The fact that we have an undocumented hack like that in our scanner clearly isn't great. We need to document it. <div class="gmail_quote">On Sep 3, 2014 8:55 PM, "Boyan Ding" <<a href="mailto:stu_dby@126.com">stu_dby@126.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Hi, It is actually not a fault in wayland, instead it is designed to be so. new_id's without interface specified in the protocol (such as the one in wl_registry::bind) must come with 3 arguments (s: interface name, u: version, n: the actual new_id). That's why 'n' turns into 'sun'. If a language binding generates the wrong code, please contact the author of the language binding and let him correct it. If you are changing how wayland-scanner (the official C code generator) works, the whole world will break before you. Refer to "How protocol objects are created" section in [1] to see the details. Regards, Boyan Ding [1] <a href="http://ppaalanen.blogspot.fi/2014/07/wayland-protocol-design-object-lifespan.html" target="_blank">http://ppaalanen.blogspot.fi/2014/07/wayland-protocol-design-object-lifespan.html</a> On Wed, 2014-09-03 at 22:32 -0500, Paul Sbarra wrote: > I tried to start some discussion on this topic previously, but it > apparently didn't make it through the moderator, so I'm trying again > having now joined the list. > > > I've recently taken an interest in the gowl implementation of the > wayland protocol and noticed that the specification doesn't match the > interface that gets generated by the wayland scanner. > > > If I attempt to code-gen gowl using the wayland.xml file the > wl_registry bind interface is missing a couple arguments that weston > expects, resulting in the following runtime error: > libwayland: message too short, object (2), message bind(usun) > > > However, the spec indicates a bind request signature of "un". > > > I tracked this down into some curious logic in the scanner and have > been working on a patch to try and allow the corrected signature to be > specified in the protocol. Unfortunately this has become more > interesting then I'd anticipated. > > > Attached is my naive attempt at resolving this issue. I'm horrified > by the string->interface lookup and the wl_registry_bind api change > but I'm not sure what else one can do. Any feedback or additional > help would be appreciated. > > > Note that this patch applies to the 1.5.91 tagged commit and compiles > cleanly (for wayland). Weston didn't like having the lookup function > defined in multiple files, so maybe there would be a better place to > put such functionality. > > Thanks, > Paul > > _______________________________________________ > wayland-devel mailing list > <a href="mailto:wayland-devel@lists.freedesktop.org">wayland-devel@lists.freedesktop.org</a> > <a href="http://lists.freedesktop.org/mailman/listinfo/wayland-devel" target="_blank">http://lists.freedesktop.org/mailman/listinfo/wayland-devel</a> _______________________________________________ wayland-devel mailing list <a href="mailto:wayland-devel@lists.freedesktop.org">wayland-devel@lists.freedesktop.org</a> <a href="http://lists.freedesktop.org/mailman/listinfo/wayland-devel" target="_blank">http://lists.freedesktop.org/mailman/listinfo/wayland-devel</a> </blockquote></div>