[PATCH] faster directory walking

[Sanjoy Mahajan]

> Just to be sure, did my current patch add this regression or does it
> already exist?

I'm pretty sure it already exists as I haven't incorporated the
patches (see copy of my email below).  Jeff Sipek first reported it a
couple days ago, and said he was running hg 9344f5dd4488.  It happens
for me running, let me see...9344f5dd4488 (the same as Jeff).  From
'hg log', to give you the date:

changeset:   1043:9344f5dd4488
tag:         tip
parent:      1042:23f9d71ab9ae
parent:      -1:000000000000
manifest:    1043:bc6c9f4ffce5907883086de13e7204c83f54599c
user:        mpm at selenic.com
date:        Wed Aug 24 22:19:35 2005 -0700
summary:     Update tests

-Sanjoy


Date:    Fri, 26 Aug 2005 18:46:11 BST
From:    Sanjoy Mahajan <sanjoy at mrao.cam.ac.uk>
To:      Jeff Sipek <jeffpc at optonline.net>
cc:      mercurial at selenic.com
Subject: Re: Bug: ignoring files 
X-Bogosity: Ham, spamicity=0.000000, tests=bogofilter-0.92.8

Hmm, I think this used to work (0.6b I think) because I added a pdf
file with 'hg add file.pdf' even though \.pdf$ is in my .hgignore.

But I tried a late-model hg and it behaves as you say.  I agree, it's
a bug.  I don't know if it's easy to fix because the hgignore is wired
deeply into the treewalk code.  I tried a cheap patch that I thought
would work (making matchfn=util.always in commands.py:makewalk() if hg
add was passed a filename).  But it didn't do anything.

-Sanjoy