Any ex ClearCase or VSS users out there?

John Arbash Meinel john at arbash-meinel.com
Wed Oct 21 16:32:23 BST 2009


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Krzysztof Nowicki wrote:
> John Arbash Meinel wrote:
>>> I think however that with a bit of hacking it should be possible to produce a
>>> full revision log of a file using the data stored in the repository, since
>>> Bazaar tracks files using file ids. Am I right?
>>
>> So 'bzr qlog branch1 branch2' or 'bzr qlog repository' can all show you
>> the history of the whole repository across branches. I don't think qlog
>> supports showing the history for a single file across branches yet,
>> though it certainly would be possible.
>>
>> Probably the biggest difference is that the *act* of branching is not a
>> versioned event. So those rectangles don't really exist for us.
> 
> As branching is concerned it's worth noting that ClearCase creates a branch of a
> file when the file is first checked-out and not when the branch is created
> (unless view profiles are used, which put a timestamp rule into the configspec).
> This effectively means that the branchpoint for a file is invisible until the
> file is first checked out. This looks similar to Bazaar, where you'd see the
> branch point once you commit your first revision on the branch. The only corner
> case is when you check-out a file in CC and undo the checkout leaving a branch
> with just version 0. This is however considered bad practice in the CC world anyway.
> 
>> Clearcase also choses to show more information inline, and side-by-side.
>> I think I prefer qlog's formatting, though it would sometimes be nice to
>> have parallel lines rather than one-line-per revision.
> 
> If I remember correctly, graph-ancestry from bzrtools should provide such
> feature, only it doesn't allow to perform any actions on the graph. The problem
> is that I never actually got it working anyway.
> 
>> John
>> =:->

graph ancestry writes out a GraphViz 'dot' file and then lets dot sort
out where things should go. Interesting in theory, but it really doesn't
scale well to 25k revisions. (I once tried to pass a object reference
graph to 'dot' with 300k objects, and after 1.3GB of memory and 8 hours
of churning, I just killed it.)

The other issue is that once 'dot' completes, you often end up with a
.png file that your image viewers can't handle, because it represents an
enormous viewing area.

So I've tried it a few times, and mostly have stopped trying. Because
'dot' doesn't seem to scale up very well. (per file graphs would be a
bit simpler, since they rarely change on *every* commit.)

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkrfKYcACgkQJdeBCYSNAAMyngCfYExji7BqJBuEYWs1NkqzsQRC
CxcAoLEynEdoWXy8xXp10SQ0qbGovEAt
=bsAq
-----END PGP SIGNATURE-----



More information about the bazaar mailing list