Commit Graph

2155 Commits

Author SHA1 Message Date
Simon Ser
b264ec7767 linux-dmabuf-v1: add basic helpers for feedback 2023-02-06 19:14:15 +00:00
Simon Ser
5cd14dfc91 linux-dmabuf-v1: stop using const in feedback
Instead of using const pointers, use structs owned by the tranche.
This will allow wlroots to expose helpers to build feedback
objects.
2023-02-06 19:14:15 +00:00
Consolatis
f6008ffff4 xdg-activation: accept pointer focus for new tokens
This patch allows surfaces without keyboard focus
but with pointer focus to receive valid tokens.

This can be relevant for applications using an older version of
the layershell protocol which only provided the choice between
no keyboard focus and exclusive keyboard focus.
2023-02-06 15:52:02 +01:00
Alexander Orzechowski
7b32c25a4f wlr_scene: Rename wlr_scene_surface_from_buffer
This renames it to wlr_scene_surface_try_from_buffer to be more clear
that this function can return NULL. This is inline with the rest of
wlroots[1].

[1] https://gitlab.freedesktop.org/wlroots/wlroots/-/merge_requests/3991
2023-02-04 02:57:42 -05:00
Kirill Primak
dc7b5ab21e layer-shell: forbid changing the existing popup parent 2023-02-03 19:29:11 +03:00
Simon Ser
92eedb84c1 output: don't attach buffer on first commit if disabled
In output_ensure_buffer() we create a swapchain and attach an empty
buffer to the output if necessary. We do that during the first commit.
This is fine when the first commit enables the output, however this breaks
when the first commit disables the output. A commit which disables an
output and has a buffer attached is invalid (see output_basic_test()), and
makes the DRM backend crash:

    00:00:00.780 [wlr] [backend/drm/drm.c:622] connector eDP-1: Turning off
    ../subprojects/wlroots/backend/drm/drm.c:652:44: runtime error: member access within null pointer of type 'struct wlr_drm_crtc'
    AddressSanitizer:DEADLYSIGNAL
    =================================================================
    ==2524==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7f22e894afc1 bp 0x7ffe1d57c550 sp 0x7ffe1d57c420 T0)
    ==2524==The signal is caused by a READ memory access.
    ==2524==Hint: address points to the zero page.
        #0 0x7f22e894afc1 in drm_connector_commit_state ../subprojects/wlroots/backend/drm/drm.c:652
        #1 0x7f22e894b1f5 in drm_connector_commit ../subprojects/wlroots/backend/drm/drm.c:674
        #2 0x7f22e89e8da9 in wlr_output_commit_state ../subprojects/wlroots/types/output/output.c:756
        #3 0x555ab325624d in apply_output_config ../sway/config/output.c:517
        #4 0x555ab31a1aa1 in handle_new_output ../sway/desktop/output.c:974
        #5 0x7f22e9272f6d in wl_signal_emit_mutable (/usr/lib/libwayland-server.so.0+0x9f6d)
        #6 0x7f22e899b012 in new_output_reemit ../subprojects/wlroots/backend/multi/backend.c:161
        #7 0x7f22e9272f6d in wl_signal_emit_mutable (/usr/lib/libwayland-server.so.0+0x9f6d)
        #8 0x7f22e895a153 in scan_drm_connectors ../subprojects/wlroots/backend/drm/drm.c:1488
        #9 0x7f22e893c2e4 in backend_start ../subprojects/wlroots/backend/drm/backend.c:24
        #10 0x7f22e892ed00 in wlr_backend_start ../subprojects/wlroots/backend/backend.c:56
        #11 0x7f22e8999b83 in multi_backend_start ../subprojects/wlroots/backend/multi/backend.c:31
        #12 0x7f22e892ed00 in wlr_backend_start ../subprojects/wlroots/backend/backend.c:56
        #13 0x555ab317d5cc in server_start ../sway/server.c:316
        #14 0x555ab317748d in main ../sway/main.c:400
        #15 0x7f22e783c28f  (/usr/lib/libc.so.6+0x2328f)
        #16 0x7f22e783c349 in __libc_start_main (/usr/lib/libc.so.6+0x23349)
        #17 0x555ab3134c84 in _start (/home/simon/src/sway/build/sway/sway+0x377c84)

Fixes: 3be6658ee7 ("output: allocate swapchain on first commit")
Closes: https://github.com/swaywm/sway/issues/7373
2023-02-02 16:48:35 +00:00
Kirill Primak
911648f430 subcompositor: fix mapped parent check 2023-02-02 17:26:25 +03:00
Simon Ser
eb3e8f08a8 subcompositor: convert to try_from
References: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/884
2023-02-01 20:14:39 +01:00
Simon Ser
c5f7f8ab98 ext-session-lock-v1: convert to try_from
References: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/884
2023-02-01 20:14:12 +01:00
Simon Ser
f9bd416d41 layer-shell-v1: convert to try_from
References: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/884
2023-02-01 20:13:53 +01:00
Simon Ser
49cb85ad72 input-method-v2: convert to try_from
References: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/884
2023-02-01 20:13:30 +01:00
Simon Ser
711a1a3ed4 xdg-shell: convert to try_from
References: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/884
2023-02-01 20:12:49 +01:00
Simon Ser
0f24d27e0b xdg-shell: rename wlr_xdg_popup.committed to sent_initial_configure
We made a similar change to wlr_xdg_toplevel.
2023-01-31 16:27:05 +01:00
Simon Ser
df0c926a7a xdg-shell: rename wlr_xdg_toplevel.added to sent_initial_configure
This is more descriptive, and avoids the confusion with
wlr_xdg_surface.added.
2023-01-31 16:20:33 +01:00
Simon Ser
e1a1dc4be5 xdg-shell: remove unnecessary return
We're at the end of the function anyways.
2023-01-31 16:20:33 +01:00
Simon Ser
e69dd909f7 xdg-shell: reset added/committed flag on unmap
When a client attaches a NULL buffer to its wl_surface, it's
unmapped. This resets the xdg_surface in its initial state. An
extra NULL commit makes the compositor send an initial configure
event.

Note, wlr_xdg_toplevel.added is separate from wlr_xdg_surface.added.

Closes: https://github.com/swaywm/sway/issues/7397
2023-01-31 16:20:10 +01:00
Simon Ser
0e54b861e9 seat/keyboard: constify wlr_seat_keyboard_notify_enter() 2023-01-26 15:06:45 +00:00
Simon Ser
695d28c250 seat/keyboard: constify wlr_seat_keyboard_notify_modifiers() 2023-01-26 15:06:45 +00:00
Simon Ser
e44834595d seat/keyboard: constify args in wlr_seat_keyboard_enter() 2023-01-26 15:06:45 +00:00
Simon Ser
12e04d8f83 seat/keyboard: constify wlr_seat_keyboard_send_modifiers() 2023-01-26 15:06:45 +00:00
Simon Ser
f41dcb3a97 seat/keyboard: skip wl_array allocation
We can just create a wl_array from the array passed by the user.
Here, wl_array is only necessary to make libwayland happy.
2023-01-26 15:06:45 +00:00
Alexander Orzechowski
5007e713b4 wlr_scene: Send intersecting list of scene outputs for outputs_update signal 2023-01-21 16:23:30 -05:00
Kirill Primak
843b874f22 output-layout: improve closest point for no outputs case
Without outputs, all points are equally invalid anyway, but for e.g.
cursor warping it makes more sense to preserve the original position.
2023-01-21 12:48:43 +00:00
Alexander Orzechowski
5f264a7d6c subcompositor: Clean up subsurface_consider_map
check_parent was unecessary: it only served to skip a trivial check
and added more complexity than it was worth.
2023-01-04 04:24:17 -05:00
Alexander Orzechowski
84aee1d708 subcompositor: Clarify comment on subsurface mapped consideration 2023-01-04 04:24:17 -05:00
Isaac Freund
869af1cfbb
session-lock: send more protocol errors
The invalid_destroy and invalid_unlock protocol errors aren't currently
sent by wlroots and instead left up to the compositor. However, we can
handle these as well without much additional complexity.

This also adds a missing wl_resource_destroy() call if the lock is inert
in lock_handle_unlock_and_destroy().
2022-12-23 16:56:42 +01:00
Kirill Primak
b28a9afd4b presentation-time: finish surface addon
Fixes: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/3554
2022-12-22 18:35:53 +03:00
bi4k8
790fb93794 cursor: compare to scaled coords 2022-12-20 20:01:56 +00:00
xiliuya
7f6d646e0a keyboard: only update LEDs when changed 2022-12-14 17:51:06 +01:00
Kirill Primak
79248e4961 xdg-shell: use defunct_role_object error 2022-12-07 13:54:31 +00:00
Ronan Pigott
f8e70af318 wlr_xdg_activation_v1: add new_token event 2022-12-06 05:45:37 +00:00
Kirill Primak
097ea84cda output-layout: improve API
- wlr_output_layout_add{,_auto}() now return a bool indicating whether
  the function has succeeded.
- wlr_output_layout_move() is removed.

Closes: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/1083
2022-12-05 18:21:04 +00:00
Kirill Primak
27ea18cc33 output-layout: remove wlr_output_layout_output_state
wlroots uses "// private state" comments to denote structure fields
which shouldn't be accessed by compositors, so let's drop
wlr_output_layout_output_state and inline its fields into
wlr_output_layout_output; this also simplifies layout output creation.
2022-12-05 18:21:04 +00:00
Simon Ser
512deebd82 compositor: add wlr_surface.events.precommit 2022-12-05 10:51:46 +01:00
Simon Ser
d92f92f2d3 wl-drm: only advertise formats supporting implicit modifiers
With the Vulkan renderer we don't support implicit modifiers.
However wl_drm only supports implicit modifiers.

Stop advertising wl_drm formats when implicit modifiers are
unsupported.
2022-12-02 14:27:07 +00:00
Simon Ser
f36a5915da wl-drm: don't store wlr_renderer
Query the formats at init time, then forget about the renderer.
This will allow wl_drm to be created with a list of formats instead
of a renderer, and will behave better after a GPU reset.
2022-12-02 14:27:07 +00:00
Simon Ser
c8eb24d30e output: drop enable/mode events
The backend no longer changes the output state behind the
compositor's back. Instead, compositors can listen to the "commit"
event and check for WLR_OUTPUT_STATE_ENABLED/WLR_OUTPUT_STATE_MODE.

Closes: https://gitlab.freedesktop.org/wlroots/wlroots/-/issues/2300
2022-12-01 20:05:04 +00:00
Simon Ser
4287f70dc3 output-layout: stop listening for output mode events
These will go away.
2022-12-01 20:05:04 +00:00
Simon Ser
19dfe99496 output-damage: stop listening for output mode events
These will go away.
2022-12-01 20:05:04 +00:00
Simon Ser
fecbaa2bd6 scene: stop listening for output enable/mode events
These will go away.
2022-12-01 20:05:04 +00:00
Alexander Orzechowski
fe891ebd4e output_init_render: Allow re-initialization
This lets the compositor call this function after the fact to replace
the renderer/allocator after a renderer context lost.
2022-12-01 10:01:35 +00:00
Simon Ser
f103dc74d8 linux-dmabuf-v1: introduce wlr_linux_dmabuf_v1_create()
Some compositors may want to use the linux-dmabuf-v1 implementation
with a completely custom renderer. Add a function to create the
global with a default feedback.
2022-11-30 14:19:46 +01:00
Simon Ser
9c04fd2496 linux-dmabuf-v1: drop wlr_renderer field
This isn't used anymore.
2022-11-30 14:19:46 +01:00
Simon Ser
88493d1f7c linux-dmabuf-v1: don't use wlr_renderer to send legacy format list
The wlr_renderer field will go away in a subsequent commit.

Build the legacy device list from the default feedback instead.
2022-11-30 14:19:46 +01:00
Simon Ser
d45cffb865 linux-dmabuf-v1: don't use wlr_renderer to sanity check DMA-BUFs
The wlr_renderer field will go away in a subsequent commit.

Instead of trying to create a texture, try to import the DMA-BUF
into the DRM device FD.
2022-11-30 14:14:07 +01:00
Simon Ser
c99c60090c linux-dmabuf-v1: add version arg to create()
To be able to add support for newer versions without breaking
changes.
2022-11-30 14:14:04 +01:00
Simon Ser
444d94ef89 linux-dmabuf-v1: add "_with_renderer" suffix to create() function
Make it clear this is a helper consuming a wlr_renderer. We'll
add a lower-level create() function which doesn't take it in the
next commit.
2022-11-30 14:14:02 +01:00
Simon Zeni
0611c10909 types/wlr_seat: finish keyboard_state during wlr_seat_destroy 2022-11-29 15:21:18 -05:00
Johan Malm
338e0a3976 scene/layer_shell_v1.c: remove redundant comment 2022-11-29 18:01:57 +00:00
Johan Malm
8e80432a72 scene/layer_shell_v1.c: set exclusive zone correctly
...when only one edge is anchored.

The layer-shell protocol specifies that a positive exclusive-zone value is
'meaningful' if the surface is anchored to either:

1. one edge
2. one edge and both perpendicular edges.

For example, if you wish to position a layer-shell client along the top
edge and make it exclusive, you should be able to either set anchor=TOP or
anchor=TOP|LEFT|RIGHT. It appears that many panels/bars use the latter
approach (anchor to an edge and also both perpendicular edges) which is
probably why this has not been reported previously.

This patch adds support for the first case and thereby makes exclusive
zone behaviour consistent with the protocol and also with sway's extant
layer-shell implementation.
2022-11-29 18:01:57 +00:00