| graesslin | |
| mart |
| Plasma |
Adds XDGV6 support for the kwin side.
Popup placement support is limited to the stuff v5 had, a simple offset, rather than the awesome new positioner. But Qt doesn't make use of it yet either. Also ideally we should do all the positioning before sending the first configure, but again Qt doesn't actually do anything with that anyway.
Some menus are broken, but they are broken with wl-shell integration too.
gtk3-demo works nicely.
| Automatic diff as part of commit; lint not applicable. |
| Automatic diff as part of commit; unit tests not applicable. |
Ping. Even if I now can't merge till 5.11 splits, it'd be nice to have this approved.
There's still a lot of work to be done on top of this, which is currently blocked.
| shell_client.cpp | ||
|---|---|---|
| 236 | maybe instead of the three casts just store a local variable? auto global = (static_cast<XdgShellInterface *>(m_xdgShellSurface->global()); | |
| 245–253 | You should only kill if the ping reason is close. This would also ping for focus, wouldn't it? | |
| 297–300 | this needs to go through our PointerInputRedirection code, otherwise SeatInterface and PointerInputRedirection get out of sync. From my understanding that should be ::popupGrab? | |
| 642–645 | just wondering: isn't there a close() missing? It's only pinging... | |
| shell_client.h | ||
| 52 | please use enum class for new enums. | |
| shell_client.cpp | ||
|---|---|---|
| 642–645 | the close is actually beaing done at line 244: at the pongReceived (if the ping reason was closewindow, then it closes it there) | |
| shell_client.cpp | ||
|---|---|---|
| 297–300 | Ack, Initially I thought this event could happen at any time, but on re-reading it's to separate tooltips and menus, at which point it's on of the initial properties that should be sent before the first buffer. (I'll try to get that clarified in the spec for v7) At which point it falls in nicely into how kwin already handles things checking hasPopupGrab when the toplevel is added. | |
| shell_client.cpp | ||
|---|---|---|
| 642–645 | i copied from the x11 client, where on client::CloseWindow it pings beforehand, but i can remove it and just close | |
i removed the ping to close window.. however it still doesn't sound right, as the client could not answer a close request, there may need a close and a ping at the same time, then whoever wins, either the window gets gracefully closed or brutally killed
now pings and tries to close at the same time, so either will be closed cleanly or killed if not responsive
Now it looks good to me!
What I would like to see is unit tests for all of that. This should be fairly simple as the tests are prepared to be run for different kind of shell surfaces. Many tests already have a _data method where it is set to be run for wl_shell and xdg_shell_unstable_v5. This should be easy to be extended for additional v6.
I really would like to see this go in now into master! But we need to have at least a few unit tests. If that's too difficult I can write them, but I would prefer if one of you extends it. It's really not difficult in this case as we just need to add the basics to run the same tests also as xdg shell v6.
It's really not difficult
Ha. Extending those tests was not dificult, and if you look you'll see that was committed in the XDG branch mid last week.
But we also have ping mixed in here (something I now regret), and that proved much more "interesting".
Socket mode would have a fun crash as killWindow() will implicitly delete "this", also in the process I made a made the kill test represent a far more real blocked app, and that broke the existing killWindow socket mode test, as removing the connection doesn't actually kill a frozen app.
All fixed now.