Re: [RFC] introducing git replay

[Elijah Newren <newren@gmail.com>]

Hi Sergey,

On Mon, Apr 18, 2022 at 12:30 AM Sergey Organov <sorganov@gmail.com> wrote:
>
> Elijah Newren <newren@gmail.com> writes:
>
[...]
> > Replaying merges is something I've put a little thought into, so allow
> > me to provide some pointers that may help.  Merges need special
> > handling for replaying, and in my opinion, doing either just a new
> > merge of the new trees (what rebase --rebase-merges does), or just
> > reusing existing trees (what you proposed to start this thread) are
> > both suboptimal, though the former is likely to just be annoying and
> > require potentially unnecessary user refixing,
>
> It silently drops user changes as well, and that's the worst thing about
> it, not annoyance.

Yes, I mentioned that later in the email, but omitted it in the
summary you highlight here just because the fixed-tree case was so
much more likely to do it.  Anyway, sorry for the inaccuracy in the
summarized version.

> > whereas the latter can silently discard changes or reintroduce
> > discarded changes and could be dangerous. More details on both of
> > these...
>
> Please consider yet another option:

I linked to where I had given another option.

> https://public-inbox.org/git/87r2oxe3o1.fsf@javad.com/
>
> that at least is safe with respect to user changes.

If you read the suggestion I made (which I'll reinclude here at [1]),
you'll note that I read the old thread you link to with both your and
Phillips' suggestions.  I dug into them with some examples, and came
to the conclusion that we needed something better, as I briefly
commented when proposing my suggested alternative (at [1]).  I
appreciate your suggestion and the time you put into it, but based on
my earlier investigation, I believe my suggestion would be a better
way of preserving user changes in merges and I'll be implementing it.
The fact that Martin (in this thread) independently came up with the
same basic idea and implemented it in jj (though he apparently has
some further tweaks around the object model) and it works well
suggests to me that the idea has some real world testing too that
gives me further confidence in the idea.

[1] https://lore.kernel.org/git/CABPp-BGW39_5r8Lbt3ymR+F_=hWJcf=2e7O75vFNJ=3CEL5s=g@mail.gmail.com/