philip@tigris.org writes:
> Author: philip
> Date: 2002-09-22 10:03:46 -0500 (Sun, 22 Sep 2002)
> New Revision: 3206
[snip]
> Continuing issue 749. Pass the access baton to svn_wc_entries_read,
> svn_wc_entry and svn_wc__entries write. Incoroprate an idea from
> Benjamin Pflugmann <benjamin-svn-dev@pflugmann.de> to handle revert
> of schedule add directories that are missing.
>
> Introduce a simple entries caching mechanism. This is currently
> disabled as it's experimental and has an adverse effect on client
> memory use. However the regression tests pass when it's enabled.
> Compile with SVN_WC_ADM_CACHE_ENTRIES defined to try it.
I didn't originally intend to introduce caching with this patch, but
it was so simple I could not resist trying! For the intrepid amongst
you, if you want to try it you need to rebuild libsvn_wc/lock.c and
libsvn_wc/entries.c with the pre-processor symbol mentioned in the log
message.
Caching is not complete, it works best for read-only operations.
Things like 'svn diff' and 'svn st' on a Subversion trunk working copy
run four or five time faster on my machine. The the regression test
suite uses about %5 less CPU and runtime. That's not such a big
change because not all the tests use libsvn_wc, the working copies are
small and the caching is less effective on read-write operations like
checkout and update.
Heed the warning about memory use, I really mean it! If you try to
checkout a Subversion working copy with caching enabled you may well
run out of memory.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Sep 22 17:39:12 2002