Re: [kernel.org users] Please remember to run 'git gc' on your repositories

[Steven Rostedt <rostedt@goodmis.org>]

[ took off users.kernel.org and added LKML ]

There's a thread started on LKML, search for the subject:


"[RFC] to rebase or not to rebase on linux-next" (thanks Ingo for the
typo fix).

And continue this conversation there.

Thanks!

-- Steve



On Fri, 2009-10-23 at 22:55 +0200, Stefan Richter wrote:
> Nicolas Pitre wrote:
> > On Fri, 23 Oct 2009, Ingo Molnar wrote:
> > 
> >> linux-next should not be used as a 'test and rebase' feedback loop for 
> >> build-broken, unbisectable crap. Why do you think it should be used for 
> >> that?
> > 
> > I've been under the impression that linux-next has been created exactly 
> > for that: a tree that pulls everything together for testing and early 
> > merge conflict resolutions, and to be thrown away the next day when the 
> > process is repeated again.  It is indeed exactly that: a "test and 
> > rebase" feedback tree where the big picture can be assessed before stuff 
> > is headed for Linus's tree where commits are then stable.
> > 
> > To my knowledge, unlike Linus's tree, linux-next was never meant to be a 
> > stable tree itself, nor its components.
> [...]
> 
> The trees which are pulled into linux-next are supposed to be
> unit-tested.  It is material which you would ask Linus to pull if he had
> a merge window open today.  The only thing that this material may still
> lack is integration testing (and exposure to mainline's huge userbase of
> course).
> 
> I have doubts that the majority of build errors which linux-next
> frequently encounters are purely integration related.