Pushing after merge considered harmful

[Stefan Monnier]

>>> I really can't think of any kind of branch where this option would
>>> hurt (if you ignore uncommit, obviously).
>> Branches where mainline has no meaning, I presume.  No idea what
>> use-cases this would cover, though.

"mainline has no meaning" doesn't mean that "append_only" would hurt, tho.

> Some people continuously develop in a single local branch and pull trunk
> into that when the current code has landed.

That would be either a "pull --overwrite" or something like
a "merge --pull" (tho maybe they just typed "pull" because they somehow
happen to know that it would work as well, of course).

I would consider the "pull --overwrite" to be in the same camp as
"uncommit", so I think it's a separate issue.

> It would also generally be incompatible with people using "bzr merge
> --pull". (Which are people who assign no meaning to the mainline.)

Right, these use cases would be affected, but would they be hurt?
...hmmm... yes, I see that I've been in such a case once and indeed
I didn't care about the mainline and I suffered from bzr's handling of
the mainline.  So I think that the default "bzr log" behavior should
depend on the "append-only" flag: if that flag is not set (I'd have it
be set by default), then "bzr log" should show details of merges rather
than only show the mainline commits.


        Stefan