Re: Missing changes in merges from emacs-25 to master

[Eli Zaretskii]

> Date: Tue, 22 Mar 2016 05:35:32 +0200
> From: Eli Zaretskii <address@hidden>
> Cc: address@hidden, address@hidden, address@hidden
> 
> > From: John Wiegley <address@hidden>
> > Cc: address@hidden,  address@hidden,  address@hidden
> > Date: Mon, 21 Mar 2016 14:27:54 -0700
> > 
> > As for "reason for the omission", I think there are two possible answers:
> > 
> >   1. gitmerge.el auto-marked it to be skipped, and despite my examination of
> >      each such mark, I failed to see that should not be skipped.
> > 
> >   2. I incorrectly resolved a merge conflict that seemed simple enough to me
> >      that I didn't e-mail the author of the commit about it.
> 
> But in these cases, the commit we lost should have been mentioned in
> the list of skipped one.  It wasn't, AFAICS.

I think I know what happened.  It goes like this:

  . Phillip worked on test selector for the test suite.  When he
    finished, he made a local commit 549a765 on the emacs-25 branch.

  . Phillip then said "git pull", followed by "git push".  This
    produced the merge-commit bb0cd31, which incorporated 9 other
    commits that were pushed there since the time Phillip started
    working on 549a765.  One of these 9 commits was my commit that
    fixed resize_frame_windows.

  . When John merged all of that to master, bb0cd31 was skipped.
    Skipping it omitted all the 9 commits that were merged by bb0cd31.

So the question (to John) now becomes: why was bb0cd31 skipped?  Was
that something that gitmerge.el suggested?  If so, we should fix
gitmerge.el not to make such dangerous proposals in the future.

We should also audit all the other commits merged in bb0cd31, and
probably also in any other similar merge-commit that was skipped by
John in the merge to master.