dotted revision numbers oddity
John Arbash Meinel
john at arbash-meinel.com
Sun Apr 8 23:06:30 BST 2007
Erik Bågfors wrote:
> I belive they are numbered after where they where when committed, not
> where they are now. That is. at rev 55, someone created a new branch,
> commited 56, 57, .....
>
> Then in "this" branch, someone commited 56, and then merged the other
> branch.
> so 55.1.1 has a "parrent" and that is 55.
Correct. The graph would look something like:
55 - - - -,
| \ \
| 55.1.1 55.2.1
| | |
56 55.1.2 55.2.2
| / |
57 /
| . - - -'
58
So things are numbered based on their source, not their merge target.
If you want, you can see a huge discussion on this in the archives. Both
have merit, but ultimately the nice thing about these numbers is that
they are 100% stable from a given mainline[1] and easy to compute.
John
=:->
[1] Any future actions will not change the numbering as long as the
left-hand parents are kept. So if you merge 58 into a different branch
of 55.2.2, then the numbers would change, since 55.2.2 might become the
mainline #57.
>
> /Erik
>
> On 4/8/07, Matthieu Moy <Matthieu.Moy at imag.fr> wrote:
>> Hi,
>>
>> I have a very strange case with dotted revision numbers :
>>
>> $ bzr log http://greensocs.sourceforge.net/pinapa/download/bzr/pinapa/
>> [...]
>> revno: 57
>> message:
>> [...]
>> ------------------------------------------------------------
>> revno: 55.1.5
>> merged:
>> Arch-1:Matthieu.Moy at imag.fr--pinapa%pinapa--arrays--0--patch-4
>> [...]
>> ------------------------------------------------------------
>> revno: 55.1.4
>> [...]
>> ------------------------------------------------------------
>> revno: 55.1.1
>> [...]
>> message:
>> tag of Matthieu.Moy at imag.fr--pinapa/pinapa--main--0--patch-41
>> ------------------------------------------------------------
>> revno: 56
>> [...]
>> ------------------------------------------------------------
>> revno: 55
>> committer: Matthieu Moy <Matthieu.Moy at imag.fr>
>> timestamp: Mon 2005-04-04 08:59:06 +0000
>> message:
>> Create .libs/ before running "ln -s" in gcc.mk
>> ------------------------------------------------------------
>>
>> I would have expected revisions showing up after 57 to be numbered
>> 57.something. Here, they are 55.{1,2,3,4,5}.
>>
>> Is it a bug, or a surprising feature?
>>
>> (as you might have guessed, this is an import from an arch repo)
>>
>> Thanks,
>>
>> --
>> Matthieu
>>
>>
>
>
More information about the bazaar
mailing list