Input triggers spike by tarek-y-ismail · Pull Request #4410 · canonical/mir

tarek-y-ismail · 2025-11-13T15:45:34Z

What's new?

Initial implementation of the input triggers protocols. Allows privileged clients to reserve a key combination (modifiers + key) and be notified whenever that combination is pressed/release, regardless of which client has focus.

How to test

Direct communication with privileged client:

Run mir with --add-wayland-extensions=ext_input_trigger_action_manager_v1:ext_input_trigger_registration_manager_v1
Inside the running compositor, launch a terminal
Run ./build/bin/trigger_client
At the moment, it's hardcoded to request Ctrl + Shift + C and ALT + x for key syms, and Alt + z for scan codes. If you press any of them. "Hey from " will be printed
Letting go will result in "Bye from " being printed.

To test Alt + z, run mir with --keymap fr. This should use the azerty layout and make the z key print w, and the q key print z. Even in the azerty layout, Alt + z with the azerty z will not work, but the same combination with the qwerty z will. Note: Right alt is used for special letters, so use left alt :)

Gatekeeper:

Run mir with --add-wayland-extensions=ext_input_trigger_action_manager_v1:ext_input_trigger_registration_manager_v1
Inside the running compositor, launch a terminal
Run python3 examples/gatekeeper/gatekeeper.py --show-ui
Run python3 examples/gatekeeper/test_client.py
The gatekeeper should request to register two shortcuts. You're free to reject, modify them, or accept them as is.
Pressing the shortcuts with the test client not focused will print different messages when begin and end are received.

Checklist

Tests added and pass
- WLCS tests require the addition of keyboard access. Probably for another PR.
Final rebase: once the PR is approved

tarek-y-ismail · 2025-11-14T13:42:13Z

The alphabetic key specified in the client request has to be the last key to be pressed. Pressing C + Shift + Ctrl or other combinations where C isn't the last character will not work.

We could solve this in two ways: Explicitly require users to press the alpha key at the end, or manually keep track of the pressed keysyms, consuming only the event where all the keys the user specified are down, which seems to be what i3 does on my machine. For example:

ctrl is pressed, goes to the focused client
c is pressed, goes to the focused client. We record that c is down
shift is pressed. We check if all modifiers and alpha keys are down and if so, we consume this event and call begin.

Edit: I'm opting to implement the second solution

tarek-y-ismail · 2025-12-04T17:21:13Z

Big TODO: The token authority should be modified to create a new token type (issue_trigger_token?) to allow multiple triggers to be registered.

Right now, tokens are used to store the relationship between a trigger, the input filter, and the corresponding action (that's what I remember, the architecture is a bit hazy in my head right now). So having tokens that are invalidated after a short time would require replacing this with something else.

The first idea that pops up into my head is using the token to just establish the relationship between the three objects, but not holding onto it as I do right now.

examples/trigger_client/CMakeLists.txt

tarek-y-ismail · 2026-01-20T14:49:40Z

src/server/frontend_wayland/CMakeLists.txt

+  input_trigger_registration_v1.cpp          input_trigger_registration_v1.h
+  input_trigger_action_v1.cpp   input_trigger_action_v1.h


Ugly indentation. But probably something for another PR

src/server/frontend_wayland/input_trigger_action_v1.cpp

src/server/frontend_wayland/wayland_default_configuration.cpp

CMakeLists.txt

examples/gatekeeper/gatekeeper.py

examples/gatekeeper/test_client.py

Copilot

Pull request overview

This PR implements a spike/prototype for an input trigger protocol that allows Wayland clients to register global keyboard shortcuts. The protocol consists of two parts: trigger registration (defining shortcuts) and action management (handling shortcut events).

Changes:

Implements Wayland protocol extensions for input trigger registration and action management
Adds support infrastructure including token-based action tracking and event filtering
Includes example client and Python-based gatekeeper for testing the protocol

Reviewed changes

Copilot reviewed 20 out of 21 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
src/server/frontend_wayland/wayland_default_configuration.cpp	Registers the two new input trigger protocol extensions in the Wayland extension system
src/server/frontend_wayland/wayland_connector.h	Adds InputTriggerData and Clock dependencies to WaylandExtensions context
src/server/frontend_wayland/wayland_connector.cpp	Instantiates InputTriggerData and passes it to extension context
src/server/frontend_wayland/input_trigger_registration_v1.h	Defines protocol interfaces for trigger registration and KeyboardSymTrigger class
src/server/frontend_wayland/input_trigger_registration_v1.cpp	Implements registration manager, action controls, and modifier mapping logic
src/server/frontend_wayland/input_trigger_data.h	Defines shared data structure for tracking pending trigger actions
src/server/frontend_wayland/input_trigger_action_v1.h	Defines protocol interfaces for action management
src/server/frontend_wayland/input_trigger_action_v1.cpp	Implements action manager, keyboard event filtering, and modifier matching
src/server/frontend_wayland/CMakeLists.txt	Adds new source files to build system
src/common/input/xkb_mapper.cpp	Moves expand_modifiers function to namespace scope for external use
include/common/mir/input/xkb_mapper.h	Exports expand_modifiers function for use in trigger protocol
examples/trigger_client/main.cpp	Implements standalone C++ Wayland client to test trigger registration
examples/trigger_client/CMakeLists.txt	Build configuration for trigger_client example
examples/gatekeeper/test_client.py	Python test client using D-Bus portal interface
examples/gatekeeper/generate_protocols.py	Script to generate Python bindings for Wayland protocols
examples/gatekeeper/gatekeeper.py	Python-based UI for managing global shortcuts via D-Bus portal
examples/CMakeLists.txt	Includes trigger_client subdirectory in build
debian/mir-demos.install	Adds trigger_client to package install list
rpm/mir.spec	Adds trigger_client to RPM package
CMakeLists.txt	Adds coverage build configuration and fixes platform path in LGPL test

src/server/frontend_wayland/input_trigger_registration_v1.cpp

src/server/frontend_wayland/input_trigger_action_v1.cpp

src/server/frontend_wayland/wayland_default_configuration.cpp

examples/trigger_client/main.cpp

tarek-y-ismail

Quick review for the gatekeeper and test client

Gatekeeper is a bit spaghettified, test client is alright.

tarek-y-ismail · 2026-01-27T08:41:02Z

examples/gatekeeper/gatekeeper.py

+# Add current directory to path to find generated protocols
+sys.path.append(os.path.dirname(os.path.abspath(__file__)))


Is this necessary?

tarek-y-ismail · 2026-01-27T08:42:48Z

examples/gatekeeper/gatekeeper.py

+# Minimal Keysym definitions (XKB)
+KEY_A = 0x0061
+# ... add more as needed
+
+MOD_SHIFT = 0x08
+MOD_CTRL = 0x100
+MOD_ALT = 0x01
+MOD_META = 0x800


Hmmm, maybe there's a library to handle this?

tarek-y-ismail · 2026-01-27T08:44:25Z

examples/gatekeeper/gatekeeper.py

+        lbl = Gtk.Label(label="An application wants to register the following shortcuts:")
+        box.append(lbl)


Unnecessary variable.

Suggested change

lbl = Gtk.Label(label="An application wants to register the following shortcuts:")

box.append(lbl)

box.append(Gtk.Label(label="An application wants to register the following shortcuts:"))

tarek-y-ismail · 2026-01-27T08:51:49Z

examples/gatekeeper/gatekeeper.py

+class GlobalShortcutsPortal:
+    def __init__(self, app):
+        self.app = app
+        self.sessions = {} # session_handle -> { 'app_id': str, 'shortcuts': dict, 'sender': str }


If we need a comment with the structure, might as well use type hints.

tarek-y-ismail · 2026-01-27T08:52:58Z

examples/gatekeeper/gatekeeper.py

+    # [CHANGE] Added sender argument
+    def CreateSession(self, request_handle, session_handle, app_id, options, sender):
+        logger.info(f"CreateSession: {session_handle} for {app_id} (Sender: {sender})")
+        self.sessions[session_handle] = {
+            'app_id': app_id,
+            'shortcuts': {},
+            'sender': sender  # [CHANGE] Store the sender


Leftovers from when I was vibing this

Suggested change

# [CHANGE] Added sender argument

def CreateSession(self, request_handle, session_handle, app_id, options, sender):

logger.info(f"CreateSession: {session_handle} for {app_id} (Sender: {sender})")

self.sessions[session_handle] = {

'app_id': app_id,

'shortcuts': {},

'sender': sender # [CHANGE] Store the sender

def CreateSession(self, request_handle, session_handle, app_id, options, sender):

logger.info(f"CreateSession: {session_handle} for {app_id} (Sender: {sender})")

self.sessions[session_handle] = {

'app_id': app_id,

'shortcuts': {},

'sender': sender

tarek-y-ismail · 2026-01-27T09:38:24Z

examples/gatekeeper/test_client.py

+from gi.repository import Gio, GLib
+
+# XML for the interface we (the client) will expose to receive events
+# [FIXED] Removed session_handle argument to match Gatekeeper's call signature (sua{sv})


Suggested change

# [FIXED] Removed session_handle argument to match Gatekeeper's call signature (sua{sv})

tarek-y-ismail · 2026-01-27T09:41:27Z

examples/gatekeeper/test_client.py

+    def on_client_method_call(self, connection, sender, object_path, interface_name, method_name, parameters, invocation):
+        # Handle the callbacks from Gatekeeper
+        if method_name == "Activated":
+            # [FIXED] Unpack 3 arguments: shortcut_id, timestamp, options


Suggested change

# [FIXED] Unpack 3 arguments: shortcut_id, timestamp, options

tarek-y-ismail · 2026-01-27T09:41:35Z

examples/gatekeeper/test_client.py

+            invocation.return_value(None)
+
+        elif method_name == "Deactivated":
+            # [FIXED] Unpack 3 arguments


Suggested change

# [FIXED] Unpack 3 arguments

tarek-y-ismail · 2026-01-27T09:42:43Z

examples/gatekeeper/test_client.py

+    def setup_client_listener(self):
+        # We simply register the object. Gatekeeper knows our bus unique name (e.g. :1.99)
+        # because it captured it when we called CreateSession.
+        session_path = "/org/freedesktop/portal/desktop/session/test_client/1"


This probably means we can't have multiple test clients?

examples/gatekeeper/test_client.py

src/server/frontend_wayland/input_trigger_action_v1.h

tarek-y-ismail · 2026-01-27T13:35:32Z

src/server/frontend_wayland/input_trigger_data.h

+/// Tracks keyboard state shared among all keyboard event filters
+class KeyboardStateTracker
+{
+public:
+    KeyboardStateTracker() = default;
+
+    /// Update pressed keysyms on key down
+    void on_key_down(uint32_t keysym);
+
+    /// Update pressed keysyms on key up
+    void on_key_up(uint32_t keysym);
+
+    /// Check if a keysym exists in the pressed set
+    auto keysym_is_pressed(uint32_t keysym, bool case_insensitive) const -> bool;
+
+    /// Check if protocol modifiers match event modifiers
+    static auto protocol_and_event_modifiers_match(uint32_t protocol_modifiers, MirInputEventModifiers event_mods)
+        -> bool;
+
+private:
+    std::unordered_set<uint32_t> pressed_keysyms;


Not sure if the comments being in a different style is reason enough to change them...

src/server/frontend_wayland/input_trigger_data.h

src/server/frontend_wayland/input_trigger_registration_v1.cpp

src/server/frontend_wayland/input_trigger_registration_v1.h

src/server/frontend_wayland/wayland_connector.h

tarek-y-ismail · 2026-01-27T16:25:30Z

Ok, I think this is ready to review now. One thing that could be improved is that the data inside InputTriggerData could be encapsulated better instead of just being public, but that's a minor thing in my opinion.

AlanGriffiths

OK as a throwaway spike: it works without too much pain.

But for maintainable code that we want to land:

There are apparently spurious changes to xkb_mapper;
poor encapsulation of logic; and,
a failure to leverage the threading mode.

debian/mir-demos.install

examples/trigger_client/main.cpp

include/common/mir/input/xkb_mapper.h

src/common/input/xkb_mapper.cpp

src/server/frontend_wayland/wayland_connector.h

src/server/symbols.map

src/server/frontend_wayland/input_trigger_data.h

src/server/frontend_wayland/input_trigger_action_v1.cpp

…code`

This reverts commit b54f810.

tarek-y-ismail · 2026-03-03T15:55:25Z

I'll admit that I'm not 100% happy with how modifiers are being handled. But I ran out of time and ideas for today.

AlanGriffiths

Progress: I'm beginning to see how the code should have been structured. Sorry it has taken this long to disentangle

src/server/frontend_wayland/wayland_connector.h

AlanGriffiths · 2026-03-04T17:14:56Z

src/server/frontend_wayland/input_trigger_common.h

+    shell::TokenAuthority& token_authority;
+    std::vector<wayland::Weak<InputTriggerAction const>> actions;
+    std::optional<std::pair<uint32_t, std::string>> timestamp_and_token;
+};


ActionGroup isn't an informative name: there's nothing in the name to connect it to input triggers. Maybe if it were namespaced by putting it in the InputTrigger class?

Compare with the naming of InputTriggerAction, that is clearly related to input triggers.

But also, look at the interface it provides: add(), end() and begin(). Anyone looking at the Group name, and add(), begin() & end() methods would expect it to be a collection.

Maybe both InputTriggerAction and ActionGroup should be in InputTriggerRegistry?

And maybe, ActionGroup should be ActionControl as that is the part of the protocol it relates to.

class InputTriggerRegistry { public: class Action; // nee `InputTriggerAction` class ActionControl; // nee `ActionGroup` class ActionControlRegistration; // nee `ActionGroupManager` class Trigger; // nee `InputTrigger` ... }; class InputTriggerRegistry::Action ... class InputTriggerRegistry::ActionControl { public: ActionControl(shell::TokenAuthority& token_authority); void add(wayland::Weak<Action const> action); void send_begin(MirEvent const& event); void send_end(MirEvent const& event); ... }; class InputTriggerRegistry::ActionControlRegister ...

I contemplated naming it InputTriggerActionGroup but felt like we had too many long names already. In any case, I agree with all of your suggestion except overloading ActionControl further. We already have an instance of that name in the protocol, another instance in input_trigger_registration_v1.cpp. A third would be too confusing.

I originally chose this name because this simply was a way to wrap a group of actions, hence ActionGroup. Maybe under InputTriggerRegistry that name would work?

I chose ActionControl because the functionality corresponds with the protocol object

I disagree. The protocol's ActionControl object associates triggers with a specific action. This class groups a set of actions together.

src/server/frontend_wayland/input_trigger_common.h

tarek-y-ismail

Progress: I'm beginning to see how the code should have been structured. Sorry it has taken this long to disentangle

My bad, I'm the one responsible for tangling it in the first place xD

tarek-y-ismail · 2026-03-04T18:52:12Z

src/server/frontend_wayland/input_trigger_common.h

+    shell::TokenAuthority& token_authority;
+    std::vector<wayland::Weak<InputTriggerAction const>> actions;
+    std::optional<std::pair<uint32_t, std::string>> timestamp_and_token;
+};


I contemplated naming it InputTriggerActionGroup but felt like we had too many long names already. In any case, I agree with all of your suggestion except overloading ActionControl further. We already have an instance of that name in the protocol, another instance in input_trigger_registration_v1.cpp. A third would be too confusing.

I originally chose this name because this simply was a way to wrap a group of actions, hence ActionGroup. Maybe under InputTriggerRegistry that name would work?

src/server/frontend_wayland/input_trigger_common.h

github-actions · 2026-03-05T10:26:37Z

TICS Quality Gate

❌ Failed

mir

Coding Standards: ❌ 6 Blocking Issues

❌ Condition “No new Coding Standard Violations for level 1, 2, 3 with respect to Previous analysis” failed 2 times.

File	Issues
File	🪲 Total	❌ Blocking
src/server/frontend_wayland/wayland_connector.cpp	3	+3
src/server/frontend_wayland/wl_seat.cpp	3	+3

See the results in the TICS Viewer

The following files have been checked for this project

src/server/frontend_wayland/input_trigger_action_v1.cpp
src/server/frontend_wayland/input_trigger_action_v1.h
src/server/frontend_wayland/input_trigger_common.cpp
src/server/frontend_wayland/input_trigger_common.h
src/server/frontend_wayland/input_trigger_registration_v1.cpp
src/server/frontend_wayland/input_trigger_registration_v1.h
src/server/frontend_wayland/wayland_connector.cpp
src/server/frontend_wayland/wayland_connector.h
src/server/frontend_wayland/wayland_default_configuration.cpp
src/server/frontend_wayland/wl_seat.cpp
src/server/frontend_wayland/wl_seat.h

TICS / TICS / Run TICS analysis

`WaylandConnector::Context`

tarek-y-ismail requested review from AlanGriffiths, RAOF and mattkae November 13, 2025 15:45

tarek-y-ismail self-assigned this Nov 13, 2025

tarek-y-ismail force-pushed the input-triggers-spike branch 2 times, most recently from 76a9c52 to 29f6da5 Compare November 14, 2025 12:32

tarek-y-ismail force-pushed the input-triggers-spike branch from 4d9958a to 60cd1ab Compare November 17, 2025 15:41

tarek-y-ismail force-pushed the MIRENG-1493-input-triggers branch from 41c94a9 to 3599e06 Compare November 18, 2025 08:01

tarek-y-ismail force-pushed the input-triggers-spike branch from 60cd1ab to 99780c8 Compare November 18, 2025 08:02

tarek-y-ismail force-pushed the input-triggers-spike branch 2 times, most recently from f488d7b to 5e01560 Compare December 4, 2025 17:17

tarek-y-ismail force-pushed the input-triggers-spike branch from 5531782 to d880af4 Compare December 11, 2025 09:50

AlanGriffiths force-pushed the MIRENG-1493-input-triggers branch from 7c58613 to 90f9c87 Compare January 16, 2026 11:09

Base automatically changed from MIRENG-1493-input-triggers to main January 20, 2026 11:07

tarek-y-ismail force-pushed the input-triggers-spike branch 4 times, most recently from 1cf3f74 to f8c1c47 Compare January 20, 2026 14:41

tarek-y-ismail commented Jan 20, 2026

View reviewed changes

tarek-y-ismail requested a review from Copilot January 20, 2026 16:00

Copilot AI reviewed Jan 20, 2026

View reviewed changes

tarek-y-ismail force-pushed the input-triggers-spike branch from f8c1c47 to 5b72aeb Compare January 23, 2026 13:21

tarek-y-ismail commented Jan 27, 2026

View reviewed changes

tarek-y-ismail marked this pull request as ready for review January 27, 2026 16:25

tarek-y-ismail requested a review from a team as a code owner January 27, 2026 16:25

AlanGriffiths requested changes Jan 28, 2026

View reviewed changes

tarek-y-ismail added 14 commits March 3, 2026 15:55

InputTriggerActionV1 doesn't need to inherit from LiftimeTracker

fb249b2

Replace InputTriggerModifiers::raw_value with contains

c6331cf

Make InputTriggerModifiers::value const

a0bb508

Privatise KeyboardSymTrigger::keysym and `KeyboardCodeTrigger::scan…

9c3fc3c

…code`

Fix formatting of std::transform block

0ebe223

Remove unnecessary unordered_set include

b476afd

Revert "Delete input_modifier_untils in favor of xkb_mapper"

afcd16d

This reverts commit b54f810.

Fix indentdation in frontend_wayland/CMakeLists.txt

f0fd224

Fix includes in input_trigger_action_v1.cpp

c3e0c17

Format if-statement

60551dc

Use static_cast instead of std::string ctor

1b1d5ca

Remove ActionGroupManager::token_revoked

c68b01b

Refactor input trigger modifier handling

66d0426

Reorganize input_trigger_registration_v1.cpp

0f8a5b3

tarek-y-ismail requested a review from AlanGriffiths March 3, 2026 15:28

AlanGriffiths requested changes Mar 4, 2026

View reviewed changes

tarek-y-ismail commented Mar 4, 2026

View reviewed changes

tarek-y-ismail and others added 6 commits March 5, 2026 11:05

Remove ActionGroup::any_trigger_active

92341c8

Move input trigger types into InputTriggerRegistry namespace

ac93b5a

Refactor input_trigger_common.cpp to reorder class method definitions

724d099

Rename ActionGroup::begin/end to send_begin/send_end

e23a35b

Add class comment for RecentTokens describing its purpose

7e908c9

pre-commit: apply automatic fixes

9b19bcf

tarek-y-ismail mentioned this pull request Mar 5, 2026

Add keyboard state tracker #4746

Open

3 tasks

tarek-y-ismail and others added 4 commits March 6, 2026 13:52

Remove case-insensitive keysym matching from KeyboardStateTracker

e90d816

Move KeyboardStateTracker from InputTriggerRegistry to

634666e

`WaylandConnector::Context`

Move KeyboardStateTracker to a separate header and source file

6b07261

pre-commit: apply automatic fixes

aef0170

		input_trigger_registration_v1.cpp input_trigger_registration_v1.h
		input_trigger_action_v1.cpp input_trigger_action_v1.h

		# Add current directory to path to find generated protocols
		sys.path.append(os.path.dirname(os.path.abspath(__file__)))

		lbl = Gtk.Label(label="An application wants to register the following shortcuts:")
		box.append(lbl)

	lbl = Gtk.Label(label="An application wants to register the following shortcuts:")
	box.append(lbl)
	box.append(Gtk.Label(label="An application wants to register the following shortcuts:"))

Conversation

tarek-y-ismail commented Nov 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What's new?

How to test

Checklist

Uh oh!

tarek-y-ismail commented Nov 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tarek-y-ismail commented Dec 4, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tarek-y-ismail left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tarek-y-ismail commented Jan 27, 2026

Uh oh!

AlanGriffiths left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tarek-y-ismail commented Nov 13, 2025 •

edited

Loading

tarek-y-ismail commented Nov 14, 2025 •

edited

Loading

tarek-y-ismail Mar 4, 2026 •

edited

Loading

tarek-y-ismail Mar 4, 2026 •

edited

Loading