[PATCH weston 17/17] xdg-shell: Bump unstable version to 6

[Jonas Ådahl]

On Thu, Apr 30, 2015 at 06:54:34PM +0300, Pekka Paalanen wrote:
> On Tue,  7 Apr 2015 17:01:32 +0800
> Jonas Ådahl <jadahl at gmail.com> wrote:
> 
> > Signed-off-by: Jonas Ådahl <jadahl at gmail.com>
> > ---
> >  clients/simple-damage.c | 2 +-
> >  clients/simple-egl.c    | 2 +-
> >  clients/simple-shm.c    | 2 +-
> >  clients/window.c        | 2 +-
> >  desktop-shell/shell.c   | 2 +-
> >  protocol/xdg-shell.xml  | 2 +-
> >  6 files changed, 6 insertions(+), 6 deletions(-)
> 
> Trying to remember all the reviews I just sent for this series, there
> was not a single case which absolutely requires bumping the version
> here.
> 
> Not bumping the version would be very attractive, because it would let
> wider choice between Gtk+ etc. versions for people who do already rely
> on xdg-shell working. (Even if you might claim they shouldn't, but they
> do.)
> 
> Is it reasonable to expect that nothing will seriously break if we
> don't bump the version, and people mix and match implementations before
> and after this patch series?
> 
> I do not care about build failures, those are easily patched, and I
> assume projects carry their own copies of xdg-shell.xml anyway, so
> upgrading means work in any case. All I care is the "binary"
> compatilibity. Like said, changing error codes is also not a reason,
> because errors are fatal anyway.
> 
> If you do see the need for bumping the version, I really really hope we
> can get it in before 1.8-alpha. It's such a big breakage, that I'd hate
> to do it later when we are trying to stabilize a Weston release.

Already discussed this on IRC, but anyway: the reason for the bump was
because the error conditions were moved. They change ABI, but only in an
already fatal path, so lets drop that patch.


Jonas

> 
> Sorry for getting here sooner.
> 
> 
> To anyone wondering: xdg-shell uses a special experimental versioning
> scheme, where every bump literally breaks everything. It is not
> following the normal Wayland interface versioning rules yet, and
> likely won't be for quite a long time.
> 
> 
> Thanks,
> pq