Speedup with history-db

John Arbash Meinel john at arbash-meinel.com
Tue May 31 09:47:55 UTC 2011


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


...
>> Counting the whole revision table is stupid. I just committed what
>> should be a fix for that (revno 135).
> 
> Thanks.  This slashed the time significantly, but bzr consistently
> says it imports 4 revisions each time I do a "bzr up" with no new
> revisions available in the remote repo:
> 
>   9.360  fetch up to rev {rgm at gnu.org-20110527071815-pnm36i0e38mb8wte}
>   11.766  history_db post-change-hook took 0.246s (0.002s to get_config, 0.096s to init, 0.148s to import)
>   11.766  Stats:
>   {'num_search_tips': 0,
>    'step mainline': 1,
>    'step mainline added': 4,
>    'step mainline cache missed': 1,
>    'step mainline initial': 1}
>   [ 3688] 2011-05-27 18:06:55.062 INFO: Tree is up to date at revision 104384 of branch bzr+ssh://eliz@bzr.savannah.gnu.org/emacs/trunk
> 
> It looks like something else is at work here.

1) That isn't actually imported, that is how much data we read from the
db to do our work. I should probably clarify the strings, but I'm not
really sure how to do so.

2) rev 139 has a few fixes
   a) I used to explicitly check that a tip wasn't already imported,
      apparently that got removed, but I brought it back.
   b) handle the 'bzr up' case when nothing changes. Bazaar fires the
      "post-branch-tip-changed" hook, even though old_revid ==
      new_revid. I added the shortcut so we don't even have to parse
      config files, etc. It should be an uncommon case, but we might as
      well have a no-op be a no-op.

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

iEYEARECAAYFAk3kuUsACgkQJdeBCYSNAAOWmgCfeF/IziYFwugzUYgn6j6Gjq7+
VHUAoIqLgZG+WQBxD3DSqeVZDugHR1E1
=6OdM
-----END PGP SIGNATURE-----



More information about the bazaar mailing list