On Tue, Apr 23, 2013 at 8:18 PM, Philip Martin
<philip.martin_at_wandisco.com>wrote:
> Stefan Fuhrmann <stefan.fuhrmann_at_wandisco.com> writes:
>
> > On Tue, Apr 23, 2013 at 6:54 PM, Philip Martin
> > <philip.martin_at_wandisco.com>wrote:
> >> At the start of the log
> >> you say it is essential to use consistent configs. Does using NULL
> >> result in consistent configs?
> >>
> >
> > Consistent as in "same as in all other API calls potentially opening a
> FS".
> > You can pass NULL consistently to all of them (in which case the defaults
> > apply).
> >
> > svn_repos_verify_fs2 accepts an already open FS and will now pass the
> > same config to svn_fs_verify - whether NULL or not. It has been forced to
> > use the defaults in the past, which could be inconsistent with the ones
> used
> > by original FS instance.
>
> I guessed it was something like but I had to deduce that from the fact
> that this patch was written. I don't think the fs_config restrictions
> are clear to somebody reading the API documentation.
>
> If I have understood correctly: for any given filesystem all calls to
> svn_fs_create, svn_fs_open and svn_fs_verify and and svn_fs_new need to
> ensure that the same fs_config is used (perhaps NULL). I don't think
> that restriction is explicitly documented in svn_fs.h.
>
There is no such restriction. This is all about being *able* to
use the same settings everywhere. In particular, you want to
be able to use consistent settings through all stages of an
'svnadmin verify' and svn_repos_verify_fs2 now can and does
just that.
The only way to produce actually "wrong" results by inconsistent
FS configs is if one of them e.g. uses the default cache key
space ("namespace") when it intended to use a separate one.
-- Stefan^2.
--
*Join one of our free daily demo sessions on* *Scaling Subversion for the
Enterprise <http://www.wandisco.com/training/webinars>*
*
*
Received on 2013-04-23 20:30:52 CEST