refactor: type escaped OS strings and fix the final audit findings

Escaped config strings (ghostty_os::osbytes) kept reaching OS
boundaries undecoded, and decoding was occasionally applied to plain
text. The root cause was that both were plain `String`s. Now:
- `EncodedOs` is a newtype (no Display/AsRef<Path>) that decodes only
  via to_os/to_path/to_bytes/to_wide; plain text enters via
  from_plain_text. OS helpers (env, home, xdg, passwd, resources dir,
  path expansion), config::Path, WorkingDirectory, theme/file_load/edit
  paths, terminal pwd, OpenUrl/Pwd actions and termio cwd use it, so the
  compiler forces decoding at every boundary. os::open takes &OsStr.
  Config output is unchanged.

Fixes all rows of docs/rewrite/notes/review-rv5-codex.md (final-Z7.md):
screen-file copy; regex/OSC 8 URLs no longer decoded (regression);
lossless Windows WTF-8/WTF-16 for args, env and CreateProcessW; Windows
resources dir/TERMINFO; color OSC request lists and X11 color lookup no
longer allocate globally (C zero-allocation test extended); macOS Mach
handler resumes all threads before KERN_FAILURE (hang regression) and
SIGABRT minidumps go through the exception thread; plus open-config,
background images, crash dir, ssh cache, exe dir/PATH, fontconfig
filenames, file links, kitty TMPDIR, .hushlogin and benchmark paths.
The typing also surfaced and fixed gtk-custom-css, bell audio, bash
HISTFILE, OSC 52 writes and the C config getters.

Test Plan:
- cargo test --workspace (3376 passed)
- just check-macos (both), just check-windows, just check-wasm
- diff <(ghostty-rs +show-config --default --docs) \
  <(ghostty +show-config --default --docs)
This commit is contained in:
ddidderr committed 2026-10-12 10:06:07 +02:00
1 parent 7a3f0b0f2c
commit f7fec6f44d
92 files changed
+2967 -992

No files matched your search

+2 -2
View File
@@ -131,7 +131,7 @@ pub fn action_to_c(action: &Action, keep: &mut Keep) -> ghostty_action_s {
Action::SetTabTitle(t) => v.set_tab_title.title = keep.cstr(&t.title),
Action::PromptTitle(p) => v.prompt_title = *p as c_int,
Action::Pwd(p) => {
v.pwd.pwd = keep.cstr_bytes(&ghostty_os::osbytes::config_str_to_bytes(&p.pwd));
v.pwd.pwd = keep.cstr_bytes(&p.pwd.to_bytes());
}
Action::MouseShape(s) => v.mouse_shape = *s as c_int,
Action::MouseVisibility(m) => v.mouse_visibility = *m as c_int,
@@ -185,7 +185,7 @@ pub fn action_to_c(action: &Action, keep: &mut Keep) -> ghostty_action_s {
v.open_url.kind = u.kind as c_int;
// The URL may be the escaped form of a non-UTF-8 path; hand the
// embedder the real bytes.
let url = ghostty_os::osbytes::config_str_to_bytes(&u.url);
let url = u.url.to_bytes();
v.open_url.url = keep.cstr_bytes(&url);
v.open_url.len = url.len();
}
+24 -15
View File
@@ -7,27 +7,27 @@ use ghostty_bench::{args::Args, benches};
/// Split the argument string like `std.process.ArgIteratorGeneral(.{})`:
/// whitespace separated, with double quotes grouping and backslash
/// escapes inside quotes.
fn split_args(s: &str) -> Vec<String> {
/// escapes inside quotes. Works on raw bytes: file names need not be UTF-8.
fn split_args(s: &[u8]) -> Vec<std::ffi::OsString> {
let mut out = Vec::new();
let mut cur = String::new();
let mut cur: Vec<u8> = Vec::new();
let mut have = false;
let mut in_quote = false;
let mut chars = s.chars().peekable();
while let Some(c) = chars.next() {
let mut bytes = s.iter().copied();
while let Some(c) = bytes.next() {
match c {
'"' => {
b'"' => {
in_quote = !in_quote;
have = true;
}
'\\' if in_quote => {
if let Some(n) = chars.next() {
b'\\' if in_quote => {
if let Some(n) = bytes.next() {
cur.push(n);
}
}
c if c.is_whitespace() && !in_quote => {
c if c.is_ascii_whitespace() && !in_quote => {
if have {
out.push(std::mem::take(&mut cur));
out.push(ghostty_os::osbytes::os_from_bytes(std::mem::take(&mut cur)));
have = false;
}
}
@@ -38,7 +38,7 @@ fn split_args(s: &str) -> Vec<String> {
}
}
if have {
out.push(cur);
out.push(ghostty_os::osbytes::os_from_bytes(cur));
}
out
}
@@ -62,7 +62,7 @@ pub unsafe extern "C" fn ghostty_benchmark_cli(
log::warn!("unknown action={action}");
return false;
}
let argv = split_args(&args.to_string_lossy());
let argv = split_args(args.to_bytes());
let parsed = match Args::parse(argv) {
Ok(a) => a,
Err(err) => {
@@ -85,9 +85,18 @@ mod tests {
#[test]
fn split_args_basic() {
assert_eq!(split_args("a b c"), ["a", "b", "c"]);
assert_eq!(split_args("--x=\"a b\" c"), ["--x=a b", "c"]);
assert_eq!(split_args(" ").len(), 0);
assert_eq!(split_args(b"a b c"), ["a", "b", "c"]);
assert_eq!(split_args(b"--x=\"a b\" c"), ["--x=a b", "c"]);
assert_eq!(split_args(b" ").len(), 0);
}
#[cfg(unix)]
#[test]
fn split_args_keeps_non_utf8_bytes() {
use std::os::unix::ffi::OsStrExt;
let v = split_args(b"--data=/tmp/a\xffb x");
assert_eq!(v[0].as_bytes(), b"--data=/tmp/a\xffb");
assert_eq!(v[1], "x");
}
#[test]
+40 -10
View File
@@ -16,7 +16,7 @@ use ghostty_config::{
use ghostty_input::binding::{Action as BindingAction, TriggerC, TriggerCTag};
use crate::{
main_c::{EMPTY_STRING, string_from_owned},
main_c::{EMPTY_STRING, string_from_bytes},
types::{
ghostty_command_s,
ghostty_config_color_list_s,
@@ -103,6 +103,15 @@ fn cstring(s: &str) -> CString {
CString::new(s.replace('\0', "")).unwrap_or_default()
}
/// A C string for a config *value*: config strings carry non-UTF-8 bytes in
/// the reversible [`ghostty_os::osbytes`] encoding, so the embedder gets the
/// decoded bytes (Zig hands out the original byte slice).
fn cstring_value(s: &str) -> CString {
let mut bytes = ghostty_os::osbytes::config_str_to_bytes(s);
bytes.retain(|b| *b != 0);
CString::new(bytes).unwrap_or_default()
}
/// Intern a string for the lifetime of the process (enum tag names).
fn intern(s: &str) -> *const c_char {
static POOL: OnceLock<Mutex<HashSet<Box<CStr>>>> = OnceLock::new();
@@ -122,8 +131,8 @@ fn intern(s: &str) -> *const c_char {
fn build_backing(value: &CValue) -> Backing {
let mut b = Backing::default();
match value {
CValue::String(s) => b.strings.push(cstring(s)),
CValue::Path { path, .. } => b.strings.push(cstring(path)),
CValue::String(s) => b.strings.push(cstring_value(s)),
CValue::Path { path, .. } => b.strings.push(cstring_value(path)),
CValue::ColorList(list) => {
b.colors = list
.iter()
@@ -136,10 +145,10 @@ fn build_backing(value: &CValue) -> Backing {
}
CValue::Commands(cmds) => {
for c in cmds {
b.strings.push(cstring(&c.action_key));
b.strings.push(cstring(&c.action));
b.strings.push(cstring(&c.title));
b.strings.push(cstring(&c.description));
b.strings.push(cstring_value(&c.action_key));
b.strings.push(cstring_value(&c.action));
b.strings.push(cstring_value(&c.title));
b.strings.push(cstring_value(&c.description));
}
// The CStrings' heap buffers don't move when the Vec grows.
b.commands = b
@@ -252,9 +261,12 @@ pub unsafe extern "C" fn ghostty_config_load_file(ptr: ghostty_config_t, path: *
// Non-UTF-8 file names are valid on Unix; keep the bytes (reversible
// escape understood by the config file loader).
let path =
ghostty_os::osbytes::bytes_to_config_string(unsafe { CStr::from_ptr(path) }.to_bytes());
ghostty_os::osbytes::EncodedOs::from_bytes(unsafe { CStr::from_ptr(path) }.to_bytes());
if let Err(err) = c.config.load_file(&path) {
log::error!("error loading config from file path={path} err={err}");
log::error!(
"error loading config from file path={} err={err}",
path.as_encoded()
);
}
}
@@ -574,7 +586,8 @@ pub unsafe extern "C" fn ghostty_config_get_diagnostic(
#[unsafe(no_mangle)]
pub extern "C" fn ghostty_config_open_path() -> ghostty_string_s {
match ghostty_config::edit::open_path() {
Ok(path) => string_from_owned(&path),
// The embedder gets the real bytes of the path.
Ok(path) => string_from_bytes(&path.to_bytes()),
Err(err) => {
log::error!("error opening config in editor err={err}");
EMPTY_STRING
@@ -588,6 +601,23 @@ mod tests {
use super::*;
#[test]
fn config_values_reach_c_as_decoded_bytes() {
use ghostty_os::osbytes::EncodedOs;
let enc = EncodedOs::from_bytes(b"/a\xffb");
assert_ne!(enc.as_encoded().as_bytes(), b"/a\xffb");
let b = build_backing(&CValue::Path {
path: enc.as_encoded().to_owned(),
optional: false,
});
assert_eq!(b.strings[0].as_bytes(), b"/a\xffb");
let b = build_backing(&CValue::String(enc.into_encoded()));
assert_eq!(b.strings[0].as_bytes(), b"/a\xffb");
// Plain text (even with the escape char) is unchanged.
let b = build_backing(&CValue::String("h\u{e9}llo".to_owned()));
assert_eq!(b.strings[0].to_str().unwrap(), "h\u{e9}llo");
}
fn get<T: Copy>(cfg: ghostty_config_t, key: &str, init: T) -> (bool, T) {
let mut out = init;
// SAFETY: `out` is a valid destination of the key's C type.
+5 -4
View File
@@ -619,7 +619,7 @@ impl RtSurface for SurfaceRt {
true
}
fn set_clipboard(&self, clipboard: Clipboard, contents: &[(String, String)], confirm: bool) {
fn set_clipboard(&self, clipboard: Clipboard, contents: &[(String, Vec<u8>)], confirm: bool) {
let Some(cb) = self.app.opts.write_clipboard_cb else {
return;
};
@@ -628,7 +628,8 @@ impl RtSurface for SurfaceRt {
.map(|(m, d)| {
(
CString::new(m.replace('\0', "")).unwrap_or_default(),
CString::new(d.replace('\0', "")).unwrap_or_default(),
CString::new(d.iter().copied().filter(|b| *b != 0).collect::<Vec<u8>>())
.unwrap_or_default(),
)
})
.collect();
@@ -889,7 +890,7 @@ unsafe fn apply_surface_options(config: &mut Config, opts: &ghostty_surface_conf
&& !wd.is_empty()
{
// Raw bytes are kept (reversible escape in the String based config).
let wd_str = osbytes::bytes_to_config_string(wd);
let wd_str = osbytes::EncodedOs::from_bytes(wd);
let wd_os = osbytes::os_from_bytes(wd.to_vec());
let wd_disp = String::from_utf8_lossy(wd);
match std::fs::metadata(&wd_os) {
@@ -2207,7 +2208,7 @@ mod tests {
.as_ref()
.and_then(|w| w.value())
.unwrap();
assert_eq!(osbytes::config_str_to_bytes(got_wd), dir);
assert_eq!(got_wd.to_bytes(), dir);
let Some(Command::Shell(c)) = &config.command else {
panic!("no command");
};
+7 -1
View File
@@ -21,9 +21,15 @@ static ARGS: OnceLock<Vec<String>> = OnceLock::new();
/// Build the `ghostty_string_s` for an owned string (Zig:
/// `String.fromSlice` of an allocated `[:0]const u8`). Free with
/// `ghostty_string_free`.
#[cfg(test)]
pub(crate) fn string_from_owned(s: &str) -> ghostty_string_s {
string_from_bytes(s.as_bytes())
}
/// [`string_from_owned`] for raw bytes (paths need not be UTF-8).
pub(crate) fn string_from_bytes(s: &[u8]) -> ghostty_string_s {
let mut bytes = Vec::with_capacity(s.len() + 1);
bytes.extend_from_slice(s.as_bytes());
bytes.extend_from_slice(s);
bytes.push(0);
let boxed = bytes.into_boxed_slice();
let len = boxed.len() - 1;