On Fri, Mar 22, 2013 at 05:52:10PM +0200, Daniel Shahaf wrote:
> C. Michael Pilato wrote on Fri, Mar 22, 2013 at 11:25:29 -0400:
> > On 03/22/2013 11:13 AM, Daniel Shahaf wrote:
> > > svnadmin.c has this code:
> > >
> > > /* We can't create repository with a version newer than what
> > > the running version of Subversion supports. */
> > > if (! svn_version__at_least(&latest,
> > > compatible_version->major,
> > > compatible_version->minor,
> > > compatible_version->patch))
> > > {
> > > err = svn_error_createf(SVN_ERR_UNSUPPORTED_FEATURE, NULL,
> > > _("Cannot guarantee compatibility "
> > > "beyond the current running
> > > version "
> > > "(%s)"),
> > > SVN_VER_NUM );
> > > return EXIT_ERROR(err);
> > > }
> > >
> > > Should the condition be relaxed to (compatible_version->major !=
> > > latest->major)?
> > >
> > > After all, if someone passes "1.9.0", that's fine --- we'll just create
> > > a 1.8.0-style repository, and our compatibility rules imply that 1.9.0
> > > will know to read it. Similarly, if someone passes "1.8.2", then
> > > regardless of whether this svnadmin is 1.8.0 or 1.8.4, e can just go
> > > ahead and create a 1.8-stlye repository; 1.8.2 will be able to read it.
> > >
> > > In other words, the "is my version number at least X" behaviour might be
> > > useful, but --compatible-version= doesn't seem to be the right place for
> > > it.
> >
> > Makes sense to me.
>
> Cool. r1459849. (I committed that before Julian replied.)
... and reverted since IRC demonstrated conflicting opinions.
