Unicode parsing crash on non-unicode data

[cplir-c]

Describe the bug
Dumping significant amounts of binary data into kitty results in a crash from failed parsing?

To Reproduce

1. run a second kitty instance (with another instance group, if you use -1) and switch to it
2. Open a directory with several PNG files.
3. run sort *.png.
4. The second Kitty crashes, and the first kitty instance will contain the log.

Some probably-representative output:

[10.275] [PARSE ERROR] Unhandled charset related escape code: 0x2a 0x95
[10.275] [PARSE ERROR] Unknown char after ESC: 0x66
[10.275] [PARSE ERROR] Unknown char after ESC: 0x35
[10.275] [PARSE ERROR] Unrecognized APC code: 0xf0
[10.275] [PARSE ERROR] Unknown char after ESC: 0x6f
...
[10.275] [PARSE ERROR] Unknown char after ESC: 0x3b
[10.275] [PARSE ERROR] Invalid character in CSI: 0x9b (0x9b), ignoring the sequence
[10.275] [PARSE ERROR] Unknown char after ESC: 0xd
[10.275] [PARSE ERROR] Unrecognized APC code: 0x7d
[10.276] [PARSE ERROR] Unknown char after ESC: 0x36
...
[10.277] [PARSE ERROR] Unknown char after ESC: 0x66
[10.277] [PARSE ERROR] Unknown char after ESC: 0xe9
[10.277] [glfw error 65544]: Wayland: fatal display error: Broken pipe

Environment details

kitty 0.41.1 created by Kovid Goyal
Linux void 6.12.26_1 #1 SMP PREEMPT_DYNAMIC Sun May  4 21:20:14 UTC 2025 x86_64
Void 6.12.26_1 (void) (/dev/tty)

Running under: Wayland (sway version 1.10) missing: blur icon window-state-suspended
OpenGL: '4.6 (Core Profile) Mesa 24.3.4' Detected version: 4.6
Frozen: False
Fonts:
  medium: DejaVuSansMono: /usr/share/fonts/TTF/DejaVuSansMono.ttf:0
          Features: ()
    bold: DejaVuSansMono-Bold: /usr/share/fonts/TTF/DejaVuSansMono-Bold.ttf:0
          Features: ()
  italic: DejaVuSansMono-Oblique: /usr/share/fonts/TTF/DejaVuSansMono-Oblique.ttf:0
          Features: ()
      bi: DejaVuSansMono-BoldOblique: /usr/share/fonts/TTF/DejaVuSansMono-BoldOblique.ttf:0
          Features: ()
Paths:
  kitty: /usr/bin/kitty
  base dir: /usr/lib/kitty
  extensions dir: /usr/lib/kitty/kitty
  system shell: /bin/fish
System color scheme: no_preference. Applied color theme type: none
Loaded config files:
  /home/user/.config/kitty/kitty.conf

Config options different from defaults:
input_delay          5
linux_display_server wayland
repaint_delay        15

Important environment variables seen by the kitty process:
	PATH                                /usr/local/sbin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/bin:/home/user/.local/bin:/home/user/.cargo/bin
	LANG                                en_US.UTF-8
	EDITOR                              nvim
	SHELL                               /bin/fish
	DISPLAY                             :0
	WAYLAND_DISPLAY                     wayland-1
	USER                                user
	XCURSOR_SIZE                        24
	XDG_SESSION_TYPE                    wayland
	LC_COLLATE                          C
	XDG_RUNTIME_DIR                     /run/user/user
	XDG_DATA_DIRS                       /home/user/.local/share/flatpak/exports/share:/var/lib/flatpak/exports/share:/usr/local/share:/usr/share

Additional context
Also reproduced with --config NONE.

[kovidgoyal]

Does not reproduce for me with those steps on sway. Failed parsing is to be expected when you
dump binary data into a terminal.

From your log I don't see a crash, just wayland breaking the connection.
If you have an actual crash post the back trace. If it's just wayland
flaking, then run kitty with --dump-bytes and post the resulting bytes
and I will try to reproduce. Also post the log with --debug-rendering in
the latter case.

Closing till I get a back trace or a reproducer.

[alex-huff]

I was able to reproduce this by running sort *.png in a folder with .png files. For me, kitty crashed because it tried to set the title to a sequence of bytes that was not valid utf-8. This is possible because set_os_window_title calls _strip_csi without first making sure the title is less than 2048 bytes. This can cause the title to end in the middle of a multi-byte utf-8 codepoint.

python -c 'import sys; sys.stdout.buffer.write(b"\x1b]2;%b\xff\a" % (b"_" * 2046))'

[1007928.582] {Display Queue} wl_display#1.error(xdg_toplevel#32, 4294967295, "xdg_toplevel title is not valid UTF-8")
xdg_toplevel#32: error -1: xdg_toplevel title is not valid UTF-8
[2.788] [glfw error 65544]: Wayland: fatal display error: Protocol error

[kovidgoyal]

set_os_window_title calls glfwsetwindowtitle which in turn calls
ut8_strndup() which is supposed ot copy only valid utf-8, presumably a
bug in that fucntion.