Sprint summary/impressions?

[Alexander Belchenko]

Ian Clatworthy пишет:
> Is anyone planing to write up the outcomes of the Brussels sprint? What 
> hacking got done and what topics were rigorously debated over a few 
> beers? :-)
> 
> It would be nice to see some blogs as well from folk who attended, 
> giving their impressions of the event.

I'm still planning to write a post about Bazaar sprint for my russian 
bzr-day blog, so I'd better to put my english impressions here (if you 
don't mind).

The Bazaar sprint was the part of UDS-Maverick event, and this made some 
feeling (for me) that bzr is part of something bigger. This is great 
thing. But in the same time there were many times when part of the whole 
bzr team going to other discussions with adjacent projects. Something I 
did not expect beforehand.

Who was on the sprint: Martin Pool, Robert Collins, John Meinel, Aaron 
Bentley, Andrew Bennetts, Vincent Ladeuil, Martin [gz], Gary van der 
Merwe, Marius Kruger, Parth Malwankar, Alexander Belchenko. Sometimes 
other people joined us: Jelmer Vernoij, Michael Hudson (mwhudson), James 
Westby and Jonathan Lange. Also in Wednesday another one man joined us 
(QBzr team!): Luis Arias (kaaloo), who interesting in hacking on 
Explorer and QBzr. It was very cool.

I'm very sad that some other people were not be there :-( :-( :-(

It was my first bzr sprint, so I did not know what to expect and how it 
should be. I've planned to spend some time to discuss with Gary our QBzr 
plans (discussion started couple of months earlier by Ian and other QBzr 
users, see qbzr ML). Also I was in hope to spend some time to work on 
Windows installer or maybe other windows-related patched. I can't say 
I've reached these goals for 100%. I feel I did much less hacking than 
I'd like or I've planned, but instead almost all the time tried[*] to 
talk with different people. Some discussions had no final point or 
decision though, or maybe it's my poor English... I've tried to talk a 
bit about versioned properties, but I'm not sure what outcome of that 
discussion was. Really.

In the monday Vincent has presented current state of babune. The other 
days he and Martin gz spent many time working together on fixing issues 
in selftest on windows platform. It was great that windows slave started 
to run at least some part of test suite. Also Martin Pool has presented 
different statistics on bzr project about fixed bugs and reviewed merge 
proposals. It was very interesting.

I've put some gobby bzr-related sessions to the wiki: 
http://wiki.bazaar.canonical.com/UdsM . (BTW, gobby server seems to lost 
or delete some pages, fortunately I've made some copies; but maybe I've 
miss something). (Hopefully) You can get some impression about the 
topics we discussed. The very interesting discussion was about which 
changes bzr need to make in its internal and user model, see 
http://docs.google.com/Doc?docid=0AQ3S65cphSMHZGd4NWY0cHdfMTBkanRmbjljdg&hl=en_GB
Again, I'm not sure what is the conclusion on this topic.

Many discussions touched UDD, and this understandable. As I said before, 
there was very strong feeling that bzr is not just a DVCS, it's the part 
of Launchpad and Ubuntu.

There was interesting presentation on Tuesday plenary about Diffamation:
http://www.youtube.com/watch?v=x3pB2_Rs4Ao
It have inspired many ideas in Gary and me about such approach to 
history visualization. The other day we discussed merge review queue of 
bzr on Launchpad, and in the discussion Gary said there is no visible 
way to see the progress on the merge proposal after Needs Information 
vote or Needs Fixing. Gary, Aaron and me have proposed to use 
Diffamation-like approach to visualize this, i.e. show merge review 
activity as history events, e.g. "Needs Fixing, New commits, Approve". 
There was bug report filed by Aaron, but I don't remember its number.

We spend some time with Gary discussing our vision and our plans on 
future QBzr versions. Unfortunately we did not reach any conclusion re 
QBzr 1.0. So for now we're planning to work in the same way as before: 
make releases in sync with bzr, improve QBzr as we can. I've tried to 
present my vision of possible QBzr logo (smiling sun as Q letter). Luis 
said he might try to help us with better design.

I spend big part of Wednesday by explaining to Luis about internals of 
QBzr project. He started to work on small bugfix for QBzr (bug #486946).
The other day Luis bring a Mac computer, so he and Gary have spent some 
time fixing Mac-related issues. It was great for all of us.

I've spent some time with Martin gz by reviewing his latest 
win32-related patches. The I've managed to fix problem with command-line 
parser in 2.1 and problem with double backslashes. Unfortunately I've 
failed to create minimal change for this problem and ended up by 
rewriting John's code, because I have to handle many weird cases. At 
Friday I've started to work on patch re diff header encoding for windows 
(bug 382699), thanks to Aaron for a little help to get this ball moving 
forward.

Wednesday, Thursday and part of Friday we spent talking with Gary about 
QBzr. Some important points on which we both agreed:

* I would like to unify command windows in the way they shown 
information about current branch/tree.
* Gary would like to use toolbar instead of main menu for browse 
windows, and put all options currently available as buttons and 
comboboxes (in qannotate, qdiff, etc) on that toolbar.
* Gary made some progress on threading model for QBzr. It seems good.
* We made some sketches on redesign of qdiff window (Gary said he will 
put that on Wiki), and because we have big number of bugs/wishes re 
qdiff, Gary want to concentrate the efforts on improving it.
* I've re-read some old QBzr discussions and proposals from different 
people, and mockups from Craig. All this good stuff is not forgotten! We 
still planning to use your ideas. We just need more time to implement 
it. Help is welcome, of course.

Other things I've discussed with different people.

We discussed with Marius some details about nested trees. It seems my 
understanding of the Aaron design was incomplete. This is really hard 
problem with many edge use cases.

We discussed with Martin gz and Aaron the problem with checkout on 
Windows the tree which contains symlinks and case-collisions (foo vs 
Foo). For symlinks it seems possible to use views to filter out 
symlinks. I'm not sure how good views are working though. For 
case-collisions me and Martin would like to use approach of 
auto-renaming clashing file to Foo.moved and mark it as conflict. 
Unfortunately, the logic behind creating Foo.moved is not easily 
available in create checkout codepath (only in merge code), so I'm not 
sure how hard it will be to implement.

I've tried to discuss with Martin Pool new global option `--encoding` to 
allow user control the output of the bzr (desired encoding). Again this 
problem with very important for Windows, and maybe less important for 
UTF-8 environments. It seems possible and not very hard. I'd like to 
work on it after my diff patch. It seems Martin Pool did some work on 
it, I'm not quite sure.

After chat with Martin gz about Windows installer I've felt I need to 
start maintaining installer again.

Group photo: 
http://picasaweb.google.com/105838898120883430819/UDSM#5474815759644134850

Alexander

* "Tried" -- because my English speaking skill level is less than good. 
:-( I need more training and practice.