Re: Format bump to f21

"Hyrum K. Wright" <hyrum_wright_at_mail.utexas.edu> writes:

> As part of moving tree conflict information storage to the victim
> nodes, I'll be bumping the format to f21 shortly. It is pretty
> unobtrusive, and the upgrade code and test has already been committed
> in r1034155. The upgrade should be painless for folks, but I figured
> I'd just give fair warning.
>
> If nobody has concerns, I'll do the bump tomorrow.

This allows actual nodes without base or working nodes. We need to
decide what

   svn_wc__db_read_info(&status, .... )

should do for such nodes; there is no sensible status for such nodes.
We probably don't need to do this before the upgrade but we do need to
have a plan for how it should behave.

In the past it returned an error: SVN_ERR_WC_CORRUPT if there was an
actual node with no base or working, or SVN_ERR_WC_PATH_NOT_FOUND if
there was no node.

In the new code we do

   svn_wc__db_read_info(NULL, .... &conflicted, ...)

and expect to get the conflict status of actual-only nodes. I've made
svn_wc__db_read_info return SVN_NO_ERROR and set status to normal but
this should be regarded as temporary. Maybe we should only return an
error if the caller asks for status? That doesn't sound like an easy
interface to use. Maybe we should introduce a new API for getting
conflicted? That might mean multiple calls to get the node information
which doesn't sound good either.

-- 
Philip