Weekly update: 0.12

John Arbash Meinel john at arbash-meinel.com
Mon Oct 9 01:31:12 BST 2006


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

This is just a general update to indicate how the 0.12 release is
shaping up. We've had a few good patches land, and we have several more
in the queue.

Remember, next Monday (Oct 16th) is the feature-freeze date. So all
non-trivial changes need to be reviewed and merged this week.

I'd like to say that things seem to be moving forward well. 0.12 has the
possibility to include a few big changes, though at the moment it may be
more of a polish and minor improvements release.


Things that have landed:

cleanup progress bar - A nice cleanup by Aaron, to properly finalize
   progress bars when interrupted during fetch.

reverting files removes them - More handling of cases when files that
   have not been modified can be removed. Which allows a simple api for
   removing working trees.

lock write will wait - when acquiring a write lock, the user is prompted
   and the default is to wait for 5 minutes before giving up. Should
   help reduce user pain when committing to a shared location.

test case with memory transport - A nice base class that can speed up
   simple test cases considerably (by not requiring any disk i/o).


Things that have been submitted and should be landed soon:

unique tree roots - We stalled for a while on whether a 'Tree' can be
   rootless. But Robert has finally agreed that it should be possible
   to have no root. I've reviewed Aaron's changes, and IIRC with a
   little bit of cleanup, it is ready to be merged.

help topics - There has been a couple patches which seem to have
   stalled. Goffredo submitted some updates that did not seem to get
   reviewed. I'm hoping he can re-submit them to the mailing list, and
   we can get this landed

registry - Also something that I've submitted a couple times, and just
   needs final approval

lazy imports - I've started a series of refactoring, to get the bzr
   startup time down. A few patches have been submitted, though this is
   an ongoing change. If this is going to be merged, I'd like to see it
   happen early this week, since it is something that can cause hidden
   import order dependencies.

urllib keep alive - Vincent Ladeuil's work to bring webdav support has
   also provided us with a keep alive implementation. This should let us
   make urllib the default http provider, and avoid some of the
   negatives of pycurl (downloading a large file cannot be interrupted
   is the only big one I know of)

cascading support for config files- This one seems to be stalled. I'll
   try to talk to Martin and Robert to see if we can get it unstuck.

update --revision - This also seems to be stalled. With difficulty
   deciding whether --revision should be evaluated locally or remotely.

cleanup use message for ssh errors - Richard Wilbur thought this was
   good, but I haven't heard from anyone else.

Transport.get() is defined that it can raise a lazy exception on read()-
   This was done by Robert because sftp has that property, and it is
   important that the rest of the codebase be aware of this fact. While
   I was originally unsure, I know think it is a good thing. And though
   I thought I gave it a +1, I'll give it one now, just in case.

Show full name of plugin- I think this is a good direction, but probably
   needs a little bit more work. Frequently plugins will have dots in
   their name, which aren't version numbers (bzr.webdav comes to mind).
   Robert has made it clear that he feels plugins should have a
   canonical name, which must be a valid python variable name. It does
   have a lot of benefits, so I would like to see us work towards this.

bzr+http:// and served over fastcgi - Andrew Bennetts has done good work
   in making the smart server work over http. While it needs a little
   bit of cleanup, I don't see any reason why it shouldn't land for
   0.12.

Things that are in progress:

Martin-
  Updated benchmark display, to make it easier to see what is going on.

  Tags- We've been discussing this a lot, and unfortunately, a clear
        path has still not emerged. The current proposal has some merit
        and may also introduce some other benefits to make it work well.

Robert-
  'dirstate' changes. He and I have been working on it while I'm in AU,
  and we have been sending a few refactoring patches. With some luck,
  we'll be able to see a new working tree format in 0.12.

Aaron-
  support path-kind changing - This claims to be 'blocked', though I'm
    not sure what it is blocked on. Aaron, can you clarify? I know you
    sent in a patch a while ago which would at least allow 'commit' to
    function, which was a pre-requisite for converting some branches
    from Arch.

  nested tree support - I think this is blocked on unique roots, which
    should be cleared for landing.

John-
  commit performance - This is an ongoing thing. I still have my patch
    which removes the hash prefixes for knit storage. Which does seem
    to improve performance under certain circumstances.  I cleaned it
    up a bit for bzr.dev, though it still needs upgrade logic to be
    added.

  bug triaging - One of my big goals is to have several stabs at the bug
    tracker, and try to get fixed bugs marked as such, and get trivial
    bugs squashed. Some of this can happen after the feature freeze,
    though I'd like to have at least one bug-day this week, in case
    there are bugs that involve less trivial changes.


I'm sure there is more that I've overlooked. If you feel like I've
missed something that you have worked on, please post a follow up. I'm
trying to get a good summary, so that we can make sure things aren't
being dropped. And if anyone feels like they need a little bit of extra
support, just send me an email. At times I have limited access, but I've
been trying to download everything, and do all my responding while
offline. I'd like to be around IRC more, but that probably won't happen
this week.

Overall, I'm very pleased with how 0.12 is shaping up, and I'd like to
thank everyone for their involvement.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFKZhQJdeBCYSNAAMRAoHMAJ9bnHnQvuEhQtPGhIr2yoJv0VPL7ACfTM0i
6kf51yTiJ7Mc9PaXBKsCLDM=
=h9cR
-----END PGP SIGNATURE-----




More information about the bazaar mailing list