actual merge window.

Yes, I know. That's what I'm talking about :-)

#11 lore raw
Quoted 8 lines · On Wed, 21 May 2008 17:50:00 +0300 Tarkan Erimer <[email protected]> wrote:
Quoted 5 lines · Arjan van de Ven

sorry but you're wrong ;(
calling something -test won't change anything. It just means it'll get tested less, not more.

It's just a suggestion. Of course, I may be wrong ;-)
Why do you think that calling something test means that less testing ?

because fewer people will use it compared to a final release.

Quoted 5 lines
Quoted 4 lines

Also the "Slow DOWN please" had NOTHING to do with releases, but only with what happens during the actual merge window.

Yes, I know. That's what I'm talking about :-)

no you're not. You're talking about the tail of a cycle, while the Slow Down thing was ONLY about the first 2 weeks merge window.

Since this discussion has all the signs of a derailing thread.. I would strongly suggest we end it here. Not trying to just cut you off, but this has been discussed a lot, and your posts aren't adding something new to that.

#12 lore raw
In reply to Tarkan Erimer
Quoted 2 lines · * Tarkan Erimer ([email protected]) wrote:

To make kernel releases more stable/tested via "testX" series, as recently discussed by David Miller's "Slow DOWN, please!!!" thread.

Arjan already pointed out that this was about the merge window, not the tail.

Quoted 6 lines

These "testX"
releases will give more opportunity and time to test more these new releases. Also, with the aim of the "testX" series, the new kernel release schedules/merge windows will be not slow down. Even, Linus can releases new kernels more quickly. Because, testing purpose should be handled by someone else like Chris did for 2.6.x.y.

(BTW, Greg does 2.6.x.y as well). What you are missing is that it is not Greg and I who are doing the bulk of testing for 2.6.x.y. It's the entire community, and adding additional stage to the tail or the release cycle will not increase testing coverage. History shows it'll just postpone it.

thanks,
-chris

#13 lore raw
Quoted 6 lines · Chris Wright

(BTW, Greg does 2.6.x.y as well). What you are missing is that it is not Greg and I who are doing the bulk of testing for 2.6.x.y. It's the entire community, and adding additional stage to the tail or the release cycle will not increase testing coverage. History shows it'll just postpone it.

Yep,sorry for forgetting Greg ;-) I meant testing maintenance(applying bug fixes etc.), of course not testing by yourself.

Tarkan

#14 lore raw
In reply to Tarkan Erimer
Quoted 2 lines · On Wed, 21 May 2008 16:59:56 +0300 Tarkan Erimer <[email protected]> wrote:

Even, Linus can releases new kernels more quickly. Because, testing purpose should be handled by someone else like Chris did for 2.6.x.y.

If Linus were to go straight from merge window to merge window, with stabilization happening in parallel, then there would be no stable basis at the start of each merge window and bugs can just pile up.

Speeding things up could lead to the same kind of code quality issues we used to have in the 1.3, 2.1, 2.3 and 2.5 kernel series.

Slowing down regularly is good, not bad.

Signature · 1 line
All rights reversed.
#15 lore raw
In reply to Tarkan Erimer
Quoted 16 lines · Tarkan Erimer

Hi,

After long discussions about kernel release methodology, an idea has came to ( Any comments welcomed :-) ) my mind :

- 2 weeks of merge window for each "rc" releases should remain the same .
- When decided the 2.6.xx-rcx is ready to become 2.6.xx, it should be 2.6.xx-test1 instead of 2.6.xx
- Chris Wright (or whoever will handle these "testX" maintaining) can maintain these "testX" releases as like 2.6.x.y maintaining.
- After releasing the "testX" releases, the new kernel release process should start.
- When Chris decided it is stable enough, it should release as 2.6.xx (stable).
- Of course, 2.6.x.y releases follows then as usually.

Quoted 9 lines

Please share your opinions. Thanks :-)

Tarkan
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

now:
linux-2.6.x -> linux-2.6.(x+1)-rcY -> linux-2.6.(x+1)

         \->2.6.x.y
       \->2.6.(x+1).y
                  \->2.6.x.(y+1)-rc1->2.6.x.(.y+1)
           \->2.6.x.(y+1)-rc1->2.6.x.(.y+1)

your idea:
linux-2.6.x -> linux-2.6.(x+1)-rcY -> linux-2.6.(x+1)

         \->2.6.x.y
       \->2.6.(x+1).y
                  \->2.6.x.(y+1)-test1->2.6.x.(.y+1)
             \->2.6.x.(y+1)-test1->2.6.(x+1).(.y+1)

only the namig is was other, rc -> test,

Signature · 2 lines
Thanks,
Oliver