Arch Linux

Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines

Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!
Tasklist

FS#17513 - [kernel26] continue kernel26 updates even if a new kernel hits [testing]

Attached to Project: Arch Linux
Opened by Emmanuel (bkk_drs) - Tuesday, 15 December 2009, 11:47 GMT
Last edited by Tobias Powalowski (tpowa) - Sunday, 31 October 2010, 06:31 GMT
Task Type Feature Request
Category Packages: Core
Status Closed
Assigned To Tobias Powalowski (tpowa)
Thomas Bächler (brain0)
Architecture All
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description: Current version of the kernel is very unlikely to be updated once a new kernel version has hit [testing]. Therefore, we could be several minor versions behind the latest stable kernel for weeks before the [testing] kernel goes [current].

To be clear, I am in no way complaining about packages being behind (this happens, for many different reasons and it is definitely not my point) but rather questionning [1] the process of totally stopping the [current] kernel development, which seems to be the standard. This is where this FR applies.

I just think it would benefit all users to get all those precious bugfixes even if some shiny new kernel is already on the way...

[1] it's the only word I could find, English is not my native language so I hope it doesn't sound too aggressive.
This task depends upon

Closed by  Tobias Powalowski (tpowa)
Sunday, 31 October 2010, 06:31 GMT
Reason for closing:  Implemented
Comment by Pierre Schmitz (Pierre) - Tuesday, 15 December 2009, 12:59 GMT
This is a general problem as our repo'/devtools don't really work with branches but only some kind of linear development.
Comment by Emmanuel (bkk_drs) - Tuesday, 15 December 2009, 13:06 GMT
Even if the packages are on a different repo (in this case, core+testing)?
Comment by Emmanuel (bkk_drs) - Tuesday, 15 December 2009, 13:15 GMT
Please let me rephrase this as I'm afraid there's some misunderstanding here. The point is not to keep several kernel versions at any time but only to update the current kernel until the testing kernel goes current. (If that was clear in the first place then it's just that I shouldn't have had this 4th beer and you can happily ignore this comment :) )
Comment by Tobias Powalowski (tpowa) - Thursday, 17 December 2009, 16:19 GMT
We cannot have 2 packages with the same name in testing repository, kernel package is a core package which needs to signoff.
Testing is used for this, in the meantime please use abs for building updated outdated kernels.
Comment by Thomas Bächler (brain0) - Thursday, 17 December 2009, 16:52 GMT
If we skip the signoff, we can manipulate the PKGBUILD in the repos/core-*/ folders without going through trunk, commit manually and run db-core.
Comment by Paul Mattal (paul) - Saturday, 06 February 2010, 14:26 GMT
This is a general structural problem with our svn layout.

It seems to me the right model for testing to be on its own branch, but this is a major structural change to the developer workflow.

I'm adding Allan, as he's the guy who has done the last few releases of devtools.

Who else should think about this issue?
Comment by Allan McRae (Allan) - Saturday, 06 February 2010, 15:15 GMT
The case of wanting to update a package for one branch while a newer branch is in [testing] is quite rare. So I do not think we need a whole work-flow change to deal with this one package. Anyway, that would still leave the sign-off issue...

The real question is, if there was a way around the issues, do any of the devs who package the kernel want to do the extra packaging? If not, then this is a non-issue.
Comment by Paul Mattal (paul) - Saturday, 06 February 2010, 16:15 GMT
I think it's rare now, because the workflow discourages it.

For example, it would be nice to have a several-month run up to firefox, thunderbird, postgres, eclipse-- you name it-- which do significant major releases which tend to break some stuff.

I'm not suggesting that's reason enough to implement a whole system for it (or that anyone should), I'm just pointing out that I think one of the reasons devs don't do a smoother testing run up to major package upgrades is because the system doesn't provide an efficient way to do it.
Comment by Allan McRae (Allan) - Sunday, 07 February 2010, 03:07 GMT
> For example, it would be nice to have a several-month run up to firefox, thunderbird, postgres, eclipse-- you name it-- which do significant major releases which tend to break some stuff.

That approach is called Debian/Ubuntu/Fedora... Bleeding edge is suppose to cause bleeding occasionally.
Comment by Paul Mattal (paul) - Sunday, 07 February 2010, 03:10 GMT
I'm not talking about Debian/Ubuntu/Fedora. I'm talking about doing the same quality and amount of work (e.g. figuring out the quirks of a new major release) in parallel while doing the minor version bumps on the previous revs.

Such as system would actually make us *more* bleeding edge, as it we'd be able to test beta packages leading up to a major release in advance.
Comment by Allan McRae (Allan) - Sunday, 07 February 2010, 03:39 GMT
I was being a little facetious above, but that is hard to covey in text format. :P

The KDE guys already manage this by setting up a temporary repo for their testing and not using trunk for their commits. Alternatively, openoffice provides a beta package. So there is already ways to do this.

Talking about "easier" approaches to handle this is a waste of time unless packagers are actually going to use it. We used to have an [unstable] repo for packaging beta release but it was unused so was removed. As far as I can tell, none of the kernel packagers have expressed interest in doing this extra packaging so...
Comment by Paul Mattal (paul) - Saturday, 06 March 2010, 21:06 GMT
It sounds like we're leaning away from a special case mechanism for doing this, as well as from making any larger change that would allow it in general.

Unless someone objects, I'll close this as "won't implement" on April 2010 bug day.
Comment by Tobias Powalowski (tpowa) - Sunday, 31 October 2010, 06:31 GMT
we ship those kernels now too and bypass testing repository.

Loading...