On Mon, Jun 25, 2012 at 9:14 PM, Stefan Sperling <stsp_at_elego.de> wrote:
> On Mon, Jun 25, 2012 at 09:03:47PM +0200, Stefan Sperling wrote:
>> on the same working copy. E.g. a 1.7 client might run into tree conflicts
>> which it cannot understand because a 1.8 client flagged a conflict involving
>> a move. I believe we should bump to avoid such problems.
>
> FYI, here is what this looks like:
>
> With trunk:
>
> $ svn status
> ! C alpha
> > local moved away and edited, incoming delete upon update
> A + alpha2
> Summary of conflicts:
> Tree conflicts: 1
>
> With 1.7.x:
>
> $ svn status
> subversion/svn/status-cmd.c:344: (apr_err=155016)
> subversion/svn/util.c:981: (apr_err=155016)
> subversion/libsvn_client/status.c:490: (apr_err=155016)
> subversion/libsvn_wc/status.c:2421: (apr_err=155016)
> subversion/libsvn_wc/status.c:2421: (apr_err=155016)
> subversion/libsvn_wc/status.c:1200: (apr_err=155016)
> subversion/svn/status.c:210: (apr_err=155016)
> subversion/svn/status.c:210: (apr_err=155016)
> subversion/libsvn_wc/wc_db.c:5814: (apr_err=155016)
> subversion/libsvn_wc/wc_db.c:5814: (apr_err=155016)
> subversion/libsvn_wc/tree_conflicts.c:249: (apr_err=155016)
> subversion/libsvn_wc/tree_conflicts.c:130: (apr_err=155016)
> svn: E155016: Unknown enumeration value in tree conflict description
>
> I don't see a way to avoid this problem for 1.7 clients, apart from either
> reverting the tree conflict description changes or bumping the format.
Okay I see. That pretty much breaks it (unless we backport the
"unknown enumeration value" to 1.7, but that feels a bit wonky (and we
can't keep backporting changes to older releases to keep them working
with newer wc formats forever anyway)).
How about the other way around? Make 1.8 work with 1.7-format wc's,
without requiring an upgrade. The upgrade would be optional, enabling
new features and improvements. Would that be at all possible? That
would make it possible to mix 1.7 and 1.8 svn clients, as long as the
format stays on 1.7-format.
I guess that would be pretty hard with our current codebase, to make
some (client-side) features conditional on wc-format. Not to mention
additional complexity to document this for users. But still, it's
worth a discussion I think. As a user, I'd definitely like to have
this flexibility ...
--
Johan
Received on 2012-06-25 23:02:00 CEST