Effort for upgrading from 1.3 to 1.5

[Konopelko, Pavel (P.)]

Pekka, Jason,

Jason Ekstrand wrote on 2014-03-20:
> 
> On Mar 20, 2014 9:59 AM, "Pekka Paalanen" <ppaalanen at gmail.com>
> wrote:
>> 
>> On Thu, 20 Mar 2014 13:31:31 +0000
>> "Konopelko, Pavel (P.)" <pkonopel at visteon.com> wrote:
>> 
>>> Hello everybody,
>>> 
>>> Question: Given that somebody has Wayland/Weston 1.3 already
>>> integrated in their system, what would it take to upgrade to the
>>> upcoming Wayland/Weston 1.5?  Is this just a matter of re-building it
>>> and everything will continue working out of the box?  Are there any
>>> adjustments in the graphics stack (drivers, EGL support, etc.) needed
>>> to support 1.5?  Are there any adjustments on the application side
>>> needed to work with 1.5?
>> 
>> Hi,
>> 
>> oh, 1.3 seems like ages ago, I can't remember. But, looking at the git
>> history of few selected files should be enlightening, for instance in
>> Wayland src/wayland-client.h and src/wayland-server.h, and also the
>> protocol specification in protocol/wayland.xml. In Weston
>> src/compositor.h.
>> 
>> For clients, things are kept backward-compatible. For Weston plugins,
>> change from 1.n to 1.n+1 is not guaranteed to be compatible.
>> 
>> Libwayland API is stable and backward-compatible. The Wayland protocol
>> is also kept stable, but we are moving from wl_shell to xdg_shell,
>> though that probably does not concern you.
>> 
>> On EGL front, you probably should be looking at the history of the
>> specification files in Mesa, e.g.:
>> 
> http://cgit.freedesktop.org/mesa/mesa/log/docs/specs/WL_bind_wayl
> and_display.spec
>> to get an overview.
>> 
>> I think application side should just work, all in all.
> 
> As Pekka said, from the libwayland and client-side everything
> *should* be stable.  We've put a lot of work into keeping things
> API and ABI-stable as we've changed them.  EGL implementations
> should be following the EGL API's that haven't changed.
> 
> On the weston side, things are not so stable.  In particular, the
> weston_view structure was added in 1.4 and this changed the
> weston scenegraph somewhat substantially.  In particulare, the
> concepts of "node in the scenegraph" and "client surface" were
> split apart so that the same surface can appear at multiple
> locations at once.  If you are writing your own shell plugin this
> change may be small or large depending on how the plugin is
> written.  My guess, given the little I understand about IVI is
> that it shouldn't be too hard, but it won't be trivial.

Thanks a bunch for your feedback.  I will go through the individual hints above to get a better picture.

>>> Background: The question is related to the weston-ivi-shell patch
>>> series submitted by Nobuhiko Tanibata.  This work is connected to the
>>> work that GENIVI [1] previously did in the area of IVI Layer
>>> Management. On the GENIVI side, the original plan was to propose the
>>> ivi- shell and the corresponding protocol to the Wayland project.
>>> However, it was expected that it will take time to review and agree
>>> the protocols.  Until then the plan provided for adopting an out-
>>> of-tree patch based on Wayland/Weston 1.3 for GENIVI purposes.  Once
>>> the ivi-shell would be integrated in Wayland/Weston, GENIVI would
>>> switch to the upstream version.
>>> 
>>> Now that the patch series was well accepted and that only a few
>>> changes were suggested to the original protocol, it would make more
>>> sense for GENIVI to align with the upstream right away.  The only
>>> catch is that the content of the next GENIVI release is being
>>> finalized right now and the decision must be done quickly. During the
>>> original reviews in GENIVI, using Wayland/Weston 1.3 as the basis was
>>> already approved.  The above question is an attempt to estimate the
>>> impact of adopting 1.5 instead.  Any hints and information on possible
>>> upgrade problems would be therefore greatly
> appreciated.
>> 
>> Sorry, I think "well accepted" may still be a slight overstatement at
>> the moment. It's just so hard to find time to review anything. When you
>> get a "looks good to me" then that's accepted by that particular
>> person, but even that is not enough to get into Wayland/Weston
>> upstream. For that you need to convince Kristian, who seems to be
>> extremely busy nowadays.
> 
> Agreed.  Personally, I like the direction the protocol has taken.
> Unfortunately, I haven't had time to review any of the code or the
> library api.  As Pekka said, to a certain extent, we're waiting on
> Kristian to be able to take a look at it and give an opinion on how it
> ought to fit into the Wayland ecosystem. However, he's been pretty busy
> with the X.org and GNOME merge windows coming up so I wouldn't expect a
> lot for a little bit yet.

OK, I see.  It's always tempting to declare a job done.


Thanks,
--Pavel Konopelko