branch.cache

[Ezra.Smith at bentley.com]

Sure. The exception, which shows at the bottom of web pages, is:
Exception exceptions.AttributeError: "'atomictempfile' object has no
attribute 'closed'" in <bound method atomictempfile.__del__ of
<mercurial.util.atomictempfile object at 0x00C4C790>> ignored

-----Original Message-----
From: Matt Mackall [mailto:mpm at selenic.com] 
Sent: Monday, January 07, 2008 3:44 PM
To: Ezra Smith
Cc: dustin at spy.net; mercurial at selenic.com
Subject: RE: branch.cache


On Mon, 2008-01-07 at 14:45 -0500, Ezra.Smith at bentley.com wrote:
> I see. Any ideas why that would be getting updated on a read/pull?

Pull does a changeset identifier lookup. As an identifier may be a
branch id, we have to check it against known branches. Finding known
branches is expensive - it requires potentially looking at every
changeset. So we keep a cache.

If this cache is out of date, we attempt to find new branches and update
the cache. If we can't write to the cache, we _should_ quietly fail.

If you're getting an exception here, please send us a copy.

-- 
Mathematics is the supreme nostalgia of our time.