list mirror (unofficial, one of many)
 help / color / Atom feed
From: Stefan Beller <>
To: Yaroslav Halchenko <>
Cc: git <>
Subject: Re: [wishlist] git submodule update --reset-hard
Date: Fri, 7 Dec 2018 13:55:15 -0800
Message-ID: <> (raw)
In-Reply-To: <>

On Thu, Dec 6, 2018 at 5:23 PM Yaroslav Halchenko <> wrote:

> > There was a proposal to "re-attach HEAD" in the submodule, i.e.
> > if the branch branch points at the same commit, we don't need
> > a detached HEAD, but could go with the branch instead.
> if I got the idea right, if we are talking about any branch, it
> would also non-deterministic since who knows what left over branch(es)
> point to that commit.  Not sure if I would have used that ;)

I would think we'd rather want to have it deterministic, i.e. something like
1) record branch name of the submodule
2) update submodules HEAD to to superprojects gitlink
3) if recorded branch (1) matches the sha1 of detached HEAD,
  have HEAD point to the branch instead.

You notice a small inefficiency here as we write HEAD twice, so it
could be reworded as:
1) compare superprojects gitlink with the submodules branch
2a) if equal, set submodules HEAD to branch
2b) if unequal set HEAD to gitlink value, resulting in detached HEAD

Note that this idea of reattaching reflects the idea (a) below.

> > a) "stay on submodule branch (i.e. HEAD still points at $branch), and
> > reset --hard" such that the submodule has a clean index and at that $branch
> > or
> > b) "stay on submodule branch (i.e. HEAD still points at $branch), but $branch is
> >    set to the gitlink from the superproject, and then a reset --hard
> >    will have the worktree set to it as well.

> NB "gitlink" -- just now discovered the thing for me.  Thought it would be
> called a  subproject  echoing what git diff/log -p shows for submodule commits.

The terminology is messy:
The internal representation in Gits object model is a "gitlink" entry in a tree
object. Once we have a .gitmodules entry, we call it submodule.

The term 'subproject' is a historic artifact and will likely not be changed
in the diff output (or format-patch), because these diffs can be applied using
git-am for example. That makes the diff output effectively a transport
protocol, and changing protocols is hard if you have no versioning in them.

More in (a rather recent new write
of a man page, going into concepts).

> > > right -- I meant the local changes and indeed reset --recurse-submodules
> > > indeed seems to recurse nicely.  Then the undesired effect remaining only
> > > the detached HEAD
> > For that we may want to revive discussions in
> >
> well, isn't that one requires a branch to be specified in .gitmodules?

Ah good point.

> >   git reset --hard --recursive=hard,keep-branch PREVIOUSPOINT
> 'keep-branch' (given aforementioned keeping the specified in .gitmodules
> branch) might be confusing.  Also what if a submodule already in a
> detached HEAD?  IMHO --recursive=hard, and just saying that it would do
> "reset --hard", is imho sufficient.  (that is why I like pure
> --reset hard   since it doesn't care and neither does anything to the
> branch)

For that we might want to first do the

  git submodule update --reset-hard

which runs reset --hard inside the submodule, no matter which
branch the submodule is on (if any) and resets to the given
superproject sha1.

See in git.git[1] in cmd_update.
We'd need to add a command line flag (`--reset-hard`
would be the obvious choice?) which would set the `update`
variable, which then is evaluated to what needs to be done in
the submodule, which in that case would be the hard reset.

Once that is done we'd want to add a test case, presumably
in t/

> > > I would have asked for
> > >    git revert --recursive <commit>...
> > >    git rebase --recursive [-i] ...
> > > which I also frequently desire (could elaborate on the use cases etc).
> > These would be nice to have. It would be nice if you'd elaborate on the
> > use cases for future reference in the mailing list archive. :-)
> ok, will try to do so ;-) In summary: they are just a logical extension
> of git support for submodules for anyone actively working with
> submodules to keep entire tree in sync.  Then quite often the need for
> reverting a specific commit (which also has changes reflected in
> submodules) arises.  The same with rebase, especially to trim away some
> no longer desired changes reflected in submodules.
> the initial "git submodule update --reset-hard" is pretty much a
> crude workaround for some of those cases, so I would just go earlier in
> the history, and redo some things, whenever I could just drop or revert
> some selected set of commits.

That makes sense.
Do you want to give the implementation a try for the --reset-hard switch?

> ah... so it is only   submodule  command which has --recursive, and the
> rest have --recurse-submodules   when talking about recursing into
> submodules?

I don't think we were that cautious in development as it was done by
different people at different times. There is also just `--submodule` for
the diff family, for reference:

  reply index

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-12-06 17:35 Yaroslav Halchenko
2018-12-06 18:29 ` Stefan Beller
2018-12-06 21:24   ` Yaroslav Halchenko
2018-12-06 21:55     ` Stefan Beller
2018-12-07  1:22       ` Yaroslav Halchenko
2018-12-07 21:55         ` Stefan Beller [this message]
2018-12-08  2:15           ` Yaroslav Halchenko
2018-12-08  4:21             ` Yaroslav Halchenko
2018-12-10 18:58               ` Stefan Beller
2018-12-10 20:14                 ` Yaroslav Halchenko
2018-12-11  4:08                 ` [PATCH 1/2] submodule: Add --reset-hard option for git submodule update Yaroslav Halchenko
2018-12-11  4:08                   ` [PATCH 2/2] RF+ENH(TST): compare the entire list of submodule status --recursive to stay intact Yaroslav Halchenko
2018-12-12 19:48                     ` Stefan Beller
2018-12-13 16:42                       ` Yaroslav O Halchenko
2018-12-13 20:44                         ` Stefan Beller
2018-12-13 22:43                           ` Yaroslav O Halchenko
2018-12-13 23:58                             ` Stefan Beller
2018-12-14  4:22                               ` Yaroslav O Halchenko

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

  List information:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link list mirror (unofficial, one of many)

Archives are clonable:
	git clone --mirror
	git clone --mirror http://ou63pmih66umazou.onion/git
	git clone --mirror http://czquwvybam4bgbro.onion/git
	git clone --mirror http://hjrcffqmbrq6wope.onion/git

Newsgroups are available over NNTP:

 note: .onion URLs require Tor:

AGPL code for this site: git clone public-inbox