Re: Having a custom merge process (was: EWW bugs should be fixed on emacs-25 branch)

[Eli Zaretskii]

> From: John Wiegley <address@hidden>
> Cc: Eli Zaretskii <address@hidden>,  address@hidden,  address@hidden
> Date: Fri, 25 Dec 2015 15:35:30 -0800
> 
> >>>>> David Engster <address@hidden> writes:
> 
> > gitmerge.el will detect cherry-picks through the cherry-mark. However, since
> > git is the C++ of version control, this works only 99% of the time. It does
> > not work when the cherry-pick could not be applied cleanly, since then the
> > patch-id changes. It also does not work when the same commit is also merged
> > later (see the beginning of the emacs-25 branch for an example, where first
> > a bunch of commits was cherry-picked and later 'master' was also merged).
> 
> OK, that's understandable. What I'd like to avoid -- as much as feasible -- is
> having so much of a custom merge process, apart from what most other projects
> using Git do, that it becomes difficult to teach newcomers about our process.

Newcomers are told to use gitmerge.el, so how that could be a problem?

Eventually, what gitmerge.el does is invoke "git merge" to merge
ranges of commits that were selected in a semi-automatic way; see
gitmerge-apply.  There's nothing mysterious or magic in that command
that I could spot; I don't even think that it's a particularly
advanced way of doing merges with Git.  Just more or less standard
stuff.

> I've used Git in many scenarios, on many projects, but some of the practices
> we're using here for Emacs development I've never seen before. This simply
> makes me wonder how much of that difference is truly necessary, or is a result
> of having evolved to this point from CVS and then Bazaar.

Can you point out those practices that surprised you?