mlua/tests/tests.rs

1325 lines
35 KiB
Rust
Raw Normal View History

use std::collections::HashMap;
use std::iter::FromIterator;
use std::panic::{catch_unwind, AssertUnwindSafe};
use std::string::String as StdString;
use std::sync::atomic::{AtomicU32, Ordering};
2020-05-05 20:32:05 -05:00
use std::sync::Arc;
use std::{error, f32, f64, fmt};
2017-05-21 18:50:59 -05:00
2019-10-14 16:21:30 -05:00
use mlua::{
ChunkMode, Error, ExternalError, Function, Lua, LuaOptions, Nil, Result, StdLib, String, Table,
UserData, Value, Variadic,
2019-10-14 16:21:30 -05:00
};
2017-05-21 18:50:59 -05:00
2022-02-19 08:15:15 -06:00
#[cfg(not(feature = "luau"))]
#[test]
fn test_safety() -> Result<()> {
let lua = Lua::new();
assert!(lua.load(r#"require "debug""#).exec().is_err());
match lua.load_from_std_lib(StdLib::DEBUG) {
Err(Error::SafetyError(_)) => {}
Err(e) => panic!("expected SafetyError, got {:?}", e),
Ok(_) => panic!("expected SafetyError, got no error"),
}
drop(lua);
let lua = unsafe { Lua::unsafe_new() };
assert!(lua.load(r#"require "debug""#).exec().is_ok());
drop(lua);
match Lua::new_with(StdLib::DEBUG, LuaOptions::default()) {
Err(Error::SafetyError(_)) => {}
Err(e) => panic!("expected SafetyError, got {:?}", e),
Ok(_) => panic!("expected SafetyError, got new Lua state"),
}
let lua = Lua::new();
match lua.load(r#"package.loadlib()"#).exec() {
Err(Error::CallbackError { ref cause, .. }) => match cause.as_ref() {
Error::SafetyError(_) => {}
e => panic!("expected SafetyError cause, got {:?}", e),
},
Err(e) => panic!("expected CallbackError, got {:?}", e),
Ok(_) => panic!("expected CallbackError, got no error"),
};
match lua.load(r#"require "fake_ffi""#).exec() {
Err(Error::RuntimeError(msg)) => assert!(msg.contains("can't load C modules in safe mode")),
Err(e) => panic!("expected RuntimeError, got {:?}", e),
Ok(_) => panic!("expected RuntimeError, got no error"),
}
drop(lua);
// Test safety rules after dynamically loading `package` library
let lua = Lua::new_with(StdLib::NONE, LuaOptions::default())?;
assert!(lua.globals().get::<_, Option<Value>>("require")?.is_none());
lua.load_from_std_lib(StdLib::PACKAGE)?;
match lua.load(r#"package.loadlib()"#).exec() {
Err(Error::CallbackError { ref cause, .. }) => match cause.as_ref() {
Error::SafetyError(_) => {}
e => panic!("expected SafetyError cause, got {:?}", e),
},
Err(e) => panic!("expected CallbackError, got {:?}", e),
Ok(_) => panic!("expected CallbackError, got no error"),
};
Ok(())
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_load() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2021-06-18 17:13:56 -05:00
2019-09-28 09:23:17 -05:00
let func = lua.load("return 1+2").into_function()?;
let result: i32 = func.call(())?;
assert_eq!(result, 3);
assert!(lua.load("").exec().is_ok());
2019-09-28 09:23:17 -05:00
assert!(lua.load("§$%§&$%&").exec().is_err());
2019-09-28 09:23:17 -05:00
Ok(())
}
2017-05-21 18:50:59 -05:00
#[test]
2019-09-28 09:23:17 -05:00
fn test_exec() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
Another major API change, out of stack space is not an Err It, ahem "should not" be possible to exhaust lua stack space in normal usage, and causing stack errors to be Err is slightly obnoxious. I have been wanting to make this change for a while, and removing the callback API from tables makes this sensible *I think*. I can think of a couple of ways that this is not technically true, but I think that they are acceptable, or should be handled differently. One, you can make arbitrarily sized LuaVariadic values. I think this is maybe a bug already, because there is an argument limit in Lua which is lower than the stack limit. I'm not sure what happens there, but if it is a stack based panic, (or any panic?) it is a bug. Two, I believe that if you recurse over and over between lua -> rust -> lua -> rust etc, and call rlua API functions, you might get a stack panic. I think for trusted lua code, this is morally equivalent to a regular stack overflow in plain rust, which is already.. well it's not a panic but it's some kind of safe crash I'm not sure, so I think this is acceptable. For *untrusted* lua code, this could theoretically be a problem if the API provided a callback that would call back into lua, then some lua script could force a stack based panic. There are so many concerns with untrusted lua code, and this library is NOT safe enough yet for untrusted code (it doesn't even provide an option to limit lua to the safe API subset yet!), so this is not currently an issue. When the library provides support for "safe lua", it should come with big warnings anyway, and being able to force a stack panic is pretty minor in comparison. I think if there are other ways to cause unbounded stack usage, that it is a bug, or there can be an error just for that situation, like argument count limits. This commit also fixes several stupid bugs with tests, stack checking, and panics.
2017-06-25 15:52:32 -05:00
let globals = lua.globals();
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
res = 'foo'..'bar'
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
assert_eq!(globals.get::<_, String>("res")?, "foobar");
2018-09-24 21:13:42 -05:00
let module: Table = lua
2019-09-28 09:23:17 -05:00
.load(
2018-08-05 08:51:39 -05:00
r#"
local module = {}
function module.func()
return "hello"
end
return module
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.eval()?;
2022-02-19 08:15:15 -06:00
println!("checkpoint");
2019-09-28 09:23:17 -05:00
assert!(module.contains_key("func")?);
2017-06-15 09:26:39 -05:00
assert_eq!(
2019-09-28 09:23:17 -05:00
module.get::<_, Function>("func")?.call::<_, String>(())?,
2017-06-15 09:26:39 -05:00
"hello"
);
2019-09-28 09:23:17 -05:00
Ok(())
2017-05-21 18:50:59 -05:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_eval() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
assert_eq!(lua.load("1 + 1").eval::<i32>()?, 2);
assert_eq!(lua.load("false == false").eval::<bool>()?, true);
assert_eq!(lua.load("return 1 + 2").eval::<i32>()?, 3);
match lua.load("if true then").eval::<()>() {
2017-10-23 15:42:20 -05:00
Err(Error::SyntaxError {
incomplete_input: true,
..
}) => {}
r => panic!(
"expected SyntaxError with incomplete_input=true, got {:?}",
r
),
2017-05-21 18:50:59 -05:00
}
2019-09-28 09:23:17 -05:00
Ok(())
2017-05-21 18:50:59 -05:00
}
#[test]
fn test_load_mode() -> Result<()> {
let lua = unsafe { Lua::unsafe_new() };
assert_eq!(
lua.load("1 + 1").set_mode(ChunkMode::Text).eval::<i32>()?,
2
);
match lua.load("1 + 1").set_mode(ChunkMode::Binary).exec() {
Ok(_) => panic!("expected SyntaxError, got no error"),
Err(Error::SyntaxError { message: msg, .. }) => {
assert!(msg.contains("attempt to load a text chunk"))
}
Err(e) => panic!("expected SyntaxError, got {:?}", e),
};
#[cfg(not(feature = "luau"))]
let bytecode = lua.load("return 1 + 1").into_function()?.dump(true);
#[cfg(feature = "luau")]
let bytecode = mlua::Compiler::new().compile("return 1 + 1");
assert_eq!(lua.load(&bytecode).eval::<i32>()?, 2);
assert_eq!(
lua.load(&bytecode)
.set_mode(ChunkMode::Binary)
.eval::<i32>()?,
2
);
match lua.load(&bytecode).set_mode(ChunkMode::Text).exec() {
Ok(_) => panic!("expected SyntaxError, got no error"),
Err(Error::SyntaxError { message: msg, .. }) => {
assert!(msg.contains("attempt to load a binary chunk"))
}
Err(e) => panic!("expected SyntaxError, got {:?}", e),
};
Ok(())
}
2017-05-21 18:50:59 -05:00
#[test]
2019-09-28 09:23:17 -05:00
fn test_lua_multi() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
lua.load(
r#"
function concat(arg1, arg2)
return arg1 .. arg2
end
2017-05-21 18:50:59 -05:00
function mreturn()
return 1, 2, 3, 4, 5, 6
end
2019-09-28 09:23:17 -05:00
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
let globals = lua.globals();
let concat = globals.get::<_, Function>("concat")?;
let mreturn = globals.get::<_, Function>("mreturn")?;
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
assert_eq!(concat.call::<_, String>(("foo", "bar"))?, "foobar");
let (a, b) = mreturn.call::<_, (u64, u64)>(())?;
2017-05-21 18:50:59 -05:00
assert_eq!((a, b), (1, 2));
2019-09-28 09:23:17 -05:00
let (a, b, v) = mreturn.call::<_, (u64, u64, Variadic<u64>)>(())?;
2017-05-21 18:50:59 -05:00
assert_eq!((a, b), (1, 2));
assert_eq!(v[..], [3, 4, 5, 6]);
2019-09-28 09:23:17 -05:00
Ok(())
2017-05-21 18:50:59 -05:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_coercion() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
int = 123
str = "123"
num = 123.0
func = function() end
2019-09-28 09:23:17 -05:00
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
let globals = lua.globals();
assert_eq!(globals.get::<_, String>("int")?, "123");
assert_eq!(globals.get::<_, i32>("str")?, 123);
assert_eq!(globals.get::<_, i32>("num")?, 123);
assert!(globals.get::<_, String>("func").is_err());
2019-09-28 09:23:17 -05:00
Ok(())
2017-05-21 18:50:59 -05:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_error() -> Result<()> {
2017-05-21 18:50:59 -05:00
#[derive(Debug)]
pub struct TestError;
impl fmt::Display for TestError {
fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result {
2017-05-21 18:50:59 -05:00
write!(fmt, "test error")
}
}
2022-02-19 08:15:15 -06:00
impl error::Error for TestError {}
2017-05-21 18:50:59 -05:00
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
Another major API change, out of stack space is not an Err It, ahem "should not" be possible to exhaust lua stack space in normal usage, and causing stack errors to be Err is slightly obnoxious. I have been wanting to make this change for a while, and removing the callback API from tables makes this sensible *I think*. I can think of a couple of ways that this is not technically true, but I think that they are acceptable, or should be handled differently. One, you can make arbitrarily sized LuaVariadic values. I think this is maybe a bug already, because there is an argument limit in Lua which is lower than the stack limit. I'm not sure what happens there, but if it is a stack based panic, (or any panic?) it is a bug. Two, I believe that if you recurse over and over between lua -> rust -> lua -> rust etc, and call rlua API functions, you might get a stack panic. I think for trusted lua code, this is morally equivalent to a regular stack overflow in plain rust, which is already.. well it's not a panic but it's some kind of safe crash I'm not sure, so I think this is acceptable. For *untrusted* lua code, this could theoretically be a problem if the API provided a callback that would call back into lua, then some lua script could force a stack based panic. There are so many concerns with untrusted lua code, and this library is NOT safe enough yet for untrusted code (it doesn't even provide an option to limit lua to the safe API subset yet!), so this is not currently an issue. When the library provides support for "safe lua", it should come with big warnings anyway, and being able to force a stack panic is pretty minor in comparison. I think if there are other ways to cause unbounded stack usage, that it is a bug, or there can be an error just for that situation, like argument count limits. This commit also fixes several stupid bugs with tests, stack checking, and panics.
2017-06-25 15:52:32 -05:00
let globals = lua.globals();
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
function no_error()
end
function lua_error()
error("this is a lua error")
end
function rust_error()
rust_error_function()
end
function return_error()
local status, res = pcall(rust_error_function)
assert(not status)
return res
end
function return_string_error()
return "this should be converted to an error"
end
function test_pcall()
local testvar = 0
pcall(function(arg)
testvar = testvar + arg
error("should be ignored")
end, 3)
local function handler(err)
2022-02-19 08:15:15 -06:00
if string.match(_VERSION, ' 5%.1$') or string.match(_VERSION, ' 5%.2$') or _VERSION == "Luau" then
-- Special case for Lua 5.1/5.2 and Luau
2019-10-14 16:21:30 -05:00
local caps = string.match(err, ': (%d+)$')
if caps then
err = caps
end
end
2019-09-28 09:23:17 -05:00
testvar = testvar + err
return "should be ignored"
2017-05-21 18:50:59 -05:00
end
2019-09-28 09:23:17 -05:00
local status, res = xpcall(function()
error(5)
end, handler)
assert(not status)
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
if testvar ~= 8 then
error("testvar had the wrong value, pcall / xpcall misbehaving "..testvar)
end
2019-09-28 09:23:17 -05:00
end
2019-09-28 09:23:17 -05:00
function understand_recursion()
understand_recursion()
end
"#,
)
.exec()?;
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
let rust_error_function =
lua.create_function(|_, ()| -> Result<()> { Err(TestError.to_lua_err()) })?;
globals.set("rust_error_function", rust_error_function)?;
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
let no_error = globals.get::<_, Function>("no_error")?;
let lua_error = globals.get::<_, Function>("lua_error")?;
let rust_error = globals.get::<_, Function>("rust_error")?;
let return_error = globals.get::<_, Function>("return_error")?;
let return_string_error = globals.get::<_, Function>("return_string_error")?;
let test_pcall = globals.get::<_, Function>("test_pcall")?;
let understand_recursion = globals.get::<_, Function>("understand_recursion")?;
2017-05-21 18:50:59 -05:00
assert!(no_error.call::<_, ()>(()).is_ok());
match lua_error.call::<_, ()>(()) {
Err(Error::RuntimeError(_)) => {}
2019-10-16 04:56:44 -05:00
Err(e) => panic!("error is not RuntimeError kind, got {:?}", e),
2017-06-17 22:50:40 -05:00
_ => panic!("error not returned"),
2017-05-21 18:50:59 -05:00
}
match rust_error.call::<_, ()>(()) {
Err(Error::CallbackError { .. }) => {}
2019-10-16 04:56:44 -05:00
Err(e) => panic!("error is not CallbackError kind, got {:?}", e),
2017-06-17 22:50:40 -05:00
_ => panic!("error not returned"),
2017-05-21 18:50:59 -05:00
}
match return_error.call::<_, Value>(()) {
Ok(Value::Error(_)) => {}
_ => panic!("Value::Error not returned"),
}
assert!(return_string_error.call::<_, Error>(()).is_ok());
2019-09-28 09:23:17 -05:00
match lua
.load("if youre happy and you know it syntax error")
.exec()
{
2017-10-23 15:42:20 -05:00
Err(Error::SyntaxError {
incomplete_input: false,
..
}) => {}
Err(_) => panic!("error is not LuaSyntaxError::Syntax kind"),
_ => panic!("error not returned"),
}
2019-09-28 09:23:17 -05:00
match lua.load("function i_will_finish_what_i()").exec() {
2017-10-23 15:42:20 -05:00
Err(Error::SyntaxError {
incomplete_input: true,
..
}) => {}
Err(_) => panic!("error is not LuaSyntaxError::IncompleteStatement kind"),
_ => panic!("error not returned"),
}
2017-05-21 18:50:59 -05:00
2019-09-28 09:23:17 -05:00
test_pcall.call::<_, ()>(())?;
2017-05-21 18:50:59 -05:00
assert!(understand_recursion.call::<_, ()>(()).is_err());
Ok(())
}
Another major API change, out of stack space is not an Err It, ahem "should not" be possible to exhaust lua stack space in normal usage, and causing stack errors to be Err is slightly obnoxious. I have been wanting to make this change for a while, and removing the callback API from tables makes this sensible *I think*. I can think of a couple of ways that this is not technically true, but I think that they are acceptable, or should be handled differently. One, you can make arbitrarily sized LuaVariadic values. I think this is maybe a bug already, because there is an argument limit in Lua which is lower than the stack limit. I'm not sure what happens there, but if it is a stack based panic, (or any panic?) it is a bug. Two, I believe that if you recurse over and over between lua -> rust -> lua -> rust etc, and call rlua API functions, you might get a stack panic. I think for trusted lua code, this is morally equivalent to a regular stack overflow in plain rust, which is already.. well it's not a panic but it's some kind of safe crash I'm not sure, so I think this is acceptable. For *untrusted* lua code, this could theoretically be a problem if the API provided a callback that would call back into lua, then some lua script could force a stack based panic. There are so many concerns with untrusted lua code, and this library is NOT safe enough yet for untrusted code (it doesn't even provide an option to limit lua to the safe API subset yet!), so this is not currently an issue. When the library provides support for "safe lua", it should come with big warnings anyway, and being able to force a stack panic is pretty minor in comparison. I think if there are other ways to cause unbounded stack usage, that it is a bug, or there can be an error just for that situation, like argument count limits. This commit also fixes several stupid bugs with tests, stack checking, and panics.
2017-06-25 15:52:32 -05:00
#[test]
fn test_panic() -> Result<()> {
fn make_lua(options: LuaOptions) -> Result<Lua> {
let lua = Lua::new_with(StdLib::ALL_SAFE, options)?;
2019-09-28 09:23:17 -05:00
let rust_panic_function =
lua.create_function(|_, msg: Option<StdString>| -> Result<()> {
if let Some(msg) = msg {
panic!("{}", msg)
}
panic!("rust panic")
})?;
lua.globals()
.set("rust_panic_function", rust_panic_function)?;
Ok(lua)
}
2017-05-21 18:50:59 -05:00
2021-06-18 17:13:56 -05:00
// Test triggering Lua error with sending Rust panic (must be resumed)
{
let lua = make_lua(LuaOptions::default())?;
match catch_unwind(AssertUnwindSafe(|| -> Result<()> {
lua.load(
r#"
_, err = pcall(rust_panic_function)
error(err)
"#,
)
.exec()
})) {
Ok(Ok(_)) => panic!("no panic was detected"),
Ok(Err(e)) => panic!("error during panic test {:?}", e),
Err(p) => assert!(*p.downcast::<&str>().unwrap() == "rust panic"),
};
// Trigger same panic again
match lua.load("error(err)").exec() {
Ok(_) => panic!("no error was detected"),
Err(Error::PreviouslyResumedPanic) => {}
Err(e) => panic!("expected PreviouslyResumedPanic, got {:?}", e),
}
}
2017-05-21 18:50:59 -05:00
// Test returning Rust panic (must be resumed)
{
let lua = make_lua(LuaOptions::default())?;
match catch_unwind(AssertUnwindSafe(|| -> Result<()> {
let _catched_panic = lua
.load(
r#"
-- Set global
_, err = pcall(rust_panic_function)
return err
"#,
)
.eval::<Value>()?;
Ok(())
})) {
Ok(_) => panic!("no panic was detected"),
Err(_) => {}
};
assert!(lua.globals().get::<_, Value>("err")? == Value::Nil);
match lua.load("tostring(err)").exec() {
Ok(_) => panic!("no error was detected"),
Err(Error::CallbackError { ref cause, .. }) => match cause.as_ref() {
Error::PreviouslyResumedPanic => {}
e => panic!("expected PreviouslyResumedPanic, got {:?}", e),
},
Err(e) => panic!("expected CallbackError, got {:?}", e),
}
}
2017-05-21 18:50:59 -05:00
2021-06-18 17:13:56 -05:00
// Test representing Rust panic as a string
match catch_unwind(|| -> Result<()> {
let lua = make_lua(LuaOptions::default())?;
2019-09-28 09:23:17 -05:00
lua.load(
r#"
local _, err = pcall(rust_panic_function)
error(tostring(err))
2019-09-28 09:23:17 -05:00
"#,
)
.exec()
2017-05-21 18:50:59 -05:00
}) {
2019-09-28 09:23:17 -05:00
Ok(Ok(_)) => panic!("no error was detected"),
Ok(Err(Error::RuntimeError(_))) => {}
Ok(Err(e)) => panic!("expected RuntimeError, got {:?}", e),
2019-09-28 09:23:17 -05:00
Err(_) => panic!("panic was detected"),
}
2019-09-28 09:23:17 -05:00
// Test disabling `catch_rust_panics` option / pcall correctness
match catch_unwind(|| -> Result<()> {
let lua = make_lua(LuaOptions::new().catch_rust_panics(false))?;
lua.load(
r#"
local ok, err = pcall(function(msg) error(msg) end, "hello")
assert(not ok and err:find("hello") ~= nil)
ok, err = pcall(rust_panic_function, "rust panic from lua")
-- Nothing to return, panic should be automatically resumed
"#,
)
.exec()
}) {
Ok(r) => panic!("no panic was detected: {:?}", r),
Err(p) => assert!(*p.downcast::<StdString>().unwrap() == "rust panic from lua"),
}
2021-06-18 17:13:56 -05:00
// Test disabling `catch_rust_panics` option / xpcall correctness
match catch_unwind(|| -> Result<()> {
let lua = make_lua(LuaOptions::new().catch_rust_panics(false))?;
lua.load(
r#"
local msgh_ok = false
local msgh = function(err)
msgh_ok = err ~= nil and err:find("hello") ~= nil
return err
end
local ok, err = xpcall(function(msg) error(msg) end, msgh, "hello")
assert(not ok and err:find("hello") ~= nil)
assert(msgh_ok)
ok, err = xpcall(rust_panic_function, msgh, "rust panic from lua")
-- Nothing to return, panic should be automatically resumed
"#,
)
.exec()
}) {
Ok(r) => panic!("no panic was detected: {:?}", r),
Err(p) => assert!(*p.downcast::<StdString>().unwrap() == "rust panic from lua"),
}
2019-09-28 09:23:17 -05:00
Ok(())
2017-05-21 18:50:59 -05:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_result_conversions() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
Another major API change, out of stack space is not an Err It, ahem "should not" be possible to exhaust lua stack space in normal usage, and causing stack errors to be Err is slightly obnoxious. I have been wanting to make this change for a while, and removing the callback API from tables makes this sensible *I think*. I can think of a couple of ways that this is not technically true, but I think that they are acceptable, or should be handled differently. One, you can make arbitrarily sized LuaVariadic values. I think this is maybe a bug already, because there is an argument limit in Lua which is lower than the stack limit. I'm not sure what happens there, but if it is a stack based panic, (or any panic?) it is a bug. Two, I believe that if you recurse over and over between lua -> rust -> lua -> rust etc, and call rlua API functions, you might get a stack panic. I think for trusted lua code, this is morally equivalent to a regular stack overflow in plain rust, which is already.. well it's not a panic but it's some kind of safe crash I'm not sure, so I think this is acceptable. For *untrusted* lua code, this could theoretically be a problem if the API provided a callback that would call back into lua, then some lua script could force a stack based panic. There are so many concerns with untrusted lua code, and this library is NOT safe enough yet for untrusted code (it doesn't even provide an option to limit lua to the safe API subset yet!), so this is not currently an issue. When the library provides support for "safe lua", it should come with big warnings anyway, and being able to force a stack panic is pretty minor in comparison. I think if there are other ways to cause unbounded stack usage, that it is a bug, or there can be an error just for that situation, like argument count limits. This commit also fixes several stupid bugs with tests, stack checking, and panics.
2017-06-25 15:52:32 -05:00
let globals = lua.globals();
2019-09-28 09:23:17 -05:00
let err = lua.create_function(|_, ()| {
Ok(Err::<String, _>(
"only through failure can we succeed".to_lua_err(),
))
})?;
let ok = lua.create_function(|_, ()| Ok(Ok::<_, Error>("!".to_owned())))?;
globals.set("err", err)?;
globals.set("ok", ok)?;
lua.load(
r#"
2019-09-28 09:23:17 -05:00
local r, e = err()
assert(r == nil)
assert(tostring(e):find("only through failure can we succeed") ~= nil)
2019-09-28 09:23:17 -05:00
local r, e = ok()
assert(r == "!")
assert(e == nil)
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
Ok(())
}
2017-06-21 16:53:52 -05:00
#[test]
2019-09-28 09:23:17 -05:00
fn test_num_conversion() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2017-06-21 16:53:52 -05:00
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_integer(Value::String(lua.create_string("1")?))?,
Some(1)
);
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_integer(Value::String(lua.create_string("1.0")?))?,
Some(1)
);
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_integer(Value::String(lua.create_string("1.5")?))?,
None
);
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_number(Value::String(lua.create_string("1")?))?,
Some(1.0)
);
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_number(Value::String(lua.create_string("1.0")?))?,
Some(1.0)
);
assert_eq!(
2019-09-28 09:23:17 -05:00
lua.coerce_number(Value::String(lua.create_string("1.5")?))?,
Some(1.5)
);
2019-09-28 09:23:17 -05:00
assert_eq!(lua.load("1.0").eval::<i64>()?, 1);
assert_eq!(lua.load("1.0").eval::<f64>()?, 1.0);
2020-05-08 06:42:40 -05:00
#[cfg(any(feature = "lua54", feature = "lua53"))]
2019-09-28 09:23:17 -05:00
assert_eq!(lua.load("1.0").eval::<String>()?, "1.0");
2022-05-30 13:26:12 -05:00
#[cfg(any(
feature = "lua52",
feature = "lua51",
feature = "luajit",
feature = "luau"
))]
2019-10-14 16:21:30 -05:00
assert_eq!(lua.load("1.0").eval::<String>()?, "1");
2019-09-28 09:23:17 -05:00
assert_eq!(lua.load("1.5").eval::<i64>()?, 1);
assert_eq!(lua.load("1.5").eval::<f64>()?, 1.5);
assert_eq!(lua.load("1.5").eval::<String>()?, "1.5");
2019-09-28 09:23:17 -05:00
assert!(lua.load("-1").eval::<u64>().is_err());
assert_eq!(lua.load("-1").eval::<i64>()?, -1);
2019-09-28 09:23:17 -05:00
assert!(lua.unpack::<u64>(lua.pack(1u128 << 64)?).is_err());
assert!(lua.load("math.huge").eval::<i64>().is_err());
2019-09-28 09:23:17 -05:00
assert_eq!(lua.unpack::<f64>(lua.pack(f32::MAX)?)?, f32::MAX as f64);
assert_eq!(lua.unpack::<f64>(lua.pack(f32::MIN)?)?, f32::MIN as f64);
assert_eq!(lua.unpack::<f32>(lua.pack(f64::MAX)?)?, f32::INFINITY);
assert_eq!(lua.unpack::<f32>(lua.pack(f64::MIN)?)?, f32::NEG_INFINITY);
2018-09-26 20:13:25 -05:00
2019-09-28 09:23:17 -05:00
assert_eq!(lua.unpack::<i128>(lua.pack(1i128 << 64)?)?, 1i128 << 64);
Ok(())
2017-06-21 16:53:52 -05:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_pcall_xpcall() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
let globals = lua.globals();
// make sure that we handle not enough arguments
2019-09-28 09:23:17 -05:00
assert!(lua.load("pcall()").exec().is_err());
assert!(lua.load("xpcall()").exec().is_err());
assert!(lua.load("xpcall(function() end)").exec().is_err());
2021-06-18 17:13:56 -05:00
// Lua >= 5.2 compatible version of xpcall for 5.1
#[cfg(feature = "lua51")]
2019-10-14 16:21:30 -05:00
lua.load(
r#"
local xpcall_orig = xpcall
function xpcall(f, err, ...)
return xpcall_orig(function() return f(unpack(arg)) end, err)
end
"#,
)
.exec()?;
// Make sure that the return values from are correct on success
2018-08-05 08:51:39 -05:00
let (r, e) = lua
2019-09-28 09:23:17 -05:00
.load("pcall(function(p) return p end, 'foo')")
.eval::<(bool, String)>()?;
assert!(r);
assert_eq!(e, "foo");
2018-08-05 08:51:39 -05:00
let (r, e) = lua
2019-09-28 09:23:17 -05:00
.load("xpcall(function(p) return p end, print, 'foo')")
.eval::<(bool, String)>()?;
assert!(r);
assert_eq!(e, "foo");
// Make sure that the return values are correct on errors, and that error handling works
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
pcall_error = nil
pcall_status, pcall_error = pcall(error, "testerror")
2019-09-28 09:23:17 -05:00
xpcall_error = nil
xpcall_status, _ = xpcall(error, function(err) xpcall_error = err end, "testerror")
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
2019-09-28 09:23:17 -05:00
assert_eq!(globals.get::<_, bool>("pcall_status")?, false);
assert_eq!(globals.get::<_, String>("pcall_error")?, "testerror");
2019-09-28 09:23:17 -05:00
assert_eq!(globals.get::<_, bool>("xpcall_statusr")?, false);
2020-05-08 06:42:40 -05:00
#[cfg(any(
feature = "lua54",
feature = "lua53",
feature = "lua52",
feature = "luajit"
))]
2019-10-16 04:56:44 -05:00
assert_eq!(
globals.get::<_, std::string::String>("xpcall_error")?,
"testerror"
);
#[cfg(feature = "lua51")]
2019-10-14 16:21:30 -05:00
assert!(globals
.get::<_, String>("xpcall_error")?
.to_str()?
.ends_with(": testerror"));
2017-08-02 14:56:16 -05:00
// Make sure that weird xpcall error recursion at least doesn't cause unsafety or panics.
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
function xpcall_recursion()
xpcall(error, function(err) error(err) end, "testerror")
end
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
2017-08-02 14:56:16 -05:00
let _ = globals
2019-09-28 09:23:17 -05:00
.get::<_, Function>("xpcall_recursion")?
2017-08-02 14:56:16 -05:00
.call::<_, ()>(());
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_recursive_mut_callback_error() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
let mut v = Some(Box::new(123));
2019-09-28 09:23:17 -05:00
let f = lua.create_function_mut::<_, (), _>(move |lua, mutate: bool| {
if mutate {
v = None;
} else {
// Produce a mutable reference
let r = v.as_mut().unwrap();
// Whoops, this will recurse into the function and produce another mutable reference!
lua.globals().get::<_, Function>("f")?.call::<_, ()>(true)?;
println!("Should not get here, mutable aliasing has occurred!");
println!("value at {:p}", r as *mut _);
println!("value is {}", r);
}
2018-08-05 08:51:39 -05:00
2019-09-28 09:23:17 -05:00
Ok(())
})?;
lua.globals().set("f", f)?;
match lua.globals().get::<_, Function>("f")?.call::<_, ()>(false) {
Err(Error::CallbackError { ref cause, .. }) => match *cause.as_ref() {
Error::CallbackError { ref cause, .. } => match *cause.as_ref() {
Error::RecursiveMutCallback { .. } => {}
ref other => panic!("incorrect result: {:?}", other),
},
ref other => panic!("incorrect result: {:?}", other),
},
other => panic!("incorrect result: {:?}", other),
};
2019-09-28 09:23:17 -05:00
Ok(())
}
2017-11-07 22:13:52 -06:00
#[test]
2019-09-28 09:23:17 -05:00
fn test_set_metatable_nil() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
lua.load(
2017-12-02 17:56:14 -06:00
r#"
2017-11-07 22:13:52 -06:00
a = {}
setmetatable(a, nil)
2017-12-02 17:56:14 -06:00
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
Ok(())
2017-11-07 22:13:52 -06:00
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_named_registry_value() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
lua.set_named_registry_value::<_, i32>("test", 42)?;
let f = lua.create_function(move |lua, ()| {
assert_eq!(lua.named_registry_value::<_, i32>("test")?, 42);
Ok(())
})?;
2019-09-28 09:23:17 -05:00
f.call::<_, ()>(())?;
2019-09-28 09:23:17 -05:00
lua.unset_named_registry_value("test")?;
match lua.named_registry_value("test")? {
2017-12-16 17:05:53 -06:00
Nil => {}
val => panic!("registry value was not Nil, was {:?}", val),
};
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_registry_value() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
let mut r = Some(lua.create_registry_value::<i32>(42)?);
let f = lua.create_function_mut(move |lua, ()| {
if let Some(r) = r.take() {
assert_eq!(lua.registry_value::<i32>(&r)?, 42);
lua.remove_registry_value(r).unwrap();
} else {
panic!();
}
Ok(())
})?;
2019-09-28 09:23:17 -05:00
f.call::<_, ()>(())?;
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_drop_registry_value() -> Result<()> {
2020-05-05 20:32:05 -05:00
struct MyUserdata(Arc<()>);
impl UserData for MyUserdata {}
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2020-05-05 20:32:05 -05:00
let rc = Arc::new(());
2019-09-28 09:23:17 -05:00
let r = lua.create_registry_value(MyUserdata(rc.clone()))?;
2020-05-05 20:32:05 -05:00
assert_eq!(Arc::strong_count(&rc), 2);
drop(r);
lua.expire_registry_values();
2019-09-28 09:23:17 -05:00
lua.load(r#"collectgarbage("collect")"#).exec()?;
2020-05-05 20:32:05 -05:00
assert_eq!(Arc::strong_count(&rc), 1);
2019-09-28 09:23:17 -05:00
Ok(())
2018-02-05 23:54:04 -06:00
}
2022-02-14 14:51:24 -06:00
#[test]
fn test_replace_registry_value() -> Result<()> {
let lua = Lua::new();
let key = lua.create_registry_value::<i32>(42)?;
lua.replace_registry_value(&key, "new value")?;
assert_eq!(lua.registry_value::<String>(&key)?, "new value");
lua.replace_registry_value(&key, Value::Nil)?;
assert_eq!(lua.registry_value::<Value>(&key)?, Value::Nil);
lua.replace_registry_value(&key, 123)?;
assert_eq!(lua.registry_value::<i32>(&key)?, 123);
// It should be impossible to replace (initial) nil value with non-nil
let key2 = lua.create_registry_value(Value::Nil)?;
match lua.replace_registry_value(&key2, "abc") {
Err(Error::RuntimeError(_)) => {}
r => panic!("expected RuntimeError, got {r:?}"),
}
2022-02-14 14:51:24 -06:00
Ok(())
}
#[test]
fn test_lua_registry_hash() -> Result<()> {
let lua = Lua::new();
let r1 = Arc::new(lua.create_registry_value("value1")?);
let r2 = Arc::new(lua.create_registry_value("value2")?);
let mut map = HashMap::new();
map.insert(r1.clone(), "value1");
map.insert(r2.clone(), "value2");
assert_eq!(map[&r1], "value1");
assert_eq!(map[&r2], "value2");
Ok(())
}
2018-02-05 23:54:04 -06:00
#[test]
2019-09-28 09:23:17 -05:00
fn test_lua_registry_ownership() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua1 = Lua::new();
let lua2 = Lua::new();
2018-02-05 23:54:04 -06:00
2019-09-28 09:23:17 -05:00
let r1 = lua1.create_registry_value("hello")?;
let r2 = lua2.create_registry_value("hello")?;
2018-02-05 23:54:04 -06:00
assert!(lua1.owns_registry_value(&r1));
assert!(!lua2.owns_registry_value(&r1));
assert!(lua2.owns_registry_value(&r2));
assert!(!lua1.owns_registry_value(&r2));
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2019-09-28 09:23:17 -05:00
fn test_mismatched_registry_key() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua1 = Lua::new();
let lua2 = Lua::new();
2019-09-28 09:23:17 -05:00
let r = lua1.create_registry_value("hello")?;
match lua2.remove_registry_value(r) {
Err(Error::MismatchedRegistryKey) => {}
r => panic!("wrong result type for mismatched registry key, {:?}", r),
};
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
fn test_registry_value_reuse() -> Result<()> {
let lua = Lua::new();
let r1 = lua.create_registry_value("value1")?;
let r1_slot = format!("{r1:?}");
drop(r1);
// Previous slot must not be reused by nil value
let r2 = lua.create_registry_value(Value::Nil)?;
let r2_slot = format!("{r2:?}");
assert_ne!(r1_slot, r2_slot);
drop(r2);
// But should be reused by non-nil value
let r3 = lua.create_registry_value("value3")?;
let r3_slot = format!("{r3:?}");
assert_eq!(r1_slot, r3_slot);
Ok(())
}
2021-09-20 18:26:48 -05:00
#[test]
fn test_application_data() -> Result<()> {
let lua = Lua::new();
lua.set_app_data("test1");
lua.set_app_data(vec!["test2"]);
let f = lua.create_function(|lua, ()| {
{
let data1 = lua.app_data_ref::<&str>().unwrap();
assert_eq!(*data1, "test1");
}
let mut data2 = lua.app_data_mut::<Vec<&str>>().unwrap();
assert_eq!(*data2, vec!["test2"]);
data2.push("test3");
Ok(())
})?;
f.call(())?;
assert_eq!(*lua.app_data_ref::<&str>().unwrap(), "test1");
assert_eq!(
*lua.app_data_ref::<Vec<&str>>().unwrap(),
vec!["test2", "test3"]
);
lua.remove_app_data::<Vec<&str>>();
assert!(matches!(lua.app_data_ref::<Vec<&str>>(), None));
Ok(())
}
#[test]
fn test_recursion() -> Result<()> {
let lua = Lua::new();
let f = lua.create_function(move |lua, i: i32| {
if i < 64 {
lua.globals()
.get::<_, Function>("f")?
.call::<_, ()>(i + 1)?;
}
Ok(())
})?;
lua.globals().set("f", f.clone())?;
f.call::<_, ()>(1)?;
Ok(())
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_too_many_returns() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
let f = lua.create_function(|_, ()| Ok(Variadic::from_iter(1..1000000)))?;
assert!(f.call::<_, Vec<u32>>(()).is_err());
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_too_many_arguments() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
lua.load("function test(...) end").exec()?;
let args = Variadic::from_iter(1..1000000);
2019-09-27 11:38:24 -05:00
assert!(lua
.globals()
2019-09-28 09:23:17 -05:00
.get::<_, Function>("test")?
2019-09-27 11:38:24 -05:00
.call::<_, ()>(args)
.is_err());
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2019-10-16 04:56:44 -05:00
#[cfg(not(feature = "luajit"))]
2021-06-18 17:13:56 -05:00
fn test_too_many_recursions() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2021-06-18 17:13:56 -05:00
2018-08-05 08:51:39 -05:00
let f = lua
2019-09-28 09:23:17 -05:00
.create_function(move |lua, ()| lua.globals().get::<_, Function>("f")?.call::<_, ()>(()))?;
2021-06-18 17:13:56 -05:00
lua.globals().set("f", f.clone())?;
assert!(f.call::<_, ()>(()).is_err());
2019-09-28 09:23:17 -05:00
Ok(())
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_too_many_binds() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
let globals = lua.globals();
2019-09-28 09:23:17 -05:00
lua.load(
r#"
2019-09-28 09:23:17 -05:00
function f(...)
end
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.exec()?;
2019-09-28 09:23:17 -05:00
let concat = globals.get::<_, Function>("f")?;
assert!(concat.bind(Variadic::from_iter(1..1000000)).is_err());
2019-09-27 11:38:24 -05:00
assert!(concat
.call::<_, ()>(Variadic::from_iter(1..1000000))
.is_err());
2019-09-28 09:23:17 -05:00
Ok(())
}
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
#[test]
fn test_ref_stack_exhaustion() {
match catch_unwind(AssertUnwindSafe(|| -> Result<()> {
let lua = Lua::new();
let mut vals = Vec::new();
for _ in 0..1000000 {
vals.push(lua.create_table()?);
}
Ok(())
})) {
Ok(_) => panic!("no panic was detected"),
Err(p) => assert!(p
.downcast::<StdString>()
.unwrap()
.starts_with("cannot create a Lua reference, out of auxiliary stack space")),
}
}
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
#[test]
2021-06-18 17:13:56 -05:00
fn test_large_args() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
let globals = lua.globals();
2019-09-28 09:23:17 -05:00
globals.set(
"c",
lua.create_function(|_, args: Variadic<usize>| {
let mut s = 0;
for i in 0..args.len() {
s += i;
assert_eq!(i, args[i]);
}
Ok(s)
})?,
)?;
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
2018-09-24 21:13:42 -05:00
let f: Function = lua
2019-09-28 09:23:17 -05:00
.load(
2018-08-05 08:51:39 -05:00
r#"
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
return function(...)
return c(...)
end
"#,
2019-09-27 11:38:24 -05:00
)
2019-09-28 09:23:17 -05:00
.eval()?;
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
assert_eq!(
2019-09-28 09:23:17 -05:00
f.call::<_, usize>((0..100).collect::<Variadic<usize>>())?,
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
4950
);
2019-09-28 09:23:17 -05:00
Ok(())
A lot of performance changes. Okay, so this is kind of a mega-commit of a lot of performance related changes to rlua, some of which are pretty complicated. There are some small improvements here and there, but most of the benefits of this change are from a few big changes. The simplest big change is that there is now `protect_lua` as well as `protect_lua_call`, which allows skipping a lightuserdata parameter and some stack manipulation in some cases. Second simplest is the change to use Vec instead of VecDeque for MultiValue, and to have MultiValue be used as a sort of "backwards-only" Vec so that ToLuaMulti / FromLuaMulti still work correctly. The most complex change, though, is a change to the way LuaRef works, so that LuaRef can optionally point into the Lua stack instead of only registry values. At state creation a set number of stack slots is reserved for the first N LuaRef types (currently 16), and space for these are also allocated separately allocated at callback time. There is a huge breaking change here, which is that now any LuaRef types MUST only be used with the Lua on which they were created, and CANNOT be used with any other Lua callback instance. This mostly will affect people using LuaRef types from inside a scope callback, but hopefully in those cases `Function::bind` will be a suitable replacement. On the plus side, the rules for LuaRef types are easier to state now. There is probably more easy-ish perf on the table here, but here's the preliminary results, based on my very limited benchmarks: create table time: [314.13 ns 315.71 ns 317.44 ns] change: [-36.154% -35.670% -35.205%] (p = 0.00 < 0.05) create array 10 time: [2.9731 us 2.9816 us 2.9901 us] change: [-16.996% -16.600% -16.196%] (p = 0.00 < 0.05) Performance has improved. create string table 10 time: [5.6904 us 5.7164 us 5.7411 us] change: [-53.536% -53.309% -53.079%] (p = 0.00 < 0.05) Performance has improved. call add function 3 10 time: [5.1134 us 5.1222 us 5.1320 us] change: [-4.1095% -3.6910% -3.1781%] (p = 0.00 < 0.05) Performance has improved. call callback add 2 10 time: [5.4408 us 5.4480 us 5.4560 us] change: [-6.4203% -5.7780% -5.0013%] (p = 0.00 < 0.05) Performance has improved. call callback append 10 time: [9.8243 us 9.8410 us 9.8586 us] change: [-26.937% -26.702% -26.469%] (p = 0.00 < 0.05) Performance has improved. create registry 10 time: [3.7005 us 3.7089 us 3.7174 us] change: [-8.4965% -8.1042% -7.6926%] (p = 0.00 < 0.05) Performance has improved. I think that a lot of these benchmarks are too "easy", and most API usage is going to be more like the 'create string table 10' benchmark, where there are a lot of handles and tables and strings, so I think that 25%-50% improvement is a good guess for most use cases.
2018-03-11 22:20:10 -05:00
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_large_args_ref() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
let f = lua.create_function(|_, args: Variadic<String>| {
for i in 0..args.len() {
assert_eq!(args[i], i.to_string());
}
Ok(())
})?;
f.call::<_, ()>((0..100).map(|i| i.to_string()).collect::<Variadic<_>>())?;
Ok(())
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_chunk_env() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2019-09-28 09:23:17 -05:00
let assert: Function = lua.globals().get("assert")?;
let env1 = lua.create_table()?;
env1.set("assert", assert.clone())?;
let env2 = lua.create_table()?;
env2.set("assert", assert)?;
lua.load(
r#"
test_var = 1
"#,
)
.set_environment(env1.clone())?
.exec()?;
lua.load(
r#"
assert(test_var == nil)
test_var = 2
"#,
)
.set_environment(env2.clone())?
.exec()?;
assert_eq!(
lua.load("test_var").set_environment(env1)?.eval::<i32>()?,
1
);
assert_eq!(
lua.load("test_var").set_environment(env2)?.eval::<i32>()?,
2
);
Ok(())
}
#[test]
2021-06-18 17:13:56 -05:00
fn test_context_thread() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
2018-09-24 21:13:42 -05:00
let f = lua
2019-09-28 09:23:17 -05:00
.load(
r#"
local thread = ...
assert(coroutine.running() == thread)
"#,
)
.into_function()?;
#[cfg(any(
feature = "lua54",
feature = "lua53",
feature = "lua52",
feature = "luajit52"
))]
2019-09-28 09:23:17 -05:00
f.call::<_, ()>(lua.current_thread())?;
2022-02-19 08:15:15 -06:00
#[cfg(any(
feature = "lua51",
all(feature = "luajit", not(feature = "luajit52")),
feature = "luau"
))]
2019-10-14 16:21:30 -05:00
f.call::<_, ()>(Nil)?;
Ok(())
}
#[test]
#[cfg(any(feature = "lua51", all(feature = "luajit", not(feature = "luajit52"))))]
2021-06-18 17:13:56 -05:00
fn test_context_thread_51() -> Result<()> {
2019-10-14 16:21:30 -05:00
let lua = Lua::new();
let thread = lua.create_thread(
lua.load(
r#"
function (thread)
assert(coroutine.running() == thread)
end
"#,
)
.eval()?,
)?;
thread.resume::<_, ()>(thread.clone())?;
2019-09-28 09:23:17 -05:00
Ok(())
}
2020-06-07 12:48:22 -05:00
#[test]
#[cfg(feature = "luajit")]
fn test_jit_version() -> Result<()> {
let lua = Lua::new();
let jit: Table = lua.globals().get("jit")?;
assert!(jit
.get::<_, String>("version")?
.to_str()?
.contains("LuaJIT"));
Ok(())
}
#[test]
fn test_load_from_function() -> Result<()> {
let lua = Lua::new();
let i = Arc::new(AtomicU32::new(0));
let i2 = i.clone();
let func = lua.create_function(move |lua, modname: String| {
i2.fetch_add(1, Ordering::Relaxed);
let t = lua.create_table()?;
t.set("__name", modname)?;
Ok(t)
})?;
let t: Table = lua.load_from_function("my_module", func.clone())?;
assert_eq!(t.get::<_, String>("__name")?, "my_module");
assert_eq!(i.load(Ordering::Relaxed), 1);
2021-11-23 13:12:27 -06:00
let _: Value = lua.load_from_function("my_module", func.clone())?;
assert_eq!(i.load(Ordering::Relaxed), 1);
let func_nil = lua.create_function(move |_, _: String| Ok(Value::Nil))?;
let v: Value = lua.load_from_function("my_module2", func_nil)?;
assert_eq!(v, Value::Boolean(true));
2021-11-23 13:12:27 -06:00
// Test unloading and loading again
lua.unload("my_module")?;
2021-11-23 09:20:54 -06:00
let _: Value = lua.load_from_function("my_module", func)?;
assert_eq!(i.load(Ordering::Relaxed), 2);
2021-11-23 13:12:27 -06:00
// Unloading nonexistent module must not fail
lua.unload("my_module2")?;
2021-11-23 09:20:54 -06:00
Ok(())
}
#[test]
fn test_inspect_stack() -> Result<()> {
let lua = Lua::new();
// Not inside any function
assert!(lua.inspect_stack(0).is_none());
let logline = lua.create_function(|lua, msg: StdString| {
let debug = lua.inspect_stack(1).unwrap(); // caller
let source = debug.source().short_src.map(core::str::from_utf8);
let source = source.transpose().unwrap().unwrap_or("?");
let line = debug.curr_line();
Ok(format!("{}:{} {}", source, line, msg))
})?;
lua.globals().set("logline", logline)?;
lua.load(
r#"
local function foo()
local line = logline("hello")
return line
end
local function bar()
return foo()
end
assert(foo() == '[string "chunk"]:3 hello')
assert(bar() == '[string "chunk"]:3 hello')
assert(logline("world") == '[string "chunk"]:12 world')
"#,
)
.set_name("chunk")?
.exec()?;
Ok(())
}
#[test]
fn test_multi_states() -> Result<()> {
let lua = Lua::new();
let f = lua.create_function(|_, g: Option<Function>| {
if let Some(g) = g {
g.call(())?;
}
Ok(())
})?;
lua.globals().set("f", f)?;
lua.load("f(function() coroutine.wrap(function() f() end)() end)")
.exec()?;
Ok(())
}
#[test]
#[cfg(feature = "lua54")]
fn test_warnings() -> Result<()> {
let lua = Lua::new();
lua.set_app_data::<Vec<(StdString, bool)>>(Vec::new());
lua.set_warning_function(|lua, msg, tocont| {
let msg = msg.to_string_lossy().to_string();
lua.app_data_mut::<Vec<(StdString, bool)>>()
.unwrap()
.push((msg, tocont));
Ok(())
});
lua.warning("native warning ...", true)?;
lua.warning("finish", false)?;
lua.load(r#"warn("lua warning", "continue")"#).exec()?;
lua.remove_warning_function();
lua.warning("one more warning", false)?;
let messages = lua.app_data_ref::<Vec<(StdString, bool)>>().unwrap();
assert_eq!(
*messages,
vec![
("native warning ...".to_string(), true),
("finish".to_string(), false),
("lua warning".to_string(), true),
("continue".to_string(), false),
]
);
// Trigger error inside warning
lua.set_warning_function(|_, _, _| Err(Error::RuntimeError("warning error".to_string())));
assert!(matches!(
lua.load(r#"warn("test")"#).exec(),
Err(Error::CallbackError { cause, .. })
if matches!(*cause, Error::RuntimeError(ref err) if err == "warning error")
));
Ok(())
}
#[test]
#[cfg(feature = "luajit")]
#[should_panic]
fn test_luajit_cdata() {
let lua = unsafe { Lua::unsafe_new() };
let _v: Result<Value> = lua
.load(
r#"
local ffi = require("ffi")
ffi.cdef[[
void *malloc(size_t size);
void free(void *ptr);
]]
local ptr = ffi.C.malloc(1)
ffi.C.free(ptr)
return ptr
"#,
)
.eval();
}
2022-05-26 16:22:23 -05:00
#[test]
#[cfg(feature = "send")]
fn test_send() {
let lua = Lua::new();
std::thread::spawn(move || {
let _lua = lua;
})
.join()
.unwrap();
}