[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: rev 3214 - branches/issue-749-caching/subversion/libsvn_wc

From: Philip Martin <philip_at_codematters.co.uk>
Date: 2002-09-24 03:10:55 CEST

philip@tigris.org writes:

> Author: philip
> Date: 2002-09-23 19:35:01 -0500 (Mon, 23 Sep 2002)
> New Revision: 3214
>
> Modified:
> branches/issue-749-caching/subversion/libsvn_wc/adm_ops.c
> branches/issue-749-caching/subversion/libsvn_wc/entries.c
> branches/issue-749-caching/subversion/libsvn_wc/lock.c
> Log:
> Enable full read-write entries caching. Hey, it passes the regression
> tests - commit it!

First the good news. Check-out of a repository containing 200 files in
a single directory runs twice as fast with read-write entries caching.
It doesn't exhibit the gradual slowdown associated with repeatedly
parsing the entries file, although it is still repeatedly writing the
file.

Now the bad news. Memory usage is a problem. Access batons tend to
live for the duration of operations like checkout, so the caches have
to use long-lived pools and memory usage tends to increase with the
size of the working copy. Check-out of the subversion trunk uses over
100MB with APR pool debugging enabled. Now with a bit of work we can
probably reduce the rate at which it grows, but I don't immediately
see how to eliminate the growth. It's not really surprising, when I
first considered using access batons to cache entries I suggested that
it might be a problem

http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgId=175861

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Sep 24 03:11:31 2002

This is an archived mail posted to the Subversion Dev mailing list.