lkml.org 
[lkml]   [2019]   [Aug]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 5.2 000/135] 5.2.10-stable review
On Fri, Aug 23, 2019 at 07:32:58PM -0700, Greg KH wrote:
>On Fri, Aug 23, 2019 at 09:18:05PM -0400, Sasha Levin wrote:
>> On Fri, Aug 23, 2019 at 10:36:27AM -0700, Greg KH wrote:
>> > On Fri, Aug 23, 2019 at 02:28:53AM -0400, Sasha Levin wrote:
>> > > On Fri, Aug 23, 2019 at 02:42:48AM +0200, Stefan Lippers-Hollmann wrote:
>> > > > Hi
>> > > >
>> > > > On 2019-08-22, Greg KH wrote:
>> > > > > On Fri, Aug 23, 2019 at 12:05:27AM +0200, Stefan Lippers-Hollmann wrote:
>> > > > > > On 2019-08-22, Greg KH wrote:
>> > > > > > > On Thu, Aug 22, 2019 at 01:05:56PM -0400, Sasha Levin wrote:
>> > > > [...]
>> > > > > > It might be down to kernel.org mirroring, but the patch file doesn't
>> > > > > > seem to be available yet (404), both in the wrong location listed
>> > > > > > above - and the expected one under
>> > > > > >
>> > > > > > https://www.kernel.org/pub/linux/kernel/v5.x/stable-review/patch-5.2.10-rc1.gz
>> > > > [...]
>> > > > > Ah, no, it's not a mirroring problem, Sasha and I didn't know if anyone
>> > > > > was actually using the patch files anymore, so it was simpler to do a
>> > > > > release without them to see what happens. :)
>> > > > >
>> > > > > Do you rely on these, or can you use the -rc git tree or the quilt
>> > > > > series? If you do rely on them, we will work to fix this, it just
>> > > > > involves some scripting that we didn't get done this morning.
>> > > >
>> > > > "Rely" is a strong word, I can adapt if they're going away, but
>> > > > I've been using them so far, as in (slightly simplified):
>> > > >
>> > > > $ cd patches/upstream/
>> > > > $ wget https://cdn.kernel.org/pub/linux/kernel/v5.x/patch-5.2.9.xz
>> > > > $ xz -d patch-5.2.9.xz
>> > > > $ wget https://www.kernel.org/pub/linux/kernel/v5.x/stable-review/patch-5.2.10-rc1.gz
>> > > > $ gunzip patch-5.2.10-rc1.gz
>> > > > $ vim ../series
>> > > > $ quilt ...
>> > > >
>> > > > I can switch to importing the quilt queue with some sed magic (and I
>> > > > already do that, if interesting or just a larger amounts of patches are
>> > > > queuing up for more than a day or two), but using the -rc patches has
>> > > > been convenient in that semi-manual workflow, also to make sure to really
>> > > > get and test the formal -rc patch, rather than something inbetween.
>> > >
>> > > An easy way to generate a patch is to just use the git.kernel.org web
>> > > interface. A patch for 5.2.10-rc1 would be:
>> > > https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git/patch/?id=linux-5.2.y&id2=v5.2.9
>> > >
>> > > Personally this patch upload story sounded to me like a pre-git era
>> > > artifact...
>> >
>> > Given that we no longer do patches for Linus's -rc releases for the past
>> > few years, maybe it is time to move to do the same for the stable
>> > releases to be consistent.
>>
>> Or tarballs? Why do we generate tarballs (and go through kup)?
>> git.kernel.org already does it for us.
>
>As I mentioned yesterday, but writing it down here for posterity,
>there's a number of reasons.
>
>First off, the release process doesn't require kup for when a "real"
>release happens, that's all now donw on git.kernel.org with a process
>involving a signed note and some other fun backend stuff. We are
>working on expanding that in the future with some other signature
>validation as well, to make it easier to verify tarballs are "correct"
>as what we do today is a bit different than other projects.

I think that I made it read like I want to remove tarballs altogether.
That's not the case: I just want to get rid of the magical signed note
process.

The way I understand it, we generate tarballs twice: once during the
magical signed note process, and once by the git interface. I'm just
suggesting we reduce that down to happen once.

Right now you can fetch tarballs from two different links on kernel.org.
For example, a 5.2.9 tarball is available at:

- https://mirrors.edge.kernel.org/pub/linux/kernel/v5.x/linux-5.2.9.tar.xz
- https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/snapshot/linux-5.2.9.tar.gz

Can't we symlink one to the other?

>As for the tarball itself, it's still needed for the same reasons we do
>so on Linus's releases:
> - distros use these. Don't want all Gentoo users hammering on
> git.kernel.org for their updated builds, that's a huge waste.

We can just place git.kernel.org generated tarballs (for some repos) on
the CDN, no?

> - mirroring works _so_ much better around the world for tarballs

Doing the above should solve this.

> - legal reasons. git is not yet "recognized" as being something
> that properly is reflective of a specific point in time while
> as online tarballs that are mirrored and stored around the
> world are finally almost properly recognized for this.

We still keep the tarballs.

> - historical, people are used to using them, and workflows are
> built up around them. People don't like rewriting scripts, as
> can be seen in my monstrosity of a mess that I use for
> releases :)

Right, this shouldn't require changing any scripts.

--
Thanks,
Sasha

\
 
 \ /
  Last update: 2019-08-24 07:50    [W:0.162 / U:5.340 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site