From 7931654a7c29b926d91b3dd46a507e1d5c8747e0 Mon Sep 17 00:00:00 2001
From: Luflosi <luflosi@luflosi.de>
Date: Mon, 17 Feb 2020 15:14:35 +0100
Subject: [PATCH] Ensure valid UTF-8 when cutting string

---
 glfw/backend_utils.c | 18 +++++++++++++++
 glfw/backend_utils.h |  1 +
 glfw/wl_window.c     |  6 ++---
 kitty_tests/glfw.py  | 52 ++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 73 insertions(+), 4 deletions(-)
 create mode 100644 kitty_tests/glfw.py

diff --git a/glfw/backend_utils.c b/glfw/backend_utils.c
index 87af03f5d..2557111db 100644
--- a/glfw/backend_utils.c
+++ b/glfw/backend_utils.c
@@ -332,3 +332,21 @@ pollForEvents(EventLoopData *eld, monotonic_t timeout, watch_callback_func displ
     }
     return read_ok;
 }
+
+// Duplicate a UTF-8 encoded string
+// but cut it so that it has at most max_length bytes plus the null byte.
+// This does not take combining characters into account.
+GLFWAPI char* utf_8_strndup(const char* source, size_t max_length) {
+    if (!source) return NULL;
+    size_t length = strnlen(source, max_length);
+    if (length >= max_length) {
+        for (length = max_length; length > 0; length--) {
+            if ((source[length] & 0xC0) != 0x80) break;
+        }
+    }
+
+    char* result = malloc(length + 1);
+    memcpy(result, source, length);
+    result[length] = 0;
+    return result;
+}
diff --git a/glfw/backend_utils.h b/glfw/backend_utils.h
index 7ae4f90eb..cdbbd3eee 100644
--- a/glfw/backend_utils.h
+++ b/glfw/backend_utils.h
@@ -95,3 +95,4 @@ unsigned dispatchTimers(EventLoopData *eld);
 void finalizePollData(EventLoopData *eld);
 bool initPollData(EventLoopData *eld, int display_fd);
 void wakeupEventLoop(EventLoopData *eld);
+char* utf_8_strndup(const char* source, size_t max_length);
diff --git a/glfw/wl_window.c b/glfw/wl_window.c
index 166683c2c..64e5799a0 100644
--- a/glfw/wl_window.c
+++ b/glfw/wl_window.c
@@ -957,12 +957,10 @@ void _glfwPlatformSetWindowTitle(_GLFWwindow* window, const char* title)
 {
     if (window->wl.title)
         free(window->wl.title);
-    window->wl.title = _glfw_strdup(title);
     // Wayland cannot handle requests larger than ~8200 bytes. Sending
     // one causes an abort(). Since titles this large are meaningless anyway
-    // ensure they do not happen. One should really truncate ensuring valid UTF-8
-    // but I cant be bothered.
-    if (title && strnlen(title, 2048) >= 2048) window->wl.title[2048] = 0;
+    // ensure they do not happen.
+    window->wl.title = utf_8_strndup(title, 2048);
     if (window->wl.xdg.toplevel)
         xdg_toplevel_set_title(window->wl.xdg.toplevel, window->wl.title);
 }
diff --git a/kitty_tests/glfw.py b/kitty_tests/glfw.py
new file mode 100644
index 000000000..be80b7548
--- /dev/null
+++ b/kitty_tests/glfw.py
@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+# vim:fileencoding=utf-8
+# License: GPL v3 Copyright: 2020, Kovid Goyal <kovid at kovidgoyal.net>
+
+import sys
+import unittest
+from . import BaseTest
+
+_plat = sys.platform.lower()
+is_macos = 'darwin' in _plat
+
+
+class TestGLFW(BaseTest):
+
+    @unittest.skipIf(is_macos, 'Skipping test on macOS because glfw-cocoa.so is not built with backend_utils')
+    def test_utf_8_strndup(self):
+        import os
+        import ctypes
+
+        base = os.path.dirname(os.path.abspath(__file__))
+        backend_utils = os.path.join(base, '..', 'kitty', 'glfw-x11.so')
+        if not os.path.exists(backend_utils):
+            raise Exception('Module x11 not found')
+        lib = ctypes.CDLL(backend_utils)
+        utf_8_strndup = lib.utf_8_strndup
+        utf_8_strndup.restype = ctypes.c_char_p
+        utf_8_strndup.argtypes = (ctypes.c_char_p, ctypes.c_size_t)
+
+        def test(string):
+            string_bytes = bytes(string, 'utf-8')
+            prev_part_bytes = b''
+            prev_length_bytes = -1
+            for length in range(len(string) + 1):
+                part = string[:length]
+                part_bytes = bytes(part, 'utf-8')
+                length_bytes = len(part_bytes)
+                for length_bytes_2 in range(prev_length_bytes + 1, length_bytes):
+                    self.ae(utf_8_strndup(string_bytes, length_bytes_2), prev_part_bytes)
+                self.ae(utf_8_strndup(string_bytes, length_bytes), part_bytes)
+                prev_part_bytes = part_bytes
+                prev_length_bytes = length_bytes
+            self.ae(utf_8_strndup(string_bytes, len(string_bytes) + 1), string_bytes)  # Try to go one character after the end of the string
+
+        self.ae(utf_8_strndup(None, 2), None)
+        self.ae(utf_8_strndup(b'', 2), b'')
+
+        test('ö')
+        test('>a<')
+        test('>ä<')
+        test('>ế<')
+        test('>𐍈<')
+        test('∮ E⋅da = Q,  n → ∞, 𐍈∑ f(i) = ∏ g(i)')