From 20587b349504187dbb646ad6bc329a69c5d5c969 Mon Sep 17 00:00:00 2001
From: Timo <16718859+toger5@users.noreply.github.com>
Date: Mon, 6 Jul 2026 23:57:26 +0800
Subject: [PATCH] Start voice call in PiP (not in fullscreen) (#34055)

* start voice call in pip

* Update jest.config.ts

* fix "join" button to use correct intent

* Add regression test for join button

* Update RoomHeader-test.tsx

* fmt

* Add playwright test

* Fix expected intent on "join" button press

* Add additional pip check to exising:
`should be able to join a ${callType} call in progress` test

* Update RoomHeader-test.tsx
---
 .../playwright/e2e/voip/element-call.spec.ts  | 33 ++++++++++-
 .../views/rooms/RoomHeader/RoomHeader.tsx     |  2 +-
 apps/web/src/stores/RoomViewStore.tsx         |  9 ++-
 .../rooms/RoomHeader/RoomHeader-test.tsx      | 27 ++++++++-
 .../unit-tests/stores/RoomViewStore-test.ts   | 57 ++++++++++++++++++-
 5 files changed, 123 insertions(+), 5 deletions(-)

diff --git a/apps/web/playwright/e2e/voip/element-call.spec.ts b/apps/web/playwright/e2e/voip/element-call.spec.ts
index 9675779c1f..c90ae306f3 100644
--- a/apps/web/playwright/e2e/voip/element-call.spec.ts
+++ b/apps/web/playwright/e2e/voip/element-call.spec.ts
@@ -217,8 +217,23 @@ test.describe("Element Call", () => {
                 const hash = new URLSearchParams(url.hash.slice(1));
                 assertCommonCallParameters(url.searchParams, hash, user, room);
 
-                expect(hash.get("intent")).toEqual("join_existing");
+                const expectedIntent = callType === "voice" ? "join_existing_voice" : "join_existing";
+                expect(hash.get("intent")).toEqual(expectedIntent);
                 expect(hash.get("skipLobby")).toEqual(null);
+
+                // pip layout check
+                switch (callType) {
+                    case "voice": {
+                        const pipContainer = page.getByTestId("widget-pip-container");
+                        await expect(pipContainer).toBeVisible();
+                        break;
+                    }
+                    case "video": {
+                        const pipContainer = page.getByTestId("widget-pip-container");
+                        await expect(pipContainer).not.toBeVisible();
+                        break;
+                    }
+                }
             });
         });
 
@@ -336,6 +351,22 @@ test.describe("Element Call", () => {
             expect(hash.get("skipLobby")).toEqual("true");
         });
 
+        test("should start a voice call in PiP", async ({ page, user, room, app }) => {
+            await app.viewRoomById(room.roomId);
+            await expect(page.getByText("Bob joined the room")).toBeVisible();
+
+            await page.getByRole("button", { name: "Voice call" }).click();
+            await page.getByRole("menuitem", { name: "Element Call" }).click();
+
+            const frameUrlStr = await page.locator("iframe").getAttribute("src");
+            await expect(frameUrlStr).toBeDefined();
+
+            // The call should be presented in the picture-in-picture container, right in the room we started it
+            // from, rather than taking over the room view.
+            const pipContainer = page.getByTestId("widget-pip-container");
+            await expect(pipContainer).toBeVisible();
+        });
+
         test("should be able to join a call in progress", async ({ page, user, bot, room, app }) => {
             await app.viewRoomById(room.roomId);
             await expect(page.getByText("Bob joined the room")).toBeVisible();
diff --git a/apps/web/src/components/views/rooms/RoomHeader/RoomHeader.tsx b/apps/web/src/components/views/rooms/RoomHeader/RoomHeader.tsx
index dbdb0f619a..4f7d0b9c1e 100644
--- a/apps/web/src/components/views/rooms/RoomHeader/RoomHeader.tsx
+++ b/apps/web/src/components/views/rooms/RoomHeader/RoomHeader.tsx
@@ -118,7 +118,7 @@ function RoomHeaderButtons({
         >
             <Button
                 size="md"
-                onClick={videoClick}
+                onClick={activeCallSessionType === CallType.Video ? videoClick : voiceClick}
                 // If we know this is a voice session, show the voice call. All other kinds of call are video calls.
                 Icon={activeCallSessionType === CallType.Voice ? VoiceCallIcon : VideoCallIcon}
                 className="mx_RoomHeader_join_button"
diff --git a/apps/web/src/stores/RoomViewStore.tsx b/apps/web/src/stores/RoomViewStore.tsx
index c4b76882dd..7305091661 100644
--- a/apps/web/src/stores/RoomViewStore.tsx
+++ b/apps/web/src/stores/RoomViewStore.tsx
@@ -52,6 +52,7 @@ import { setMarkedUnreadState } from "../utils/notifications";
 import { ConnectionState, ElementCall } from "../models/Call";
 import { isVideoRoom } from "../utils/video-rooms";
 import { ModuleApi } from "../modules/Api";
+import ActiveWidgetStore from "./ActiveWidgetStore";
 
 const NUM_JOIN_RETRY = 5;
 
@@ -377,6 +378,12 @@ export class RoomViewStore extends EventEmitter {
                     ElementCall.create(room);
                     call = CallStore.instance.getCall(payload.room_id)!;
                 }
+
+                // Custom case where we start voice calls in pip
+                if (payload.voiceOnly ?? false) {
+                    viewingCall = false;
+                    ActiveWidgetStore.instance.setWidgetPersistence(call.widget.id, room.roomId, true);
+                }
                 call.presented = true;
                 // Immediately start the call. This will connect to all required widget events
                 // and allow the widget to show the lobby.
@@ -401,7 +408,7 @@ export class RoomViewStore extends EventEmitter {
                     roomLoadError: null,
                     viaServers: payload.via_servers,
                     wasContextSwitch: payload.context_switch,
-                    viewingCall: payload.view_call ?? false,
+                    viewingCall,
                 });
                 // set this room as the room subscription. We need to await for it as this will fetch
                 // all room state for this room, which is required before we get the state below.
diff --git a/apps/web/test/unit-tests/components/views/rooms/RoomHeader/RoomHeader-test.tsx b/apps/web/test/unit-tests/components/views/rooms/RoomHeader/RoomHeader-test.tsx
index c997a65ae7..5f03ac5ae1 100644
--- a/apps/web/test/unit-tests/components/views/rooms/RoomHeader/RoomHeader-test.tsx
+++ b/apps/web/test/unit-tests/components/views/rooms/RoomHeader/RoomHeader-test.tsx
@@ -609,6 +609,30 @@ describe("RoomHeader", () => {
             expect(joinButton).not.toHaveAttribute("aria-disabled", "true");
         });
 
+        it("clicking the join button of an ongoing video call joins as a video call", async () => {
+            const user = userEvent.setup();
+            mockRoomMembers(room, 3);
+            jest.spyOn(CallStore.instance, "getCall").mockReturnValue(createMockCall(ROOM_ID, 3, CallType.Video, true));
+            render(<RoomHeader room={room} />, getWrapper());
+
+            const dispatcherSpy = jest.spyOn(dispatcher, "dispatch").mockImplementation();
+            await user.click(getByLabelText(document.body, "Join video call"));
+
+            expect(dispatcherSpy).toHaveBeenCalledWith(expect.objectContaining({ view_call: true, voiceOnly: false }));
+        });
+
+        it("clicking the join button of an ongoing voice call joins as a voice call", async () => {
+            const user = userEvent.setup();
+            mockRoomMembers(room, 3);
+            jest.spyOn(CallStore.instance, "getCall").mockReturnValue(createMockCall(ROOM_ID, 3, CallType.Voice, true));
+            render(<RoomHeader room={room} />, getWrapper());
+
+            const dispatcherSpy = jest.spyOn(dispatcher, "dispatch").mockImplementation();
+            await user.click(getByLabelText(document.body, "Join voice call"));
+
+            expect(dispatcherSpy).toHaveBeenCalledWith(expect.objectContaining({ view_call: true, voiceOnly: true }));
+        });
+
         it("join button is disabled if there is an other ongoing call", async () => {
             mockRoomMembers(room, 3);
             // Mock CallStore to return a call with 3 participants
@@ -919,6 +943,7 @@ function createMockCall(
     roomId: string = "!1:example.org",
     participantCount: number = 0,
     callType: CallType = CallType.Video,
+    isElementCall: boolean = false,
 ): Call {
     const participants = new Map();
 
@@ -936,7 +961,7 @@ function createMockCall(
     return {
         roomId,
         participants,
-        widget: { id: "test-widget" },
+        widget: { id: "test-widget", type: isElementCall ? "m.call" : undefined },
         connectionState: "disconnected",
         callType,
         on: jest.fn(),
diff --git a/apps/web/test/unit-tests/stores/RoomViewStore-test.ts b/apps/web/test/unit-tests/stores/RoomViewStore-test.ts
index 50c8393c95..06249972bd 100644
--- a/apps/web/test/unit-tests/stores/RoomViewStore-test.ts
+++ b/apps/web/test/unit-tests/stores/RoomViewStore-test.ts
@@ -48,7 +48,8 @@ import { CallStore } from "../../../src/stores/CallStore";
 import { MatrixClientPeg } from "../../../src/MatrixClientPeg";
 import MediaDeviceHandler, { MediaDeviceKindEnum } from "../../../src/MediaDeviceHandler";
 import { storeRoomAliasInCache } from "../../../src/RoomAliasCache.ts";
-import { type Call } from "../../../src/models/Call.ts";
+import { type Call, ConnectionState } from "../../../src/models/Call.ts";
+import ActiveWidgetStore from "../../../src/stores/ActiveWidgetStore";
 import { ModuleApi } from "../../../src/modules/Api";
 
 jest.mock("../../../src/Modal");
@@ -430,6 +431,60 @@ describe("RoomViewStore", function () {
         expect(call.presented).toEqual(true);
     });
 
+    it("opens a voice-intent call directly in picture-in-picture rather than maximised", async () => {
+        const call = {
+            presented: false,
+            connectionState: ConnectionState.Disconnected,
+            widget: { id: "!widget:example.org" },
+            start: jest.fn(),
+        } as unknown as Call;
+        jest.spyOn(CallStore.instance, "getCall").mockReturnValue(call);
+        const persistenceSpy = jest.spyOn(ActiveWidgetStore.instance, "setWidgetPersistence");
+        await setupAsyncStoreWithClient(CallStore.instance, MatrixClientPeg.safeGet());
+
+        dis.dispatch<ViewRoomPayload>({
+            action: Action.ViewRoom,
+            room_id: roomId,
+            view_call: true,
+            voiceOnly: true,
+            metricsTrigger: undefined,
+        });
+        await untilDispatch(Action.ViewRoom, dis);
+
+        // The call is started and marked persistent so it renders in the PiP container...
+        expect(call.presented).toEqual(true);
+        expect(persistenceSpy).toHaveBeenCalledWith("!widget:example.org", roomId, true);
+        expect(call.start).toHaveBeenCalledWith(expect.objectContaining({ voiceOnly: true }));
+        // ...but the room is not switched to the maximised call view.
+        expect(roomViewStore.isViewingCall()).toEqual(false);
+    });
+
+    it("opens a video-intent call maximised in the room", async () => {
+        const call = {
+            presented: false,
+            connectionState: ConnectionState.Disconnected,
+            widget: { id: "!widget:example.org" },
+            start: jest.fn(),
+        } as unknown as Call;
+        jest.spyOn(CallStore.instance, "getCall").mockReturnValue(call);
+        const persistenceSpy = jest.spyOn(ActiveWidgetStore.instance, "setWidgetPersistence");
+        await setupAsyncStoreWithClient(CallStore.instance, MatrixClientPeg.safeGet());
+
+        dis.dispatch<ViewRoomPayload>({
+            action: Action.ViewRoom,
+            room_id: roomId,
+            view_call: true,
+            voiceOnly: false,
+            metricsTrigger: undefined,
+        });
+        await untilDispatch(Action.ViewRoom, dis);
+
+        expect(call.presented).toEqual(true);
+        expect(persistenceSpy).not.toHaveBeenCalled();
+        expect(call.start).toHaveBeenCalledWith(expect.objectContaining({ voiceOnly: false }));
+        expect(roomViewStore.isViewingCall()).toEqual(true);
+    });
+
     it("should display an error message when the room is unreachable via the roomId", async () => {
         // When
         // View and wait for the room