Bzr 'time' measurement

[Martin Pool]

On 08/04/2006, at 4:12 AM, Will Lentz wrote:

> Hi,
>
> I'm looking for help on how to debug a bzr 0.7 speed problem on my
> redhat machine.
>
> The problem only happens once at the start of each day.  The first  
> time
> I run something like 'time bzr st' or 'time bzr info' the "real"  
> time is
> always much larger than the "user+sys" time.  For the rest of the day
> the "real" and "user" time are in pretty close agreement.  I have
> multiple bzr-versioned directories, and the problem happens on each
> directory (i.e., running 'bzr info' on 1 directory doesn't speed up
> another separate bzr-versioned directory).

As Aaron says, this is typical of needing to bring things in to  
memory from disk.

0.8pre has better much internal mechanisms (called the hash-cache) to  
avoid needing to read so many files first thing in the morning.   
We'll be making a beta soon; it'd be good if you could try that and  
see how it goes.

-- 
Martin