Re: ag/rebase-i-rewrite-todo, was Re: What's cooking in git.git (Jun 2018, #07; Thu, 28)

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi Junio,

On Fri, 6 Jul 2018, Junio C Hamano wrote:

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > I would *strongly* encourage you to allow Alban to go back to the small,
> > incremental patch series he sent before, because it will make it
> > *substantially* easier to not only review, but also develop, and for
> > you to merge.
> 
> An organization in which you can make sure that the order of dependency
> and which ones have been updated since previous rounds are clear, even
> to those who are looking from the sidelines ("these 4 patches are to
> replace patch 3, 7 and 8 from the previous round" is already hostile to
> late reviewers and doing so without a pointer to the archive is even
> worse---a full reroll with the unchanged ones marked below the
> three-dash lines would be perfect), would be good.  A random collection
> of seemingly separate but actually interdependent topics is very hard to
> work with with limited mental bandwidth.
> 
> Once the core of _a_ topic hits 'next', we can go incremental (because
> by definition things get quiet and require small updates by then), but
> not before.
> 
> I think the 7 patches in ag/rebase-i-in-c are more or less in good
> shape, modulo the issues pointed out on the list yet to be addressed,
> which I do not think require redesign.  Which is good.

You do understand that with your proposed "let's just roll them up into
one big patch series, and just add freely whatever you need on top", these
7 patches (3 of which I reviewed I think four times on the list now, and
more times on GitHub, which is quite taxing on my time) will be soon
joined by 6 more patches: https://github.com/git/git/pull/518

Of course, at that point I will have to look through those 7 patches
again, if only to verify that yes, they are still the same.

And Alban is not sitting on his hands, either.

So after reviewing those 13 patches, which undoubtedly will not be
integrated into `next` under the premise that they are still in flux, they
will most likely be joined by another dozen patches until the interactive
rebase is rewritten completely in C. After which time, I will have
reviewed the first 3 patches over 15 times.

I wish there was a better way.

Ciao,
Dscho