Re: [topgit] tg update error

[Junio C Hamano <gitster@pobox.com>]

Junio C Hamano <gitster@pobox.com> writes:

> If you *are* setting HEAD to some ref that is outside refs/heads (or even
> inside refs/heads for that matter), at that point the HEAD is *not*
> detached, so no, it obviously is *not* what is happening.
>
> I am asking why you need to use a ref to do that, *if* it is a tentative
> state while the program is running.  You are probably calling a git
> plumbing or Porcelain command that updates HEAD, and the reason why you
> point HEAD outside refs/heads/ is beause you would want the command you
> call to update one of the refs/top-bases/ ref through HEAD.  I am asking
> why you are not running these commands on a normal detached HEAD, and then
> use update-ref (not symbolic-ref) plumbing to update the refs/top-bases/
> ref you would want to update when it is done.

Ok, I did read the script (yuck).  You do break out of TopGit process when
a merge conflict prevents the update operation to complete and do give
control back to the end user, so you can leave HEAD in a state that points
at a non-branch, and you do use the fact that the HEAD is pointing at
something funny as a sign that you are in the middle of conflicted merge
resolution.

It is just like how vanilla git uses MERGE_HEAD as the marker to signal
that it is in a funny state.

While I think it is a cute idea to use which funny hierarchy HEAD points
at to indicate what funny/intermediate state your interrupted operation is
in, and it may seem to be cleaner than using a marker file like MERGE_HEAD
at first sight, I do not think it is a wise thing to do in the long run.

You can only express two pieces of information (the overall "category of
state" by which funny ref/ hierarchy HEAD points at, and one object name
by storing it in the ref pointed at by HEAD), and if you need more (such
as MERGE_MSG that stores pre-packaged log message pieces is used during a
merge, in addition to MERGE_HEAD), you would need to use more than just
the "cute HEAD" trick to store them *anyway*.  Which means that it is a
bad tradeoff to use "cute HEAD" --- it closes the possibility to detect
user error to point HEAD at an incorrect place and I do not see the
benefit of "cute HEAD" outweigh the downside.