Let's say there are two branches, and last (HEAD
) revision in branch A
is 9
, while it is 6
in branch B
.
When cd B; svn merge -r 5:8 ^/braches/A
is ran, svn will try to apply delta between 5
and 8
from branch A
on top of branch B
.
(In other words, change sets 7
and 8
will be applied to B
)
common
ancestor left right
(1)━━┱───(3)──(5)──(7)──(8)──(9) # branch A
┃ └┄┄┄┄┬┄┄┄┄┘
┃ ↓
┗━(2)━━(4)━━(6) # branch B
working
If the delta applies cleanly, it's all good.
Let's say some lines were modified in change set 3
, and same source lines were modified differently in change set 4
.
If delta (5
→8
) doesn't touch those lines, all is still good.
If delta (5
→8
) also modified what 3
and 4
did, changes cannot be merged automatically, and svn leaves a file in conflict state:
file
--- file with (working
, left
, right
) delimited
file.working
--- state of file in branch B@6
file.merge-left
--- state of file in branch A@5
file.merge-right
--- state of file in branch A@8
If you edit such a file manually, you have a few choices --- keep working
(your version), keep right
(their version; the other branch version) or merge the changes manually.
Left
is not useful in itself, there's no point to keep left
(their old version) in the file.
It is, however, useful for tools. left
→right
is the change set.
When you see, for example:
<<<<<<< .working
life_universe_and_everything = 13
||||||| .merge-left.r5
life_universe_and_everything = "13"
=======
life_universe_and_everything = "42"
>>>>>>> .merge-right.r8
In branch A
, "13"
(str) was changed to "42"
.
Branch B
had 13
(int).
Perhaps you want 42
(int) when you reconcile this conflict manually.
r
is revision number. File with extension .rXXX is that file in that given revision. – Erroneous