C API: Add JNI bridge between Java proxy classes and WebKit/WPE GObjects by alexgcastro · Pull Request #250 · Igalia/wpe-android

alexgcastro · 2026-04-06T19:19:16Z

Add Java proxy wrappers plus JNI mappings for WPEDisplay, WPEToplevel, WPEView, WebKitWebContext, WebKitNetworkSession, WebKitSettings, WebKitCookieManager, WebKitWebsiteDataManager, and WebKitWebView.

Each bridged class uses a JNI::TypedClass singleton to register native methods and cache Java method IDs at library load time, avoiding repeated JNI lookups in GLib and async callbacks. JNIMappings initialises all singletons during JNI_OnLoad.

WebKitWebView and WebKitWebContext use per-object bridge structs that own the native GObject reference, keep a Java GlobalRef back-reference where needed, and track GLib signal handlers for orderly disconnection on destruction. Async operations such as cookie-manager queries, website-data clears, and evaluateJavascript() keep callback holders alive with heap-allocated GlobalRef objects released when the callback runs.

WPEToplevelAndroid updates its SurfaceControl tree as native windows are attached and detached, and treats repeated set_window() calls for the same ANativeWindow as a size resync instead of rebuilding the layer tree.

Native callbacks may arrive off the main thread; Java-side listener and async result delivery is reposted to the main looper before application-facing callback code runs.

Add Java proxy wrappers plus JNI mappings for WPEDisplay, WPEToplevel, WPEView, WebKitWebContext, WebKitNetworkSession, WebKitSettings, WebKitCookieManager, WebKitWebsiteDataManager, and WebKitWebView. Each bridged class uses a JNI::TypedClass<T> singleton to register native methods and cache Java method IDs at library load time, avoiding repeated JNI lookups in GLib and async callbacks. JNIMappings initialises all singletons during JNI_OnLoad. WebKitWebView and WebKitWebContext use per-object bridge structs that own the native GObject reference, keep a Java GlobalRef back-reference where needed, and track GLib signal handlers for orderly disconnection on destruction. Async operations such as cookie-manager queries, website-data clears, and evaluateJavascript() keep callback holders alive with heap-allocated GlobalRef<T> objects released when the callback runs. WPEToplevelAndroid updates its SurfaceControl tree as native windows are attached and detached, and treats repeated set_window() calls for the same ANativeWindow as a size resync instead of rebuilding the layer tree. Native callbacks may arrive off the main thread; Java-side listener and async result delivery is reposted to the main looper before application-facing callback code runs.

spenap · 2026-04-09T10:10:33Z

wpeview/src/main/cpp/capi/WebKitCookieManager.cpp

+        std::unique_ptr<JNI::GlobalRef<JNIWebKitCookieManagerAcceptPolicyCallbackHolder>> holder(
+            static_cast<JNI::GlobalRef<JNIWebKitCookieManagerAcceptPolicyCallbackHolder>*>(userData));
+        auto policy = webkit_cookie_manager_get_accept_policy_finish(WEBKIT_COOKIE_MANAGER(object), result, nullptr);
+        if (*holder) {


userData might be null -- it's passed from nativeGetAcceptPolicy into webkit_cookie_manager_get_accept_policy, and it takes holder, which could be nullptr if the received callbackHolder is null.
If that happens, the unique_ptr holder here will also be null after doing the cast, so de-referencing it here would trigger undefined behaviour.
We can't do a g_return_if_fail, because you still want to call webkit_cookie_manager_get_accept_policy_finish, so you probably need to turn this into

Suggested change

if (*holder) {

if (holder && *holder) {

spenap · 2026-04-09T10:17:19Z

wpeview/src/main/cpp/capi/WebKitCookieManager.cpp

+            ? new JNI::GlobalRef<JNIWebKitCookieManagerAcceptPolicyCallbackHolder>(env, callbackHolder)
+            : nullptr;
+        webkit_cookie_manager_get_accept_policy(
+            JNI::from_jlong<WebKitCookieManager>(nativePtr), nullptr, onGetAcceptPolicyReady, holder);


It's the same holder that here might take a null value, isn't it?

spenap · 2026-04-09T10:19:38Z

wpeview/src/main/cpp/capi/WebKitWebContext.cpp

+            GObject* automationSession = static_cast<GObject*>(g_weak_ref_get(&m_automationSession));
+            if (automationSession) {
+                g_signal_handler_disconnect(automationSession, m_createWebViewSignalId);
+                g_object_unref(automationSession);
+            }


Suggested change

GObject* automationSession = static_cast<GObject*>(g_weak_ref_get(&m_automationSession));

if (automationSession) {

g_signal_handler_disconnect(automationSession, m_createWebViewSignalId);

g_object_unref(automationSession);

}

if (GObject* automationSession = static_cast<GObject*>(g_weak_ref_get(&m_automationSession))) {

g_signal_handler_disconnect(automationSession, m_createWebViewSignalId);

g_object_unref(automationSession);

}

You could do this, so that automationSession gets a more restricted scope and a future refactor doesn't mix things up down below

spenap · 2026-04-09T10:21:26Z

wpeview/src/main/cpp/capi/WebKitWebContext.cpp

+    auto* info = webkit_application_info_new();
+    webkit_application_info_set_name(info, "WPEWebView");
+    webkit_application_info_set_version(info, WEBKIT_MAJOR_VERSION, WEBKIT_MINOR_VERSION, WEBKIT_MICRO_VERSION);
+    webkit_automation_session_set_application_info(session, info);
+    webkit_application_info_unref(info);


Suggested change

auto* info = webkit_application_info_new();

webkit_application_info_set_name(info, "WPEWebView");

webkit_application_info_set_version(info, WEBKIT_MAJOR_VERSION, WEBKIT_MINOR_VERSION, WEBKIT_MICRO_VERSION);

webkit_automation_session_set_application_info(session, info);

webkit_application_info_unref(info);

g_autoptr(WebKitApplicationInfo) info = webkit_application_info_new();

webkit_application_info_set_name(info, "WPEWebView");

webkit_application_info_set_version(info, WEBKIT_MAJOR_VERSION, WEBKIT_MINOR_VERSION, WEBKIT_MICRO_VERSION);

webkit_automation_session_set_application_info(session, info);

(and if you want it to be deleted earlier, you can nest it into a separate scope

spenap · 2026-04-09T10:24:44Z

wpeview/src/main/cpp/capi/WebKitWebContext.cpp

+    GObject* previousAutomationSession = static_cast<GObject*>(g_weak_ref_get(&bridge->m_automationSession));
+    if (previousAutomationSession && bridge->m_createWebViewSignalId)
+        g_signal_handler_disconnect(previousAutomationSession, bridge->m_createWebViewSignalId);
+    if (previousAutomationSession)
+        g_object_unref(previousAutomationSession);


Suggested change

GObject* previousAutomationSession = static_cast<GObject*>(g_weak_ref_get(&bridge->m_automationSession));

if (previousAutomationSession && bridge->m_createWebViewSignalId)

g_signal_handler_disconnect(previousAutomationSession, bridge->m_createWebViewSignalId);

if (previousAutomationSession)

g_object_unref(previousAutomationSession);

if (GObject* previousAutomationSession = static_cast<GObject*>(g_weak_ref_get(&bridge->m_automationSession)) {

if (bridge->m_createWebViewSignalId)

g_signal_handler_disconnect(previousAutomationSession, bridge->m_createWebViewSignalId);

g_object_unref(previousAutomationSession);

}

spenap · 2026-04-09T10:34:04Z

wpeview/src/main/cpp/capi/WebKitWebView.cpp

+        resultStr = JNI::String(strValue);
+        g_free(strValue);
+    }
+    g_object_unref(value);


I'd also suggest g_autoptr for the JSCValue, but that would mess up the GError and gchar* suggestions.

spenap · 2026-04-09T10:36:21Z

wpeview/src/main/cpp/capi/WebKitWebView.cpp

+    if (!value) {
+        if (error)
+            g_error_free(error);
+        if (*holder) {


ditto

Suggested change

if (*holder) {

if (holder && *holder) {

spenap · 2026-04-09T10:39:18Z

wpeview/src/main/cpp/capi/WebKitWebView.cpp

+        g_free(strValue);
+    }
+    g_object_unref(value);
+    if (*holder)


Suggested change

if (*holder)

if (holder && *holder)

spenap · 2026-04-09T10:40:34Z

wpeview/src/main/cpp/capi/WPEView.cpp

+            WPEEvent* event = wpe_event_touch_new(eventType, view, WPE_INPUT_SOURCE_TOUCHSCREEN,
+                static_cast<uint32_t>(time), WPEModifiers {}, static_cast<uint32_t>(pointerIds.getReadOnlyContent()[i]),
+                static_cast<double>(pointerXs.getReadOnlyContent()[i]),
+                static_cast<double>(pointerYs.getReadOnlyContent()[i]));
+            wpe_view_event(view, event);
+            wpe_event_unref(event);


Suggested change

WPEEvent* event = wpe_event_touch_new(eventType, view, WPE_INPUT_SOURCE_TOUCHSCREEN,

static_cast<uint32_t>(time), WPEModifiers {}, static_cast<uint32_t>(pointerIds.getReadOnlyContent()[i]),

static_cast<double>(pointerXs.getReadOnlyContent()[i]),

static_cast<double>(pointerYs.getReadOnlyContent()[i]));

wpe_view_event(view, event);

wpe_event_unref(event);

g_autoptr(WPEEvent) event = wpe_event_touch_new(eventType, view, WPE_INPUT_SOURCE_TOUCHSCREEN,

static_cast<uint32_t>(time), WPEModifiers {}, static_cast<uint32_t>(pointerIds.getReadOnlyContent()[i]),

static_cast<double>(pointerXs.getReadOnlyContent()[i]),

static_cast<double>(pointerYs.getReadOnlyContent()[i]));

wpe_view_event(view, event);

spenap · 2026-04-09T10:40:55Z

wpeview/src/main/cpp/capi/WPEView.cpp

+        WPEEvent* event = wpe_event_keyboard_new(static_cast<WPEEventType>(type), view, WPE_INPUT_SOURCE_KEYBOARD,
+            static_cast<uint32_t>(time), modifiers, hardwareKeyCode, keySym);
+        wpe_view_event(view, event);
+        wpe_event_unref(event);


Suggested change

WPEEvent* event = wpe_event_keyboard_new(static_cast<WPEEventType>(type), view, WPE_INPUT_SOURCE_KEYBOARD,

static_cast<uint32_t>(time), modifiers, hardwareKeyCode, keySym);

wpe_view_event(view, event);

wpe_event_unref(event);

g_autoptr(WPEEvent) event = wpe_event_keyboard_new(static_cast<WPEEventType>(type), view, WPE_INPUT_SOURCE_KEYBOARD,

static_cast<uint32_t>(time), modifiers, hardwareKeyCode, keySym);

wpe_view_event(view, event);

alexgcastro requested review from aperezdc, carlosgcampos, juanjosanchez and spenap April 6, 2026 19:19

alexgcastro self-assigned this Apr 6, 2026

alexgcastro force-pushed the alex/capi-add-jni-bridge branch from 293daca to 5fd90f4 Compare April 6, 2026 19:25

spenap reviewed Apr 9, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

C API: Add JNI bridge between Java proxy classes and WebKit/WPE GObjects#250

C API: Add JNI bridge between Java proxy classes and WebKit/WPE GObjects#250
alexgcastro wants to merge 1 commit intomainfrom
alex/capi-add-jni-bridge

alexgcastro commented Apr 6, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

spenap Apr 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

alexgcastro commented Apr 6, 2026

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!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants