Comment 3 for bug 309374

On Mon, 2009-03-23 at 02:50 +0000, David Reitter wrote:
> Same problem in 1.13. In the example below (unlike in what I show
> above), the file in question has only been created in my branch; the
> branch that it is stacked on shouldn't know anything about this file.

We have a major overhaul of disk storage in progress at the moment in
the brisbane-core project; this is expected to have a beta quality
format available in bzr 1.14; performance issues should be tested on
that format (and we are testing with copies of openoffice, emacs, and
mysql) - so I fully expect we'll get some motion on this.

However, right now nearly all our efforts are focused on that, so its
likely that we won't get down to examining the log performance issue
you're seeing here in detail until the new format is in beta.

If you have the time, one thing you could do to help (because there may
be a simple fix we can do) would be to get an lsprof trace:

bzr --lsprof-file foo.callgrind <your operation here>, and attach the
resulting callgrind file - it will tell us what is taking the excess
time in your operation.

Thanks,
Rob