gentoo-sources vs vanilla-sources

Message

Dieter@be · Post by **Dieter@be** » Fri Jul 14, 2006 7:15 pm

I would like a little more information then "gentoo-sources are based on vanilla-sources but with some tweaks"

Because if those extra tweaks or updates are soo good, why aren't they commited to the vanilla sources? (i don't think these tweaks are good for gentoo-ers only? in fact gentoo-sources kernel can run any other linux distro, right? and on the other hand, a vanilla-sources kernel can perfectly run a gentoo system, right?)

You see i'm wondering if i shouldnt better go from gentoo-sources to vanilla-sources because in case of problems, its much easier to solve them if you can compare your situation to the rest of the world (instead of only with a part of the gentoo community), but first offcourse i would like to know the what,why, .... about those "tweaks"

Beefrum · Post by **Beefrum** » Fri Jul 14, 2006 7:24 pm

Very visible!
It will say "gentoo" on the <uname -a> command-line.
So it can count as ONE MORE victory over every other distro!

Monkeh · Post by **Monkeh** » Fri Jul 14, 2006 7:28 pm

Many of them ARE included in the official sources at a later date. Infact a lot of them are straight out of development versions of the kernel.

michel7 · Post by **michel7** » Fri Jul 14, 2006 8:14 pm

gentoo-sources are patched vanilla sources. To see all patches applied to vanilla look at http://dev.gentoo.org/~dsd/genpatches/kernels.htm

Paapaa · Post by **Paapaa** » Fri Jul 14, 2006 8:23 pm

Another questions: is there any practical need for gentoo-sources as they all basically have the exact same bug fixes eventually. I think that gentoo-sources is basically useless as there is nothing gentoo specific there AFAIK.

Monkeh · Post by **Monkeh** » Fri Jul 14, 2006 8:47 pm

Paapaa wrote:Another questions: is there any practical need for gentoo-sources as they all basically have the exact same bug fixes eventually. I think that gentoo-sources is basically useless as there is nothing gentoo specific there AFAIK.

vesafb-tng, gensplash..

Post by **dsd** » Fri Jul 14, 2006 11:13 pm

Paapaa wrote:Another questions: is there any practical need for gentoo-sources as they all basically have the exact same bug fixes eventually. I think that gentoo-sources is basically useless as there is nothing gentoo specific there AFAIK.

gentoo-sources has the bugfixes generally before they are shipped in any official stable kernel release -- try telling a user that the bug is fixed, their (e.g.) sound card works again, but they cant have the fix shipped in the portage tree until a new stable kernel comes out in 2 months time.

vanilla-sources is untouched by us, hence the name. we need gentoo-sources simply because it is a kernel tree which we control.

but yes, it is our aim for genpatches to get smaller and smaller, and it is certainly going this way.

Beefrum · Post by **Beefrum** » Fri Jul 14, 2006 11:45 pm

Think me missing the point that gentoo-sources are there to support ALL arch's Gentoo is intended to run on 'supported' as the different information keeps stating.

Dieter@be · Post by **Dieter@be** » Sat Jul 15, 2006 11:02 am

what is a genpatch? is it a patch that is proposed for the official tree, that is applied to gentoo-sources? or are they gentoo-"only" projects like the above-mentioned vesafb-tng or gensplash ?

Monkeh · Post by **Monkeh** » Sat Jul 15, 2006 3:54 pm

genpatches is the patchset used for gentoo-sources.

Dieter@be · Post by **Dieter@be** » Sat Jul 15, 2006 4:49 pm

okay, that's probably 100% correct what you say, but unfortunately still doesn't answer my question

intgr · Post by **intgr** » Sat Jul 15, 2006 4:55 pm

The patches to gentoo-sources do usually come from kernel.org, but since they do bugfix backports and minor releases relatively rarely, it's beneficial for distributions to maintain their own patchsets. At least that's my understanding of it, anyway.

Monkeh · Post by **Monkeh** » Sat Jul 15, 2006 5:21 pm

Dieter@be wrote:okay, that's probably 100% correct what you say, but unfortunately still doesn't answer my question

It does, actually. There's no such thing as a 'genpatch', genpatches is the patchset for gentoo-sources, which contains everything. Fixes from dev kernels, fixes from elsewhere, vesafb-tng, etc.

Dieter@be · Post by **Dieter@be** » Sat Jul 15, 2006 6:41 pm

oh i see, thanks

in that case, i think i'll stick to gentoo-sources

alistair · Post by **alistair** » Sun Jul 16, 2006 6:34 am

As a related note it would be interesting to see if, with respect to the patches that do make it into the vanilla-sources, if gentoo-sources (stable) ~ vanilla-sources (unstable/testing).

I say this because if we can assume that all patches within genpatches make it into vanilla-sources ( I realise this is a long shot ) it could be possible for a gentoo user to have the latest stable gentoo-sources kernel and the latest testing vanilla kernel that are logically exactly the same.

Also it would be interesting to know because the faster the patches within genpatches are 'released' into stable vanilla-sources the more the aims of gentoo-sources should change. If the situation were to arise where vanilla and gentoo sources were equal consistantly (an even greater long shot ) then genpatches should change its focus to areas that would separate it from the vanilla-sources. So in this case bringing in more performance and functionality patches.

Post by **dsd** » Sun Jul 16, 2006 8:49 am

i dont think i understand your suggestion. you are saying that because most of genpatches-2.6.17 is shipped in upstream 2.6.18-rc1, we should consider dropping gentoo-sources and just making all users switch to 2.6.18-rc1 because it should be "logically exactly the same"?

this isnt true and is highly impractical. just look at the sheer size of the 2.6.18-rc1 changelog: http://www.kernel.org/pub/linux/kernel/ ... 2.6.18-rc1

2.6.18-rc1 may include much of genpatches, but it includes many many new features and dramatic changes which are only very lightly tested. in fact there are many bugs with 2.6.18-rc1, including random crashes which are very hard to diagnose and solve. this is expected given that the code is very new and there are so many changes all over the shop. there is going to be about 6 whole weeks of bugfixing *only* (no new features until 2.6.19 now) before this code gets released as 2.6.18. even then there will probably be 1 or 2 showstopping bugs which slip through, as always.

i cant see us patching any functionality/performance stuff because we have worked very hard to gain support from the kernel development community (who typically spit at distro kernels), and if we start heavily patching then we will throw that all away. if we can get to the point where the differences really are minimal then we will simply use any time we have freed up to improve our bugfixing efforts, we already have too many bug reports coming in (or not enough maintainers) to give full attention to.

Paapaa · Post by **Paapaa** » Sun Jul 16, 2006 9:24 am

dsd, thanks for very clear explanation on why gentoo-sources are being maintained. Obviously for some users genpatches are very useful and for some users they are not needed. Of course the best situation would be that there was no need for distro-specific patchset, and that vanilla-sources was bugfree

alistair · Post by **alistair** » Sun Jul 16, 2006 10:44 am

dsd wrote:i dont think i understand your suggestion. you are saying that because most of genpatches-2.6.17 is shipped in upstream 2.6.18-rc1, we should consider dropping gentoo-sources and just making all users switch to 2.6.18-rc1 because it should be "logically exactly the same"?

this isnt true and is highly impractical. just look at the sheer size of the 2.6.18-rc1 changelog: http://www.kernel.org/pub/linux/kernel/ ... 2.6.18-rc1

I was trying to present a hypothetical situation where
1) genpatches-x-y-z is shipped in upstream-x1-y1-z1
2) upstream has been released as stable (and I mean not marked ~ in portage).
3) corresponding genpatches and upstream sources (gentoo-sources and vanilla-sources) are marked stable within a close time frame (consider if the time difference was 2 months, 1 month, 2 weeks, 1 week, 1 day, 1 hour.....)
4) This is consistantly the case.
5) A whole lot of other assumptions have been made.

So maybe to ask differently.

Under what situations do you think that genpatches would become ineffectual and should modify it goals to, for example, be a test bed of patches (and have the instability that may result) or as another example becomes a project that backports security fixes to a collection of supported older kernels.

Post by **dsd** » Sun Jul 16, 2006 11:17 am

i dont think that will ever happen

we'll always need the option (even if seldom used) of patching our officially supported kernel. for example, the most recent security bug was reported to vendor-sec and the kernel security group by ourselves. various potential fixes were thrown around, one was agreed on as a suitable fix, but at this point it was not clear whether a linux-stable maintainer was even awake to publish a new release. in this situation, there would be absolutely no point in us waiting potentially hours for 2.6.17.4 to appear, we obviously want to get on the case right that moment.

as well as that, we also need to provide completely untouched (and never to be touched) sources to keep the support from the kernel development community.

in terms of making a gentoo-specific experimental patchset, i dont think that is likely. firstly, any heavily patched kernel is a maintenance nightmare. secondly, such a test-bed should be produced upstream and shared between all distros. this is already the case: the -mm kernel is the testbed for patches.

as for using genpatches to provide security fixes for older kernels, the system is set up to allow exactly that, and it already happens to a certain extent. a couple of developers continued maintenance of hardened-sources-2.6.14 (by producing genpatches releases) up until a few weeks ago. this would happen more often if we had the manpower to do so, and also there need to be good reasons for us to spend time doing this (it is very time consuming and there are always other things to work on)

Post by **dsd** » Sun Jul 16, 2006 11:24 am

it is also worth noting that we generally do not recommend vanilla-sources for use. this is because we do not fix it if it is broken. instead, we recommend gentoo-sources, which as you have noted is becoming more and more identical to vanilla-sources, except we do fix it when bugs are reported.

it is useful that the linux-stable branch does now exist. this means that vanilla-sources are generally much less broken, since they release small incremental versions to fix the big bugs. remember this is still a fairly new thing. even though that linux-stable makes genpatches maintenance so much simpler (and the inverse is true as well: we are one of the most regular contributors to linux-stable), their patching style does not fully meet the needs of gentoo, so we have to provide a small patchset on top.

so, instead of viewing it that we potentially provide two very similar kernels (vanilla-sources and gentoo-sources), view it as the fact we only support one of them: gentoo-sources, and the other is provided for good measure and to ease development and testing.