[Gnu-arch-users] Re: Re: How to recover my work after a failed update?

[David Allouche]

On Mon, 2005-02-28 at 16:57 +0100, Matthieu MOY wrote:
> David Allouche said:
> > On Sun, 2005-02-27 at 13:56 -0800, Lyle Brulhart wrote:
> >> Thanks,
> >>
> >> It has worked perfectly!  in my case:
> >>
> >> tla redo ,,undo-1
> >
> > Just "tla redo" would have done it, by the way.
> >
> > Also, that's something that you should never have to do.
> >
> > Rationale: you should commit your changes before updating,
> 
> Err, you just can't. If you need to "tla update", it means your tree is
> not up to date, and you're not allowed to commit.
> Arch encourrages decentralized revision control, in which "update" is
> replaced by other merge operators (star-merge, replay, apply-delta ...)
> so, your rationale should be
> "Rationale: you should not use update and commit your changes before merging"

Right, but I wanted to stay simple.

An alternative could be "you should undo, then update and commit, then
redo your local changes". Which would actually work in the centralised
case.

> I wouldn't completely agree with this either: One of the strenth of arch
> is to allow both centralized and decentralized development. Some people
> may have good reasons to keep a centralized branch. (I have: I'm working
> on two different machines, both having access to the archive, but none
> being able to access the other, and I use commit/update to synchronize
> them. It's much simpler than branching each time I change of machine)

In your situation, you run no risk (or a very controlled risk) of having
conflicts on update, since you are the only committer.

To preserve integrity and "purity" of data, it is important that merges
be separated from original changes. There are many way to do it.

-- 
                                                            -- ddaa