It seems to have been inherited behavior from tinywl, but it's not clear
what purpose it serves, and it causes a couple of issues:
- A new absolute position that's discontinuous with the previous cursor
position can produce unexpectedly large relative motion deltas. This
can occur for example when multiple input devices are active, or in
nested/VM scenarios when the pointer leaves the windowed output and
re-enters at a different point.
- When the cursor position is locked via constraint, the computed deltas
continue to get larger as the absolute event position diverges further
from the locked position. This led to the mouse pointer going crazy in
applications that use the relative events, such as games under Wine/
Wayland.
Fixes an issue where the cursor would get stuck (immovable) outside the
window of a Wine/Wayland game, if it was already outside when the game
started (common with a 4:3 game on a 16:9 screen). Now one can manually
move the cursor into the game window, at which point it becomes locked.
This is a minimal/interim fix. Ideally we should warp the cursor into
the constraint area automatically, but that would be a bit more work.
The change to apply_constraint() just turns an assert-failure into a
safe no-op return, since the function is now entered for "locked" as
well as "confined" constraint types.
Background:
I rewrote the config parser in 9462457..2f414a4, but it broke certain
configurations by changing how empty strings are handled: they were mostly
just ignored before my parser rewrite, but after that, they are
interpreted as just empty strings (output="" is considered as 'output named ""').
Though that was unintentional, I believe ignoring empty strings was not a
good idea in the first place, as we already allow empty strings for
certain configurations (e.g. `<desktop prefix="">`), which makes the
parser's behavior inconsistent.
Change:
So let's clarify that we intend to read empty strings as empty strings.
As a preparation, this commit adds warnings for empty strings we are
currently ignoring, so that users can be informed that we intend to just
read empty strings (e.g. `<theme name="">`) as empty strings in the future.
I removed existing empty strings in `rc.xml.all` to avoid warnings when
reading it.
This introduces an optional "direction" argument to the Resize action,
mirroring Fluxbox's StartResizing [corner] behavior.
Supported values (case-insensitive) are:
up-left, up, up-right, left, right, down-left, down, down-right.
If no direction is specified, the existing behavior is preserved and the
resize edges are inferred from the current pointer position. The action
documentation has been updated to describe the new argument.
I think `<windowSwitcher><osd output="keyboard">` is a bit unclear and
hard to interpret as "show OSD in the output with keyboard focus". Also,
we use "cursor" instead of "pointer" in other configurations like
`<placement policy="cursor">` and `<action name="ShowMenu" atCursor="">`.
So let's replace `output="all|pointer|keyboard"` with
`output="all|cursor|focused"`. In documentation, I reordered them to
`output="all|focused|cursor"` as "focused" feels like a bit more
sophisticated and general policy.
No one remembers why we were doing this, and we do not handle the
ping_timeout event, so it seems pointless. Also it appears to confuse
some older X clients.
This allows changing the cycled order in the future, e.g. focused order vs
created order.
Functionally, this commit also changes the initially selected window;
before this commit, the previous/next of the topmost window was always
selected, but now the previous/next of the active window is selected first
if it is in the cycled list. This won't change behaviors for most users,
but this ensures that the user can go back to the focused window with
Alt-Tab + Alt-Shift-Tab even when it is not the topmost window.
This commit fixes the TODO in the previous commit by trying to preserve
the selected view when a view is destroyed during window cycling.
This commit clarifies the lifecycle of the window switcher (cycle) by:
- init_cycle(): initializes the window switcher (e.g. OSD).
- update_cycle(): updates the window switcher states including OSD,
preview and outlines.
- destroy_cycle(): clears all the window switcher states.
This commit temporarily regresses by not trying to preserve the selected
view when a view is destroyed. This will be addressed in the next commit.
With this approach, we only need to store cycle->dummy_node instead of
cycle->preview_anchor and cycle->preview_parent and we don't need to care
about annoying cases like when the previewed window has no siblings or
when a window tracked by cycle->preview_anchor is destroyed.
This fixes a known regression in 885919fc that cursor entering the
titlebar (and other SSD parts) doesn't update the keyboard focus even when
followMouse=yes.
We were using the word "osd" to describe the window switcher, but it can
be used with on-screen display (OSD) disabled by
`<windowSwitcher><osd show="false">`. Let's use "cycle" instead to avoid
confusion.
- connect/disconnect map handlers in set_surface()
- call set_surface() at time of associate/dissociate
This separates the concepts of "associate" and "map" more clearly.
It's no longer necessary to listen for wlr_surface "destroy" event,
because dissociate is always received first.
Also, view->content_tree is now destroyed and set to NULL at unmap.
Previously, we relied on wlr_scene to destroy it automatically when
the surface was destroyed, but kept a potentially dangling pointer in
view->content_tree until next map. Similar change for unmanaged.
v2: comment updates
We were only using it to allow quick bitset comparisons of sets of
outputs (such as view->outputs). We can maintain our own bit IDs for
this purpose and avoid using the private wlroots field.
Note: from my reading of wlr_scene_output_create(), it appears to
always take the lowest unused index, resulting in aggressive re-use of
index values when outputs are disconnected and reconnected. I've tried
to make re-use as infrequent as possible. This could theoretically
reduce the chance of a mix-up in view_update_outputs(), although I'm
not aware of any practical scenario where it matters.
v2: prevent adding more than 64 outputs
Map/unmap logic is currently re-used for minimize/unminimize, but lots
of it doesn't actually apply in that case. This is both confusing and
creates some extra complexity, such as:
- extra "client_request" parameter to unmap(), in which case it has to
still do some cleanup even if view->mapped is already false
- various "view->mapped || view->minimized" checks when we really just
mean "is the view mapped"
To clean this all up, let's put the logic that really is common into
a new view_update_visiblity() function, and stop using map/unmap for
minimize/unminimize.
Note that this changes the meaning of "view->mapped", which used to
mean "mapped and not minimized" but now really just means "mapped".
I left some "view->mapped" conditions as-is (rather than changing to
"view->mapped && !view->minimized") where it seemed to make sense.
v2: add view_update_visibility() as suggested by tokyo4j
Two were missing: "minimized" and "activated".
At least "minimized" can be set before map, so the initial state needs
to be read and set on the foreign-toplevel.
I think in the current code, the foreign-toplevel is always created
before a view is activated, but for future-proofing, let's not rely
on this, and set "activated" as well. There's no harm since wlroots
optimizes away any redundant state changes.
Since d7dc6e01b4 ("Chase wlroots: Unified mapping"), the map event
is emitted by the wlr_surface itself, thus it cannot be NULL.
Change the "if" to an "assert" accordingly.
Call configure_libinput() for keyboards in new_keyboard() and
seat_reconfigure().
This allows the use of the sendEventsMode configuration option on
keyboards in order to disable keyboard input, for example:
<libinput>
<device category=" RPI Wired Keyboard 1">
<sendEventsMode>no</sendEventsMode>
</device>
</libinput>
