pick: remove transaction on the whole command (issue6037)
At its core, pick is a pretty straightforward and well-behaving command, it
uses functions already in core hg, it checks that wdir is clean and that
changeset to pick is not public, it checks if there happen to be merge
conflicts and can be --continue'd later, etc.
It is very similar to graft in core (it also uses mergemod.graft function), but
it obsoletes the original changeset. However, graft does not experience this
incorrect behavior from issue 6037.
What happens in the test case for this issue when we pick a revision that
touches both "a" and "b": mergemod.graft() takes the original changeset and
tries to apply it to the wdir, which results in "b" being marked as newly added
and ready to be committed, "a" updated with the new content and being marked as
modified, but "a" also has conflicts. Pick correctly notices this and saves its
state before asking for user intervention. So far so good. However, when the
command raises InterventionRequired to print a user-facing message and exit
while being wrapped in repo.transaction() context manager, the latter partially
undoes what mergemod.graft() did: it unmarks "b" as added. And when user
continues pick, "b" is therefore not tracked and is not included in the
resulting commit.
The transaction is not useful here, because it doesn't touch wdir (it's still
dirty), it doesn't remove pickstate (and other commands will refuse to work
until pick --abort or --continue), it just makes "b" untracked.
The solution is to use repo.transaction() only to wrap code that writes data to
hg store in the final stages of the command after all checks have passed and is
not expected to fail on trivial cases like merge conflicts. For example,
committing the picked changeset. But since pick uses repo.commit() for that,
and because that function already uses a transaction, wrapping it in another
transaction doesn't make sense.
.. Copyright 2011 Pierre-Yves David <pierre-yves.david@ens-lyon.org>
.. Logilab SA <contact@logilab.fr>
-----------------------------------
Terminology of the obsolete concept
-----------------------------------
Obsolete markers
----------------
The mutable concept is based on **obsolete markers**. Creating an obsolete
marker registers a relation between an old obsoleted changeset and its newer
version.
Old changesets are called **predecessors** while their new versions are called
**successors**. A marker always registers a single *predecessor* and:
- no *successor*: the *predecessor* is just discarded.
- one *successor*: the *predecessor* has been rewritten
- multiple *successors*: the *predecessor* were splits in multiple
changesets.
.. The *predecessors* and *successors* terms can be used on changeset directly:
.. :predecessors: of a changeset `A` are changesets used as *predecessors* by
.. obsolete marker using changeset `A` as *successors*
.. :successors: of a changeset `B` are changesets used as *successors* by
.. obsolete marker using changeset `B` as *predecessors*
Chaining obsolete markers is allowed to rewrite a changeset that is already a
*successor*. This is a kind of *second order version control*.
To clarify ambiguous situations one can use **direct predecessors** or
**direct successors** to name changesets that are directly related.
The set of all *obsolete markers* forms a direct acyclic graph the same way
standard *parents*/*children* relation does. In this graph we have:
:any predecessors: are transitive predecessors of a changeset: *direct predecessors*
and *predecessors* of *predecessors*.
:any successors: are transitive successors of a changeset: *direct successors*
and *successors* of *successors*)
Obsolete markers may refer changesets that are not known locally.
So, *direct predecessors* of a changeset may be unknown locally.
This is why we usually focus on the **first known predecessors** of the rewritten
changeset. The same apply for *successors*.
Changeset in *any successors* which are not **obsolete** are called
**newest successors**..
.. note:: I'm not very happy with this naming scheme and I'm looking for a
better distinction between *direct successors* and **any successors**.
Possible changesets "type"
--------------------------
The following table describes names and behaviors of changesets affected by
obsolete markers. The left column describes generic categories and the right
columns are about sub-categories.
+---------------------+--------------------------+-----------------------------+
| **mutable** | **obsolete** | **extinct** |
| | | |
| Changeset in either | Obsolete changeset is | *extinct* changeset is |
| *draft* or *secret* | *mutable* used as a | *obsolete* which has only |
| phase. | *predecessor*. | *obsolete* descendants. |
| | | |
| | A changeset is used as | They can safely be: |
| | a *predecessor* when at | |
| | least one obsolete | - hidden in the UI, |
| | marker refers to it | - silently excluded from |
| | as predecessors. | pull and push operations |
| | | - mostly ignored |
| | | - garbage collected |
| | | |
| | +-----------------------------+
| | | |
| | | **suspended** |
| | | |
| | | *suspended* changeset is |
| | | *obsolete* with at least |
| | | one non-obsolete descendant |
| | | |
| | | Those descendants prevent |
| | | properties of extinct |
| | | changesets to apply. But |
| | | they will refuse to be |
| | | pushed without --force. |
| | | |
| +--------------------------+-----------------------------+
| | | |
| | **unstable** | **orphan** |
| | | |
| | *unstable* has | *orphan* is a changeset |
| | unresolved issue caused | with obsolete ancestors. |
| | by *obsolete* relations. | |
| | | |
| | Possible issues are | It must be rebased on a |
| | listed in the next | non *unstable* base to |
| | column. It is possible | solve the problem. |
| | for *unstable* | |
| | changeset to combine | (possible alternative name: |
| | multiple issue at once. | precarious) |
| | (a.k.a. content-divergent| |
| | and orphan) +-----------------------------+
| | | |
| | (possible alternative | **phase-divergent** |
| | names: unsettled, | |
| | troublesome | *phase-divergent* is a |
| | | changeset that tries to be |
| | | successor of a public |
| | | changeset. |
| | | |
| | | Public changeset can't |
| | | be deleted and replace |
| | | *phase-divergent* |
| | | need to be converted into |
| | | an overlay to this public |
| | | changeset. |
| | | |
| | | (possible alternative names:|
| | | mislead, naive, unaware, |
| | | mindless, disenchanting) |
| | | |
| | +-----------------------------+
| | | **content-divergent** |
| | | |
| | | *content-divergent* is a |
| | | changeset that appears when |
| | | multiple changesets are |
| | | successors of the same |
| | | predecessor. |
| | | |
| | | *content-divergent* are |
| | | solved through a three way |
| | | merge between the two |
| | | *content-divergent* , |
| | | using the last "obsolete- |
| | | -common-ancestor" as the |
| | | base. |
| | | |
| | | (*splitting* is |
| | | properly not detected as a |
| | | conflict) |
| | | |
| | | (possible alternative names:|
| | | clashing, rival, concurrent,|
| | | conflicting) |
| | | |
| +--------------------------+-----------------------------+
| | |
| | Mutable changesets which are neither *obsolete* or |
| | *unstable* are *"ok"*. |
| | |
| | Do we really need a name for it ? *"ok"* is a pretty |
| | crappy name :-/ other possibilities are: |
| | |
| | - stable (confusing with stable branch) |
| | - sane |
| | - healthy |
| | |
+---------------------+--------------------------------------------------------+
| |
| **immutable** |
| |
| Changesets in the *public* phases. |
| |
| Rewriting operation refuse to work on immutable changeset. |
| |
| Obsolete markers that refer an immutable changeset as predecessors have |
| no effect on the predecessors but may have effect on the successors. |
| |
| When a *mutable* changeset becomes *immutable* (changing its phase from draft|
| to public) it is just *immutable* and loose any property of it's former |
| state. |
| |
| The phase properties says that public changesets stay as *immutable* forever.|
| |
+------------------------------------------------------------------------------+
Command and operation name
--------------------------
Existing terms
``````````````
Mercurial core already uses the following terms:
:amend: to rewrite a changeset
:graft: to copy a changeset
:rebase: to move a changeset
Uncommit
````````
Remove files from a commit (and leave them as dirty in the working directory)
The *evolve* extension have an `uncommit` command that aims to replace most
`rollback` usage.
Fold
````
Collapse multiple changesets into a unique one.
The *evolve* extension will have a `fold` command.
Prune
`````
Make a changeset obsolete without successors.
This an important operation as it should mostly replace *strip*.
Alternative names:
- kill: shall has funny effects when you forget "hg" in front of ``hg kill``.
- obsolete: too vague, too long and too generic.
evolve
``````
Automatically resolve *troublesome* changesets
(*orphan*, *phase-divergent* and *content-divergent*)
This is an important name as hg pull/push will suggest it the same way it
suggest merging when you add heads.
Alternative names:
- solve (too generic ?)
- stabilize