threads / discuss / 17961

Counter-intuitive results for git show and git checkout during rebase with conflict.

Subject: Counter-intuitive results for git show and git checkout during rebase with conflict.

## tl;dr

2 messages between Feb 23, 2009 and Feb 23, 2009.

replies: 1people: 2as markdown or json

Nazri Ramliy· Feb 23, 2009, 04:04 UTC · lore

The scenario: We have two branches, local and master, we are now on branch local, and we would like to rebase local wrt master:

	% git rebase master
	conflictedfile: needs merge
	cannot rebase: you have unstaged changes

Obviously we have a conflict, here's the problem: these commands have counter-intuitive effect (as oppose to during a git merge with conflict):

  git show :2:conflictedfile  # shows content from master version
  git show :3:conflictedfile  # shows content from local version
  git checkout --ours conflictedfile   # gets content from master version
  git checkout --theirs conflictedfile # gets content from local version

I know why they are counter-intuitive - :2:, :3:, --ours and --theirs are relative to the current <commit>, and during a conflict due to a rebase, the current <commit> is some commit that leads to master, which is not anywhere in the path that leads to local.

So in summary:
Fact 1:
  During a conflict due to a rebase, HEAD is a commit that leads to
  the other branch.
Fact 2:
  During a conflict due to a merge, HEAD is a commit that leads to the
  current branch.
(Please correct me if the two facts above are not true)

Technically there's nothing wrong with the behavior of the commands, but wouldn't it be better if the arguments :2:conflictedfile and :3:conflictedfile to git show and the options --ours and --theirs to git checkout be made aware of the two facts above and do a more intuitive action?

Or should this be left to a higher level tools to automatically detect the reason of the conflict and adjust the arguments appropriately so that the end results are intuitive for the user?

nazri.
Jay Soffian· Feb 23, 2009, 07:16 UTC · re: Nazri Ramliy · lore

Re: Counter-intuitive results for git show and git checkout during rebase with conflict.

On Sun, Feb 22, 2009 at 11:04 PM, Nazri Ramliy <ayiehere@gmail.com> wrote:
Show 43 quoted lines
> The scenario: We have two branches, local and master, we are now on
> branch local, and we would like to rebase local wrt master:
>
>        % git rebase master
>        conflictedfile: needs merge
>        cannot rebase: you have unstaged changes
>
> Obviously we have a conflict, here's the problem: these commands have
> counter-intuitive effect (as oppose to during a git merge with
> conflict):
>
>  git show :2:conflictedfile  # shows content from master version
>  git show :3:conflictedfile  # shows content from local version
>
>  git checkout --ours conflictedfile   # gets content from master version
>  git checkout --theirs conflictedfile # gets content from local version
>
> I know why they are counter-intuitive - :2:, :3:, --ours and --theirs
> are relative to the current <commit>, and during a conflict due to a
> rebase, the current <commit> is some commit that leads to master,
> which is not anywhere in the path that leads to local.
>
> So in summary:
>
> Fact 1:
>  During a conflict due to a rebase, HEAD is a commit that leads to
>  the other branch.
>
> Fact 2:
>  During a conflict due to a merge, HEAD is a commit that leads to the
>  current branch.
>
> (Please correct me if the two facts above are not true)
>
> Technically there's nothing wrong with the behavior of the commands,
> but wouldn't it be better if the arguments :2:conflictedfile and
> :3:conflictedfile to git show and the options --ours and --theirs to
> git checkout be made aware of the two facts above and do a more
> intuitive action?
>
> Or should this be left to a higher level tools to automatically detect
> the reason of the conflict and adjust the arguments appropriately so
> that the end results are intuitive for the user?

I took a crack at fixing this issue in the UI (well, just mergetool), but that wasn't the right way to do it. See Junio's reply in this thread:

http://thread.gmane.org/gmane.comp.version-control.git/74595
So, patches welcomed. :-)
j.

← back to recent threads