[PATCH 1 of 7] histedit: simplify computation of `newchildren` during --continue

[Augie Fackler]

On Sep 27, 2012, at 6:07 PM, Pierre-Yves David <pierre-yves.david at ens-lyon.org> wrote:

> # HG changeset patch
> # User Pierre-Yves David <pierre-yves.david at ens-lyon.org>
> # Date 1348786765 -7200
> # Node ID b1a3a0b12fcfe273781e8534b06d5dce260b3b93
> # Parent  5c89e7fa5bc20597080494aea43a09903e9e7a6a
> histedit: simplify computation of `newchildren` during --continue
> 
> We are now checking for any changesets between the previous `parentctx` and the
> current working directory parent. If the current working directory parent is
> inconsistent, we abort.
> 
> This change is useful as it simplify the --continue process, easing upcoming
> changes.
> 
> diff --git a/hgext/histedit.py b/hgext/histedit.py
> --- a/hgext/histedit.py
> +++ b/hgext/histedit.py
> @@ -550,21 +550,19 @@
> 
> def bootstrapcontinue(ui, repo, parentctx, existing, replacemap, rules,
>                       tmpnodes, created, replaced, opts):
> -    currentparent, wantnull = repo.dirstate.parents()
> -    # existing is the list of revisions initially considered by
> -    # histedit. Here we use it to list new changesets, descendants
> -    # of parentctx without an 'existing' changeset in-between. We
> -    # also have to exclude 'existing' changesets which were
> -    # previously dropped.
> -    descendants = set(c.node() for c in
> -            repo.set('(%d::) - %d', parentctx, parentctx))
> -    notdropped = set(n for n in existing if n in descendants and
> -            (n not in replacemap or replacemap[n] in descendants))
> -    # Discover any nodes the user has added in the interim. We can
> -    # miss changesets which were dropped and recreated the same.
> -    newchildren = list(c.node() for c in repo.set(
> -        'sort(%ln - (%ln or %ln::))', descendants, existing, notdropped))
>     action, currentnode = rules.pop(0)
> +    # is there any new commit between the expected parent and "."
> +    # XXX does not take non linear new change in account (but previous
> +    # XXX implementation didn't used them anyway

Perhaps we should abort here if that happens? Does that seem reasonable?

(This case never occurred to me.)

> +    newchildren = [c.node() for c in repo.set('(%d::.)', parentctx)]
> +    if not newchildren:
> +        # `parentctxnode` should match but no result. This means that
> +        # currentnode is not a descendant from parentctxnode.
> +        msg = _('working directory parent is not a descendant of %s')
> +        hint = _('update to proper revision and run "hg histedit '
> +                 '--continue" again')
> +        raise util.Abort(msg % parentctx, hint=hint)
> +    newchildren.pop(0)  # remove parentctxnode
>     if action in ('f', 'fold'):
>         tmpnodes.extend(newchildren)
>     else:
> diff --git a/tests/test-histedit-edit.t b/tests/test-histedit-edit.t
> --- a/tests/test-histedit-edit.t
> +++ b/tests/test-histedit-edit.t
> @@ -66,6 +66,19 @@
>   abort: Make changes as needed, you may commit or record as needed now.
>   When you are finished, run hg histedit --continue to resume.
> 
> +Go at a random point and try to continue
> +
> +  $ hg id -n
> +  3+
> +  $ hg up 0
> +  0 files updated, 0 files merged, 3 files removed, 0 files unresolved
> +  $ HGEDITOR='echo foobaz > ' hg histedit --continue
> +  abort: working directory parent is not a descendant of 055a42cdd887
> +  (update to proper revision and run "hg histedit --continue" again)
> +  [255]
> +  $ hg up 3
> +  3 files updated, 0 files merged, 0 files removed, 0 files unresolved
> +
> commit, then edit the revision
>   $ hg ci -m 'wat'
>   created new head