18
19
uncommitted changes in your tree beforehand. This has two purposes:
19
20
if you mess up the merge you won't lose anything important; secondly
20
21
this makes it more likely that the merge will be relatively pure.
23
It is a good idea to commit as soon as a merge is complete and
24
satisfactorily resolved, so as to protect the work you did in the
25
merge and to keep it separate from later development. (Mark suggests
26
an option to automatically commit when the merge is complete.)
32
bzr records what branches have been merged so far. This is useful as
33
historical information and also for later choosing a merge ancestor.
35
For each revision we record all the other revisions which have come
36
into this tree, either by being completely merged or as cherry-picks.
38
(This design is similar to the PatchLogPruning__ draft from baz.)
40
__ http://wiki.gnuarch.org/PatchLogPruning
42
This list of merged revisions is generally append-only, but can be
43
reduced if changes are taken back out. Changes can be
44
anti-cherry-picked, which causes any successors to change from being
45
fully-merged to being cherry-picked.
47
The list of merged patches is stored delta-compressed.
53
``tla update`` performs a useful but slightly subtle change: it pulls
54
in only changes that have been made on the other branch since you last
55
merged. That is to say, it sets the merge basis as the most recent
56
merged-from point on the other branch.
58
This means that any changes which were taken from your branch into the
59
other and then reversed or modified will not be reversed. Those
60
changes will always be considered as new in your branch and will have
63
The basic idea of a merge that only brings in remote work and doesn't
64
revert your own changes is good. It could be handled by a three way
65
merge with a specified version but perhaps there is a better way.
71
Merge is conducted at two levels: merging the tree shape, and merging
72
the file contents. Merging the tree shape means accounting for
73
renames, adds, deletes, etc. This is almost the same as merging the
74
two inventories, but we need to do a smart merge on them to enforce
75
structural invariants.
82
Some tools insist that you complete the entire merge while the
83
``merge`` command is running; you cannot exit the program or restart
84
the computer because state is held in memory. We should avoid that.
85
At least when the tool is waiting for user input it should have
86
written everything to disk sufficient to pick up and continue the
87
merge from that point.
89
This suggests that there should be a command to continue a merge;
90
perhaps ``bzr resolve`` should look for any unresolved changes and
91
start resolving them. ``bzr merge`` can (by default) automatically
94
One hard aspect is transformation of the tree state such as renames,
95
directory creation, etc. This might include files swapping place,
96
etc. We would like to do atomically but cannot.
103
If a merge has been begun but not committed then ``bzr revert`` should
104
put everything back as it was in the previous revision. This includes
105
resetting the tree state and texts, and also clearing the list of
106
pending-merged revisions.
113
It should be possible to download all the data necessary to do a merge
114
from a remote branch, then disconnect and complete the merge. It
115
should be possible to interrupt and continue the merge during this
118
This implies that all the data is pulled down and stored somewhere
119
locally before the actual merge begins. It could be pulled either
120
into the revision history on non-trunk revisions, or into temporary
123
It seems useful to move all revisions and texts from the other branch
124
into the storage of this branch, in concordance with the general idea
125
of every branch moving towards complete knowledge. This allows the
126
most options for an offline merge, and also for later looking back to
127
see what was merged in and what decisions were made during the merge.
134
What does cherry-picking mean here? It means we merged the changes
135
from a revision relative to its predecessor? But what if we actually
136
want to merge the delta relative to something else? Can that be
143
ddaa says perhaps we should have three states: with respect to a
144
branch any foreign revision can be *merged*, *not-merged*, or *rejected*.
145
The difference between *not-merged* and *rejected* is that not-merged
146
patches will try to merge in when you next sync from their
147
branch, whereas rejected changes will not.
149
'rejected' seems technically equivalent to it merged with the text
150
changes not present. But perhaps there should be something more?
156
Bitkeeper has the very interesting feature of *excluded* changesets,
157
where something previously committed to or merged into this branch is
158
treated as if it never occurred. Because of their representation of
159
changes as a weave they can do this without dependency problems on
162
Although the later changes will not mechanically conflict, there is of
163
course no guarantee that excluding the patch will generate anything
166
Cool though this is, it seems that few people understand it well
167
enough to use it confidently. It complicates the merge algorithms and
168
seems to have been a source of some bugs. It may be better just to
169
apply a reversed patch.
176
Each revision has a pointer to two sets, represented as lists of
179
* ``merged-revisions``
180
* ``picked-revisions``
182
When a branch is merged in, its revision history plus merged-revisions
183
are added to the merged-revisions file. When changes are
184
cherry-picked in they are added to the picked-revisions.
186
These lists are stored in the text store and their sha1 and id is
187
stored in the revision.
189
While a merge is underway, these are stored in
190
``.bzr/pending-merged-revisions`` and
191
``.bzr/pending-picked-revisions``.