1.16 Release

[John Arbash Meinel]

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

Jonathan Lange wrote:
> Hello all,
> 
> Quite a few weeks ago, I volunteered to be the release manager for
> Bazaar 1.16. Word on the street is that my offer was accepted.
> 
> I have to apologize for being somewhat truant in my duties. There are
> reasons, but none of them good enough.
> 
> I plan on calling Martin today to make sure I understand everything I
> have to do. For starters though, how do we feel about cutting a
> release candidate this Thursday, Jun 11?
> 
> jml
> 
> 

I'll mention that I have a few patches which would be nice as part of a
1.16 release, that have been unreviewed for a bit now.

https://code.edge.launchpad.net/~jameinel/bzr/1.16-commit-fulltext/+merge/7080

Changes 'commit' to work in terms of fulltexts rather than lines. Drops
peak memory consumption when committing large files (100MB+), makes
commit a bit faster for --2.0 format repositories as it doesn't have to
split and then join again.

https://code.edge.launchpad.net/~jameinel/bzr/1.16-simple-win32/+merge/7046

A fairly trivial patch that just fixes a test case on Windows. I think
it got stuck because of conversation about code review limitations and
confusion.

https://code.edge.launchpad.net/~jameinel/bzr/1.15-pack-source/+merge/6985

Changes fetch for Pack => Pack repositories to use a custom StreamSource
rather than the InterPackRepository code. The good is that it means we
no longer download the most recent delta for *every* file in the working
tree on occasion. The small bad is that we lose progress (just like we
don't have progress for any bzr+ssh:// fetches, etc.)


https://code.edge.launchpad.net/~jameinel/bzr/1.16-chkmap-updates/+merge/7113

Improves the performance of InternalNode._iter_nodes() quite a bit. This
shows up heavily in terms of "first-commit" where we map the entire
tree. It should also show up for stuff like "bzr ls" which wants to read
a lot of pages.

https://code.edge.launchpad.net/~jameinel/bzr/1.16-chk-direct/+merge/7203

This one is slightly more iffy, but it speeds up "initial commit" by
more than 20%. It changes how we build that first inventory from calling
.map() for every file, to building a LeafNode and then splitting it as
necessary. With all of these changes, time for 'initial commit' is
finally back near --1.9 performance. (faster at committing file texts,
slower at committing a more complex inventory structure)

None of these are release blockers, though.

John
=:->

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

iEYEARECAAYFAkov3fQACgkQJdeBCYSNAANqjgCglNPR7TSCOnsbwjXRCRqNSLlR
XF8AoJayeYmOJNqkUp4Pu0SemQWg70uv
=2dSx
-----END PGP SIGNATURE-----