> -----Original Message-----
> From: Hyrum K. Wright [mailto:hyrum_wright_at_mail.utexas.edu]
> Sent: woensdag 16 september 2009 16:36
> To: Bert Huijben
> Cc: dev_at_subversion.tigris.org
> Subject: Re: svn commit: r39355 - trunk/subversion/libsvn_wc
> On Sep 16, 2009, at 4:11 AM, Bert Huijben wrote:
> > Author: rhuijben
> > Date: Wed Sep 16 02:11:16 2009
> > New Revision: 39355
> > Log:
> > Following up on r39338, r39340, add the option of flushing the
> > cache to the remove entry and set depth database calls.
> > Suggested by: gstein
> While I agree that the entries need to be flushed in these cases
> (since we are doing a write of the DB), I'm *very* leery about doing
> it conditionally. It smacks of a premature optimization, and that's
> the type of thing that we've spent *months* attempting to eliminate in
> wc-1. Plus, doing this in only one function introduces an API
> consistency across the wc_db APIs. I understand the desire to use the
> cache, and not flush it prematurely, but in this case I don't think it
> should be conditional.
Flushing always is not an option in this _temp_ function (see its name).. unless you just silently fixed all callers from svn_wc__set_depth and svn_wc__entry_remove to no longer expect an entries cache?
This temp function should be removed when we are done removing the entry write locations, but it is now a vital part of the update editor to allow removing other dependencies on the entries cache.
Maybe the code should be moved into entries.c?
The function was designed to have just that single caller until we can remove that caller completely.
Received on 2009-09-16 16:51:23 CEST