fix(jni): hand the native bridges standard UTF-8, not modified UTF-8

.github/workflows/build-natives.yaml

@@ -82,6 +82,10 @@ jobs:
         working-directory: src/native/linux
         run: bash run_pixmap_test.sh
 
+      - name: JNI strings are standard UTF-8 (issue #441)
+        working-directory: src/native/shared
+        run: bash run_jni_utf8_test.sh
+
       - name: Upload Linux library
         uses: actions/upload-artifact@v4
         with:

src/native/linux/build.sh

@@ -64,6 +64,7 @@ echo "Compiling jni_bridge.c..."
 gcc -c -o "$SCRIPT_DIR/jni_bridge.o" \
     -fPIC -O2 -Wall -Wextra -Wno-unused-parameter \
     -I "$SCRIPT_DIR" \
+    -I "$SCRIPT_DIR/../shared" \
     -I "$JNI_INCLUDE" \
     -I "$JNI_INCLUDE_LINUX" \
     "$SCRIPT_DIR/jni_bridge.c"

src/native/linux/jni_bridge.c

@@ -17,6 +17,7 @@
 #include <stdint.h>
 #include <dlfcn.h>
 
+#include "jni_utf8.h"
 #include "sni.h"
 
 /* ========================================================================== */
@@ -166,10 +167,7 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeCr
 {
     (void)clazz;
 
-    const char *tip = NULL;
-    if (tooltip) {
-        tip = (*env)->GetStringUTFChars(env, tooltip, NULL);
-    }
+    char *tip = jni_utf8_dup(env, tooltip);
 
     const uint8_t *icon_data = NULL;
     jsize icon_len = 0;
@@ -183,7 +181,7 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeCr
     sni_tray *tray = sni_tray_create(icon_data, (size_t)icon_len, tip);
 
     if (icon_buf) (*env)->ReleaseByteArrayElements(env, iconBytes, icon_buf, JNI_ABORT);
-    if (tip) (*env)->ReleaseStringUTFChars(env, tooltip, tip);
+    free(tip);
 
     return (jlong)(uintptr_t)tray;
 }
@@ -249,9 +247,9 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeSe
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return;
-    const char *utf = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
-    sni_tray_set_title(tray, utf);
-    if (utf) (*env)->ReleaseStringUTFChars(env, title, utf);
+    char *t = jni_utf8_dup(env, title);
+    sni_tray_set_title(tray, t);
+    free(t);
 }
 
 JNIEXPORT void JNICALL
@@ -261,9 +259,9 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeSe
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return;
-    const char *utf = tooltip ? (*env)->GetStringUTFChars(env, tooltip, NULL) : NULL;
-    sni_tray_set_tooltip(tray, utf);
-    if (utf) (*env)->ReleaseStringUTFChars(env, tooltip, utf);
+    char *tt = jni_utf8_dup(env, tooltip);
+    sni_tray_set_tooltip(tray, tt);
+    free(tt);
 }
 
 /* ── Callbacks ──────────────────────────────────────────────────────── */
@@ -358,11 +356,11 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeAd
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return 0;
-    const char *t = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
-    const char *tt = tooltip ? (*env)->GetStringUTFChars(env, tooltip, NULL) : NULL;
+    char *t = jni_utf8_dup(env, title);
+    char *tt = jni_utf8_dup(env, tooltip);
     uint32_t id = sni_tray_add_menu_item(tray, t, tt);
-    if (t) (*env)->ReleaseStringUTFChars(env, title, t);
-    if (tt) (*env)->ReleaseStringUTFChars(env, tooltip, tt);
+    free(t);
+    free(tt);
     return (jint)id;
 }
 
@@ -373,11 +371,11 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeAd
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return 0;
-    const char *t = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
-    const char *tt = tooltip ? (*env)->GetStringUTFChars(env, tooltip, NULL) : NULL;
+    char *t = jni_utf8_dup(env, title);
+    char *tt = jni_utf8_dup(env, tooltip);
     uint32_t id = sni_tray_add_menu_item_checkbox(tray, t, tt, checked ? 1 : 0);
-    if (t) (*env)->ReleaseStringUTFChars(env, title, t);
-    if (tt) (*env)->ReleaseStringUTFChars(env, tooltip, tt);
+    free(t);
+    free(tt);
     return (jint)id;
 }
 
@@ -398,11 +396,11 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeAd
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return 0;
-    const char *t = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
-    const char *tt = tooltip ? (*env)->GetStringUTFChars(env, tooltip, NULL) : NULL;
+    char *t = jni_utf8_dup(env, title);
+    char *tt = jni_utf8_dup(env, tooltip);
     uint32_t id = sni_tray_add_sub_menu_item(tray, (uint32_t)parentId, t, tt);
-    if (t) (*env)->ReleaseStringUTFChars(env, title, t);
-    if (tt) (*env)->ReleaseStringUTFChars(env, tooltip, tt);
+    free(t);
+    free(tt);
     return (jint)id;
 }
 
@@ -414,11 +412,11 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeAd
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return 0;
-    const char *t = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
-    const char *tt = tooltip ? (*env)->GetStringUTFChars(env, tooltip, NULL) : NULL;
+    char *t = jni_utf8_dup(env, title);
+    char *tt = jni_utf8_dup(env, tooltip);
     uint32_t id = sni_tray_add_sub_menu_item_checkbox(tray, (uint32_t)parentId, t, tt, checked ? 1 : 0);
-    if (t) (*env)->ReleaseStringUTFChars(env, title, t);
-    if (tt) (*env)->ReleaseStringUTFChars(env, tooltip, tt);
+    free(t);
+    free(tt);
     return (jint)id;
 }
 
@@ -440,9 +438,9 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeIt
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return 0;
-    const char *t = title ? (*env)->GetStringUTFChars(env, title, NULL) : NULL;
+    char *t = jni_utf8_dup(env, title);
     int ok = sni_tray_item_set_title(tray, (uint32_t)id, t);
-    if (t) (*env)->ReleaseStringUTFChars(env, title, t);
+    free(t);
     return (jint)ok;
 }
 
@@ -521,11 +519,11 @@ Java_dev_nucleusframework_composenativetray_lib_linux_LinuxNativeBridge_nativeIt
     (void)clazz;
     sni_tray *tray = (sni_tray *)(uintptr_t)handle;
     if (!tray) return;
-    const char *k = key ? (*env)->GetStringUTFChars(env, key, NULL) : NULL;
+    char *k = jni_utf8_dup(env, key);
     sni_tray_item_set_shortcut(tray, (uint32_t)id, k,
                                 ctrl ? 1 : 0, shift ? 1 : 0,
                                 alt ? 1 : 0, superMod ? 1 : 0);
-    if (k) (*env)->ReleaseStringUTFChars(env, key, k);
+    free(k);
 }
 
 /* ========================================================================== */

src/native/macos/MacTrayBridge.m

@@ -11,6 +11,7 @@
 #import <stdlib.h>
 #import <objc/runtime.h>
 #import <AppKit/AppKit.h>
+#include "jni_utf8.h"
 #include "tray.h"
 
 /* ========================================================================== */
@@ -214,13 +215,8 @@ JNIEXPORT jlong JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacN
     struct tray *t = (struct tray *)calloc(1, sizeof(struct tray));
     if (!t) return 0;
 
-    const char *iconUtf = (*env)->GetStringUTFChars(env, iconPath, NULL);
-    t->icon_filepath = strdup(iconUtf);
-    (*env)->ReleaseStringUTFChars(env, iconPath, iconUtf);
-
-    const char *tooltipUtf = (*env)->GetStringUTFChars(env, tooltip, NULL);
-    t->tooltip = strdup(tooltipUtf);
-    (*env)->ReleaseStringUTFChars(env, tooltip, tooltipUtf);
+    t->icon_filepath = jni_utf8_dup(env, iconPath);
+    t->tooltip = jni_utf8_dup(env, tooltip);
 
     t->menu = NULL;
     t->cb = NULL;
@@ -246,9 +242,7 @@ JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNa
     struct tray *t = (struct tray *)(uintptr_t)handle;
     if (!t) return;
     free((void *)t->icon_filepath);
-    const char *utf = (*env)->GetStringUTFChars(env, iconPath, NULL);
-    t->icon_filepath = strdup(utf);
-    (*env)->ReleaseStringUTFChars(env, iconPath, utf);
+    t->icon_filepath = jni_utf8_dup(env, iconPath);
 }
 
 JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNativeBridge_nativeSetTrayTooltip(
@@ -258,9 +252,7 @@ JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNa
     struct tray *t = (struct tray *)(uintptr_t)handle;
     if (!t) return;
     free((void *)t->tooltip);
-    const char *utf = (*env)->GetStringUTFChars(env, tooltip, NULL);
-    t->tooltip = strdup(utf);
-    (*env)->ReleaseStringUTFChars(env, tooltip, utf);
+    t->tooltip = jni_utf8_dup(env, tooltip);
 }
 
 JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNativeBridge_nativeSetTrayCallback(
@@ -378,17 +370,8 @@ JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNa
     free((void *)item->text);
     free((void *)item->icon_filepath);
 
-    const char *textUtf = (*env)->GetStringUTFChars(env, text, NULL);
-    item->text = strdup(textUtf);
-    (*env)->ReleaseStringUTFChars(env, text, textUtf);
-
-    if (iconPath != NULL) {
-        const char *iconUtf = (*env)->GetStringUTFChars(env, iconPath, NULL);
-        item->icon_filepath = strdup(iconUtf);
-        (*env)->ReleaseStringUTFChars(env, iconPath, iconUtf);
-    } else {
-        item->icon_filepath = NULL;
-    }
+    item->text = jni_utf8_dup(env, text);
+    item->icon_filepath = jni_utf8_dup(env, iconPath);
 
     item->disabled = (int)disabled;
     item->checked = (int)checked;
@@ -416,13 +399,7 @@ JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNa
 
     free((void *)item->key_equivalent);
 
-    if (keyEquivalent != NULL) {
-        const char *utf = (*env)->GetStringUTFChars(env, keyEquivalent, NULL);
-        item->key_equivalent = strdup(utf);
-        (*env)->ReleaseStringUTFChars(env, keyEquivalent, utf);
-    } else {
-        item->key_equivalent = NULL;
-    }
+    item->key_equivalent = jni_utf8_dup(env, keyEquivalent);
     item->key_equivalent_mod_mask = (unsigned long)modifierMask;
 }
 
@@ -541,11 +518,11 @@ JNIEXPORT void JNICALL Java_dev_nucleusframework_composenativetray_lib_mac_MacNa
     (void)clazz;
     struct tray *t = (struct tray *)(uintptr_t)handle;
     if (!t) return;
-    const char *lightUtf = (*env)->GetStringUTFChars(env, lightIcon, NULL);
-    const char *darkUtf = (*env)->GetStringUTFChars(env, darkIcon, NULL);
+    char *lightUtf = jni_utf8_dup(env, lightIcon);
+    char *darkUtf = jni_utf8_dup(env, darkIcon);
     tray_set_icons_for_appearance(t, lightUtf, darkUtf);
-    (*env)->ReleaseStringUTFChars(env, lightIcon, lightUtf);
-    (*env)->ReleaseStringUTFChars(env, darkIcon, darkUtf);
+    free(lightUtf);
+    free(darkUtf);
 }
 
 /* ========================================================================== */

src/native/macos/build.sh

@@ -44,6 +44,7 @@ build_arch() {
         -I "$JNI_INCLUDE" \
         -I "$JNI_INCLUDE_DARWIN" \
         -I "$SCRIPT_DIR" \
+        -I "$SCRIPT_DIR/../shared" \
         -fobjc-arc \
         "$SCRIPT_DIR/MacTrayBridge.m"
 

src/native/shared/jni_utf8.h

@@ -0,0 +1,87 @@
+/*
+ * jni_utf8.h – jstring to standard UTF-8 conversion, shared by all JNI bridges.
+ *
+ * GetStringUTFChars hands back JNI's "modified UTF-8", not standard UTF-8: a
+ * supplementary character (emoji, rare CJK, ...) comes out as its two UTF-16
+ * surrogates encoded separately, 3 bytes each, and U+0000 comes out as 0xC0 0x80.
+ * Native consumers that expect real UTF-8 then break: MultiByteToWideChar on
+ * Windows substitutes replacement characters, and sd-bus on Linux rejects the
+ * whole message with -EINVAL, which costs the entire menu, not just one label.
+ *
+ * Every bridge therefore reads UTF-16 with GetStringChars and encodes it here.
+ */
+
+#ifndef COMPOSENATIVETRAY_JNI_UTF8_H
+#define COMPOSENATIVETRAY_JNI_UTF8_H
+
+#include <jni.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <stdint.h>
+
+/* Encodes one code point into at most 4 bytes; returns the number written. */
+static inline size_t jni_utf8_encode(uint32_t cp, char *out) {
+    if (cp < 0x80) {
+        out[0] = (char)cp;
+        return 1;
+    }
+    if (cp < 0x800) {
+        out[0] = (char)(0xC0 | (cp >> 6));
+        out[1] = (char)(0x80 | (cp & 0x3F));
+        return 2;
+    }
+    if (cp < 0x10000) {
+        out[0] = (char)(0xE0 | (cp >> 12));
+        out[1] = (char)(0x80 | ((cp >> 6) & 0x3F));
+        out[2] = (char)(0x80 | (cp & 0x3F));
+        return 3;
+    }
+    out[0] = (char)(0xF0 | (cp >> 18));
+    out[1] = (char)(0x80 | ((cp >> 12) & 0x3F));
+    out[2] = (char)(0x80 | ((cp >> 6) & 0x3F));
+    out[3] = (char)(0x80 | (cp & 0x3F));
+    return 4;
+}
+
+/*
+ * Returns a freshly allocated, NUL-terminated standard UTF-8 copy of jstr, or
+ * NULL when jstr is NULL or the allocation fails. Free it with free().
+ *
+ * Unpaired surrogates cannot be represented in UTF-8 and become U+FFFD, so the
+ * result is always valid UTF-8 whatever the Java string contains. An embedded
+ * U+0000 truncates the copy, since the consumers are all C string APIs.
+ */
+static inline char *jni_utf8_dup(JNIEnv *env, jstring jstr) {
+    if (!jstr) return NULL;
+
+    const jchar *utf16 = (*env)->GetStringChars(env, jstr, NULL);
+    if (!utf16) return NULL;
+    jsize units = (*env)->GetStringLength(env, jstr);
+
+    /* Three bytes per UTF-16 unit is the worst case: a surrogate pair is two
+     * units and costs four bytes, every other unit costs at most three. */
+    char *out = (char *)malloc((size_t)units * 3 + 1);
+    if (!out) {
+        (*env)->ReleaseStringChars(env, jstr, utf16);
+        return NULL;
+    }
+
+    size_t written = 0;
+    for (jsize i = 0; i < units; i++) {
+        uint32_t cp = utf16[i];
+        if (cp >= 0xD800 && cp <= 0xDBFF && i + 1 < units &&
+            utf16[i + 1] >= 0xDC00 && utf16[i + 1] <= 0xDFFF) {
+            cp = 0x10000 + ((cp - 0xD800) << 10) + (utf16[++i] - 0xDC00);
+        } else if (cp >= 0xD800 && cp <= 0xDFFF) {
+            cp = 0xFFFD;
+        }
+        if (cp == 0) break;
+        written += jni_utf8_encode(cp, out + written);
+    }
+    out[written] = '\0';
+
+    (*env)->ReleaseStringChars(env, jstr, utf16);
+    return out;
+}
+
+#endif /* COMPOSENATIVETRAY_JNI_UTF8_H */

src/native/shared/run_jni_utf8_test.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+# Build and run the issue #441 modified-UTF-8 regression test.
+# Header-only: needs jni.h, nothing else.
+
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+BIN="$SCRIPT_DIR/test_jni_utf8"
+
+if [ -z "${JAVA_HOME:-}" ]; then
+    for jdk in /usr/lib/jvm/java-*-openjdk-amd64 /usr/lib/jvm/java-*-openjdk /usr/lib/jvm/default-java /usr/lib/jvm/default; do
+        if [ -f "$jdk/include/jni.h" ]; then
+            JAVA_HOME="$jdk"
+            break
+        fi
+    done
+fi
+if [ -z "${JAVA_HOME:-}" ] || [ ! -f "$JAVA_HOME/include/jni.h" ]; then
+    echo "ERROR: JAVA_HOME not found or jni.h missing. Install a JDK or set JAVA_HOME."
+    exit 1
+fi
+
+case "$(uname -s)" in
+    Darwin) JNI_MD_INCLUDE="$JAVA_HOME/include/darwin" ;;
+    *)      JNI_MD_INCLUDE="$JAVA_HOME/include/linux" ;;
+esac
+
+echo "Compiling modified-UTF-8 test..."
+cc -O2 -g -Wall -Wextra -Werror \
+    -I "$SCRIPT_DIR" \
+    -I "$JAVA_HOME/include" \
+    -I "$JNI_MD_INCLUDE" \
+    "$SCRIPT_DIR/test_jni_utf8.c" \
+    -o "$BIN"
+
+"$BIN"
+status=$?
+rm -f "$BIN"
+exit $status