Re: git trees organization

[Stephen Hemminger <stephen@networkplumber.org>]

On Thu, 14 Sep 2017 11:11:56 +0200
Nélio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:

> On Thu, Sep 14, 2017 at 10:22:23AM +0200, Thomas Monjalon wrote:
> > 14/09/2017 04:25, Stephen Hemminger:  
> > > Bisecting a tree with lots of subtree merges is terrible. That is why Linus
> > > rebases and doesn't directly take linux-next  
> > 
> > I agree, bisecting with subtree merges is not pleasant at all.
> > That's why I chose the rebase method until now.  
> 
> I don't see what is un-pleasant, if we start a bisect what we expect is to
> find the commit introducing the issue, bisect is able to do it even on large
> tree with a lot of merges.  Moreover, the probability the issue will be
> searched in a specific section within its own subtree is high which also means
> locally it will be linear, is not it equivalent to the actual situation?
> 
> > Adrien mentioned some drawbacks with the rebase method.
> > Ferruh mentioned some drawbacks and some advantages of rebase.
> > Stephen mentioned another advantage of rebase.
> > Such decisions are really difficult.
> > One thing is sure: there will be always someone unhappy,
> > no matter the decision :)
> > 
> > When we want to take such decision or re-consider it,
> > we ask the techboard to vote...  

A recent git bisect gives an example of the problem.

I needed to bisect between two daily versions of linux-next.
Linux-next is intentionally not a serial tree, it is recreated every day.

The big bisect wanted to go through 10,000 commits and back track from
4.14-rc1 into 4.13-rc5 to get down into some subtree. 

On upstream tree it nevers goes back into ancient history.