Possible benchmark issue

[John Arbash Meinel]

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

I was just thinking about benchmarks, and I wanted to mention something.

As part of the last _iter_changes updates, I also fixed a bug where we
would cache the stat value even though the timing was incorrect. (We
would re-read a file if its stat value was within the range, but if we
cache it now, and it becomes invalid within that second, and then 2-days
later we wouldn't re-read it).

Anyway, this will have an impact on some benchmarks. Because we
currently use a 3-second window for determining freshness of files
(because vfat has a 2-second minimum resolution). And if you run 'bzr
status' 2 times inside that window, we will look extra slow, because
nothing will be cached. (And with the changes in place, 'bzr status'
becomes faster, so we are more likely to run into this).

Just something to think about.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGXzErJdeBCYSNAAMRAj+MAKCTfKxludcdmOOsO39s9l+pA6ebvQCgkdxc
V4KFd+tBheyuEkRDJYMHYM4=
=Wqb8
-----END PGP SIGNATURE-----