On 1/3/08, Linus Torvalds <[email protected]> wrote:
> > 2.6.25 actually. You don't take patches this early, do you?
>
> Yeah, no, I don't take patches early. I've considered setting up another
> branch for early patches, but decided that I'm (a) lazy and (b) better off
> encouraging people in the late -rc timeframe to just concentrate on the
> -rc's, not the next version.
This is something i was thinking to suggest.
Kernel is made of a lot of different "areas" and the regression list
is a great tool for monitoring every single area so why not opening a
new branch and accepting patches only for areas which are not in the
current regression list.?
Sounds like a good way to be more strict about regressions and
incentive people to solve regressions quicker.
Ciao,
--
Paolo
http://paolo.ciarrocchi.googlepages.com/
Paolo Ciarrocchi wrote:
> This is something i was thinking to suggest.
> Kernel is made of a lot of different "areas" and the regression list
> is a great tool for monitoring every single area so why not opening a
> new branch and accepting patches only for areas which are not in the
> current regression list.?
Some regressions can't be easily associated with an "area". And when
they can, consider the overhead involved with frequently kicking out
patchsets and taking them in again, based on when regressions become
known and when they are fixed, respectively.
> Sounds like a good way to be more strict about regressions and
> incentive people to solve regressions quicker.
To create such a motivation, that branch or tree would have to have a
practical use in development. So what purpose would such a tree
fulfill, considering that we already have a myriad of topic trees and
the -mm tree for testing and preintegration?
--
Stefan Richter
-=====-==--- ---= ---==
http://arcgraph.de/sr/
On 1/3/08, Stefan Richter <[email protected]> wrote:
> Paolo Ciarrocchi wrote:
> > This is something i was thinking to suggest.
> > Kernel is made of a lot of different "areas" and the regression list
> > is a great tool for monitoring every single area so why not opening a
> > new branch and accepting patches only for areas which are not in the
> > current regression list.?
>
> Some regressions can't be easily associated with an "area". And when
> they can, consider the overhead involved with frequently kicking out
> patchsets and taking them in again, based on when regressions become
> known and when they are fixed, respectively.
>
> > Sounds like a good way to be more strict about regressions and
> > incentive people to solve regressions quicker.
>
> To create such a motivation, that branch or tree would have to have a
> practical use in development. So what purpose would such a tree
> fulfill, considering that we already have a myriad of topic trees and
> the -mm tree for testing and preintegration?
That branch/tree would relax i bit the rule of "two weeks for merging
new stuff" for people who proven to have merged good quality code.
Ciao,
--
Paolo
http://paolo.ciarrocchi.googlepages.com/