[PATCH RFC wayland-protocols] unstable: add xcursor-configuration

Simon Ser contact at emersion.fr
Thu Oct 18 10:08:36 UTC 2018

On Tuesday, October 16, 2018 2:10 PM, Johan Helsing <johan.helsing at qt.io> wrote:
> Hi Simon,
> Thanks for working on this! :)

Thanks for your feedback! I've sent a first version of this patch. I've
added per-device configuration and addressed your comments. Let me know what
you think.

> Is the intention to implement the client side inside libwayland-cursor? No
> changes needed for toolkits except calling some new API when initializing
> libwayland-cursor?

Yeah, this is what's planned. The new design of the libwayland-cursor API will
be the difficult part of this task. I'll try to send a proposal soon.

> I guess it's also possible to implement it on the toolkit level using the old
> libwayland-cursor API... Which would allow us to use the extension before
> everyone has upgraded to the new libwayland-cursor.

Aye, this is possible.

> > +    <request name="get_xcursor_configuration_seat">
> To me, this sounds a bit like we're getting some kind of seat from the
> request. How about get_seat_xcursor_configuration? Or simply
> get_xcursor_configuration?

This makes sense, I'll rename to get_device_xcursor_configuration. I'd like to
make sure we can add new requests to create xcursor configuration objects in
the future, so I'd like to keep "device" in there.

> > +    <event name="default_cursor">
> > +      <description summary="default cursor name">
> > +        The default_cursor event describes the default XCursor cursor name to be
> > +        used for this seat.
> > +      </description>
> Maybe explain the use case here?

The use-case is to be able to tell the difference between two cursors without
having to pick two different cursor themes. For instance one could choose
left_ptr for the pointer and cross for the tablet tool.

Do you think we should include this example in the protocol?

> And another thing, maybe we can explicitly say this is a hint for the client?
> I'm asking because I don't think we have a matching concept of a default
> cursor in Qt... We just have Qt::QCursorShape::ArrowCursor. We could
> make this request override what that shape is, but that might be confusing
> to application developers if some compositor decides to set the default
> cursor to a cross or some other non-arrow cursor. Anyway, this event is
> going to cause some minor issues in Qt, perhaps in other toolkits as well.

You're right, I've mentioned this in the protocol description. Everything in
this protocol is a hint, clients could use it to e.g. just choose the cursor
size and use a custom theme.

For Qt, you could e.g. use default_cursor if the app doesn't set the cursor
itself, and switch to Qt::QCursorShape::ArrowCursor if the app requests it.
Of course, having Qt::QCursorShape::DefaultCursor would be nice too.

> Otherwise, this looks great :) Looking forward to having uniform cursor
> sizes!
> Johan

More information about the wayland-devel mailing list