in-database properties

[Greg Stein <gs...@gmail.com>]

Hey all,

After my recent rounds of work, it now appears that we're "ready" to
move to in-database properties.

If you weren't aware, we record properties into the old-style files
*and* into the database. There is a TEST_DB_PROPS define in wc/props.c
that will test/assert that these properties are in lock-step each time
we try to read properties. Except for a couple probably-minor failures
in upgrade_tests, the entire test suite now passes when this define is
set.

Thus, we should be ready to being reading props only from the
database, and stop writing out the old-style props files.

To that end, I'm going to bump us from today's "format 16" to format
17 some time this week. This will be an auto-upgrade process, so there
will be no need to use 'svn upgrade'. f17 will also include the
introduce of .svn/pristine/ for the work Julian is doing.

Please let me know if you have any questions/concerns.

Cheers,
-g

Re: in-database properties

["Hyrum K. Wright" <hy...@mail.utexas.edu>]

On Tue, May 11, 2010 at 6:12 AM, Stefan Sperling <st...@elego.de> wrote:

> On Tue, May 11, 2010 at 05:03:00AM -0400, Greg Stein wrote:
> > Hey all,
> >
> > After my recent rounds of work, it now appears that we're "ready" to
> > move to in-database properties.
> >
> > If you weren't aware, we record properties into the old-style files
> > *and* into the database. There is a TEST_DB_PROPS define in wc/props.c
> > that will test/assert that these properties are in lock-step each time
> > we try to read properties. Except for a couple probably-minor failures
> > in upgrade_tests, the entire test suite now passes when this define is
> > set.
>
> yayayayay!
>
+1

This is a big step, and it looks as if a number of other issues have been
fixed along the way.  Any idea what the immediate performance benefits are?

-Hyrum

Re: in-database properties

[Stefan Sperling <st...@elego.de>]

On Tue, May 11, 2010 at 05:03:00AM -0400, Greg Stein wrote:
> Hey all,
> 
> After my recent rounds of work, it now appears that we're "ready" to
> move to in-database properties.
> 
> If you weren't aware, we record properties into the old-style files
> *and* into the database. There is a TEST_DB_PROPS define in wc/props.c
> that will test/assert that these properties are in lock-step each time
> we try to read properties. Except for a couple probably-minor failures
> in upgrade_tests, the entire test suite now passes when this define is
> set.

yayayayay!