rust/src/libcore/task.rs

1791 lines
55 KiB
Rust

/*!
* Task management.
*
* An executing Rust program consists of a tree of tasks, each with their own
* stack, and sole ownership of their allocated heap data. Tasks communicate
* with each other using ports and channels.
*
* When a task fails, that failure will propagate to its parent (the task
* that spawned it) and the parent will fail as well. The reverse is not
* true: when a parent task fails its children will continue executing. When
* the root (main) task fails, all tasks fail, and then so does the entire
* process.
*
* Tasks may execute in parallel and are scheduled automatically by the
* runtime.
*
* # Example
*
* ~~~
* do spawn {
* log(error, "Hello, World!");
* }
* ~~~
*/
import result::result;
import dvec::extensions;
import dvec_iter::extensions;
import arc::methods;
export task;
export task_result;
export notification;
export sched_mode;
export sched_opts;
export task_opts;
export task_builder;
export default_task_opts;
export get_opts;
export set_opts;
export set_sched_mode;
export add_wrapper;
export run;
export future_result;
export run_listener;
export run_with;
export spawn;
export spawn_unlinked;
export spawn_with;
export spawn_listener;
export spawn_sched;
export try;
export yield;
export failing;
export get_task;
export unkillable;
export atomically;
export local_data_key;
export local_data_pop;
export local_data_get;
export local_data_set;
export local_data_modify;
export single_threaded;
export thread_per_core;
export thread_per_task;
export manual_threads;
export osmain;
/* Data types */
/// A handle to a task
enum task { task_handle(task_id) }
/**
* Indicates the manner in which a task exited.
*
* A task that completes without failing and whose supervised children
* complete without failing is considered to exit successfully.
*
* FIXME (See #1868): This description does not indicate the current behavior
* for linked failure.
*/
enum task_result {
success,
failure,
}
/// A message type for notifying of task lifecycle events
enum notification {
/// Sent when a task exits with the task handle and result
exit(task, task_result)
}
/// Scheduler modes
enum sched_mode {
/// All tasks run in the same OS thread
single_threaded,
/// Tasks are distributed among available CPUs
thread_per_core,
/// Each task runs in its own OS thread
thread_per_task,
/// Tasks are distributed among a fixed number of OS threads
manual_threads(uint),
/**
* Tasks are scheduled on the main OS thread
*
* The main OS thread is the thread used to launch the runtime which,
* in most cases, is the process's initial thread as created by the OS.
*/
osmain
}
/**
* Scheduler configuration options
*
* # Fields
*
* * sched_mode - The operating mode of the scheduler
*
* * foreign_stack_size - The size of the foreign stack, in bytes
*
* Rust code runs on Rust-specific stacks. When Rust code calls foreign
* code (via functions in foreign modules) it switches to a typical, large
* stack appropriate for running code written in languages like C. By
* default these foreign stacks have unspecified size, but with this
* option their size can be precisely specified.
*/
type sched_opts = {
mode: sched_mode,
foreign_stack_size: option<uint>
};
/**
* Task configuration options
*
* # Fields
*
* * linked - Do not propagate failure to the parent task
*
* All tasks are linked together via a tree, from parents to children. By
* default children are 'supervised' by their parent and when they fail
* so too will their parents. Settings this flag to false disables that
* behavior.
*
* * notify_chan - Enable lifecycle notifications on the given channel
*
* * sched - Specify the configuration of a new scheduler to create the task
* in
*
* By default, every task is created in the same scheduler as its
* parent, where it is scheduled cooperatively with all other tasks
* in that scheduler. Some specialized applications may want more
* control over their scheduling, in which case they can be spawned
* into a new scheduler with the specific properties required.
*
* This is of particular importance for libraries which want to call
* into foreign code that blocks. Without doing so in a different
* scheduler other tasks will be impeded or even blocked indefinitely.
*/
type task_opts = {
linked: bool,
parented: bool,
notify_chan: option<comm::chan<notification>>,
sched: option<sched_opts>,
};
/**
* The task builder type.
*
* Provides detailed control over the properties and behavior of new tasks.
*/
// NB: Builders are designed to be single-use because they do stateful
// things that get weird when reusing - e.g. if you create a result future
// it only applies to a single task, so then you have to maintain some
// potentially tricky state to ensure that everything behaves correctly
// when you try to reuse the builder to spawn a new task. We'll just
// sidestep that whole issue by making builders uncopyable and making
// the run function move them in.
class dummy { let x: (); new() { self.x = (); } drop { } }
// FIXME (#2585): Replace the 'consumed' bit with move mode on self
enum task_builder = {
opts: task_opts,
gen_body: fn@(+fn~()) -> fn~(),
can_not_copy: option<dummy>,
mut consumed: bool,
};
/**
* Generate the base configuration for spawning a task, off of which more
* configuration methods can be chained.
* For example, task().unlinked().spawn is equivalent to spawn_unlinked.
*/
fn task() -> task_builder {
task_builder({
opts: default_task_opts(),
gen_body: |body| body, // Identity function
can_not_copy: none,
mut consumed: false,
})
}
impl private_methods for task_builder {
fn consume() -> task_builder {
if self.consumed {
fail ~"Cannot copy a task_builder"; // Fake move mode on self
}
self.consumed = true;
task_builder({ can_not_copy: none, mut consumed: false, with *self })
}
}
impl task_builder for task_builder {
/**
* Decouple the child task's failure from the parent's. If either fails,
* the other will not be killed.
*/
fn unlinked() -> task_builder {
task_builder({
opts: { linked: false with self.opts },
can_not_copy: none,
with *self.consume()
})
}
/**
* Unidirectionally link the child task's failure with the parent's. The
* child's failure will not kill the parent, but the parent's will kill
* the child.
*/
fn supervised() -> task_builder {
task_builder({
opts: { linked: false, parented: true with self.opts },
can_not_copy: none,
with *self.consume()
})
}
/**
* Link the child task's and parent task's failures. If either fails, the
* other will be killed.
*/
fn linked() -> task_builder {
task_builder({
opts: { linked: true, parented: false with self.opts },
can_not_copy: none,
with *self.consume()
})
}
/**
* Get a future representing the exit status of the task.
*
* Taking the value of the future will block until the child task
* terminates. The future-receiving callback specified will be called
* *before* the task is spawned; as such, do not invoke .get() within the
* closure; rather, store it in an outer variable/list for later use.
*
* Note that the future returning by this function is only useful for
* obtaining the value of the next task to be spawning with the
* builder. If additional tasks are spawned with the same builder
* then a new result future must be obtained prior to spawning each
* task.
*/
fn future_result(blk: fn(-future::future<task_result>)) -> task_builder {
// FIXME (#1087, #1857): Once linked failure and notification are
// handled in the library, I can imagine implementing this by just
// registering an arbitrary number of task::on_exit handlers and
// sending out messages.
// Construct the future and give it to the caller.
let po = comm::port::<notification>();
let ch = comm::chan(po);
blk(do future::from_fn {
alt comm::recv(po) {
exit(_, result) { result }
}
});
// Reconfigure self to use a notify channel.
task_builder({
opts: { notify_chan: some(ch) with self.opts },
can_not_copy: none,
with *self.consume()
})
}
/// Configure a custom scheduler mode for the task.
fn sched_mode(mode: sched_mode) -> task_builder {
task_builder({
opts: { sched: some({ mode: mode, foreign_stack_size: none})
with self.opts },
can_not_copy: none,
with *self.consume()
})
}
/**
* Add a wrapper to the body of the spawned task.
*
* Before the task is spawned it is passed through a 'body generator'
* function that may perform local setup operations as well as wrap
* the task body in remote setup operations. With this the behavior
* of tasks can be extended in simple ways.
*
* This function augments the current body generator with a new body
* generator by applying the task body which results from the
* existing body generator to the new body generator.
*/
fn add_wrapper(wrapper: fn@(+fn~()) -> fn~()) -> task_builder {
let prev_gen_body = self.gen_body;
task_builder({
gen_body: |body| { wrapper(prev_gen_body(body)) },
can_not_copy: none,
with *self.consume()
})
}
/**
* Creates and exucutes a new child task
*
* Sets up a new task with its own call stack and schedules it to run
* the provided unique closure. The task has the properties and behavior
* specified by the task_builder.
*
* # Failure
*
* When spawning into a new scheduler, the number of threads requested
* must be greater than zero.
*/
fn spawn(+f: fn~()) {
let x = self.consume();
spawn_raw(x.opts, x.gen_body(f));
}
/// Runs a task, while transfering ownership of one argument to the child.
fn spawn_with<A: send>(+arg: A, +f: fn~(+A)) {
let arg = ~mut some(arg);
do self.spawn {
let mut my_arg = none;
my_arg <-> *arg;
f(option::unwrap(my_arg))
}
}
/**
* Runs a new task while providing a channel from the parent to the child
*
* Sets up a communication channel from the current task to the new
* child task, passes the port to child's body, and returns a channel
* linked to the port to the parent.
*
* This encapsulates some boilerplate handshaking logic that would
* otherwise be required to establish communication from the parent
* to the child.
*/
fn spawn_listener<A: send>(+f: fn~(comm::port<A>)) -> comm::chan<A> {
let setup_po = comm::port();
let setup_ch = comm::chan(setup_po);
do self.spawn {
let po = comm::port();
let ch = comm::chan(po);
comm::send(setup_ch, ch);
f(po);
}
comm::recv(setup_po)
}
}
/* Task construction */
fn default_task_opts() -> task_opts {
/*!
* The default task options
*
* By default all tasks are supervised by their parent, are spawned
* into the same scheduler, and do not post lifecycle notifications.
*/
{
linked: true,
parented: false,
notify_chan: none,
sched: none
}
}
/* Spawn convenience functions */
fn spawn(+f: fn~()) {
/*!
* Creates and executes a new child task
*
* Sets up a new task with its own call stack and schedules it to run
* the provided unique closure.
*
* This function is equivalent to `task().spawn(f)`.
*/
task().spawn(f)
}
fn spawn_unlinked(+f: fn~()) {
/*!
* Creates a child task unlinked from the current one. If either this
* task or the child task fails, the other will not be killed.
*/
task().unlinked().spawn(f)
}
fn spawn_supervised(+f: fn~()) {
/*!
* Creates a child task unlinked from the current one. If either this
* task or the child task fails, the other will not be killed.
*/
task().supervised().spawn(f)
}
fn spawn_with<A:send>(+arg: A, +f: fn~(+A)) {
/*!
* Runs a task, while transfering ownership of one argument to the
* child.
*
* This is useful for transfering ownership of noncopyables to
* another task.
*
* This function is equivalent to `task().spawn_with(arg, f)`.
*/
task().spawn_with(arg, f)
}
fn spawn_listener<A:send>(+f: fn~(comm::port<A>)) -> comm::chan<A> {
/*!
* Runs a new task while providing a channel from the parent to the child
*
* Sets up a communication channel from the current task to the new
* child task, passes the port to child's body, and returns a channel
* linked to the port to the parent.
*
* This encapsulates some boilerplate handshaking logic that would
* otherwise be required to establish communication from the parent
* to the child.
*
* The simplest way to establish bidirectional communication between
* a parent in child is as follows:
*
* let po = comm::port();
* let ch = comm::chan(po);
* let ch = do spawn_listener |po| {
* // Now the child has a port called 'po' to read from and
* // an environment-captured channel called 'ch'.
* };
* // Likewise, the parent has both a 'po' and 'ch'
*
* This function is equivalent to `task().spawn_listener(f)`.
*/
task().spawn_listener(f)
}
fn spawn_sched(mode: sched_mode, +f: fn~()) {
/*!
* Creates a new scheduler and executes a task on it
*
* Tasks subsequently spawned by that task will also execute on
* the new scheduler. When there are no more tasks to execute the
* scheduler terminates.
*
* # Failure
*
* In manual threads mode the number of threads requested must be
* greater than zero.
*/
task().sched_mode(mode).spawn(f)
}
fn try<T:send>(+f: fn~() -> T) -> result<T,()> {
/*!
* Execute a function in another task and return either the return value
* of the function or result::err.
*
* # Return value
*
* If the function executed successfully then try returns result::ok
* containing the value returned by the function. If the function fails
* then try returns result::err containing nil.
*/
let po = comm::port();
let ch = comm::chan(po);
let mut result = none;
do task().unlinked().future_result(|-r| { result = some(r); }).spawn {
comm::send(ch, f());
}
alt future::get(option::unwrap(result)) {
success { result::ok(comm::recv(po)) }
failure { result::err(()) }
}
}
/* Lifecycle functions */
fn yield() {
//! Yield control to the task scheduler
let task_ = rustrt::rust_get_task();
let killed = rustrt::rust_task_yield(task_);
if killed && !failing() {
fail ~"killed";
}
}
fn failing() -> bool {
//! True if the running task has failed
rustrt::rust_task_is_unwinding(rustrt::rust_get_task())
}
fn get_task() -> task {
//! Get a handle to the running task
task_handle(rustrt::get_task_id())
}
/**
* Temporarily make the task unkillable
*
* # Example
*
* ~~~
* do task::unkillable {
* // detach / yield / destroy must all be called together
* rustrt::rust_port_detach(po);
* // This must not result in the current task being killed
* task::yield();
* rustrt::rust_port_destroy(po);
* }
* ~~~
*/
unsafe fn unkillable(f: fn()) {
class allow_failure {
let t: *rust_task;
new(t: *rust_task) { self.t = t; }
drop { rustrt::rust_task_allow_kill(self.t); }
}
let t = rustrt::rust_get_task();
let _allow_failure = allow_failure(t);
rustrt::rust_task_inhibit_kill(t);
f();
}
/**
* A stronger version of unkillable that also inhibits scheduling operations.
* For use with exclusive ARCs, which use pthread mutexes directly.
*/
unsafe fn atomically<U>(f: fn() -> U) -> U {
class defer_interrupts {
let t: *rust_task;
new(t: *rust_task) { self.t = t; }
drop {
rustrt::rust_task_allow_yield(self.t);
rustrt::rust_task_allow_kill(self.t);
}
}
let t = rustrt::rust_get_task();
let _interrupts = defer_interrupts(t);
rustrt::rust_task_inhibit_kill(t);
rustrt::rust_task_inhibit_yield(t);
f()
}
/****************************************************************************
* Internal
****************************************************************************/
/* spawning */
type sched_id = int;
type task_id = int;
// These are both opaque runtime/compiler types that we don't know the
// structure of and should only deal with via unsafe pointer
type rust_task = libc::c_void;
type rust_closure = libc::c_void;
/* linked failure */
type taskgroup_arc =
arc::exclusive<option<(dvec::dvec<option<*rust_task>>,dvec::dvec<uint>)>>;
class taskgroup {
// FIXME (#2816): Change dvec to an O(1) data structure (and change 'me'
// to a node-handle or somesuch when so done (or remove the field entirely
// if keyed by *rust_task)).
let me: *rust_task;
// List of tasks with whose fates this one's is intertwined.
let tasks: taskgroup_arc; // 'none' means the group already failed.
let my_pos: uint; // Index into above for this task's slot.
// Lists of tasks who will kill us if they fail, but whom we won't kill.
let parents: option<(taskgroup_arc,uint)>;
let is_main: bool;
new(me: *rust_task, -tasks: taskgroup_arc, my_pos: uint,
-parents: option<(taskgroup_arc,uint)>, is_main: bool) {
self.me = me;
self.tasks = tasks;
self.my_pos = my_pos;
self.parents = parents;
self.is_main = is_main;
}
// Runs on task exit.
drop {
// If we are failing, the whole taskgroup needs to die.
if rustrt::rust_task_is_unwinding(self.me) {
// Take everybody down with us.
kill_taskgroup(self.tasks, self.me, self.my_pos, self.is_main);
} else {
// Remove ourselves from the group(s).
leave_taskgroup(self.tasks, self.me, self.my_pos);
}
// It doesn't matter whether this happens before or after dealing with
// our own taskgroup, so long as both happen before we die.
alt self.parents {
some((parent_group,pos_in_group)) {
leave_taskgroup(parent_group, self.me, pos_in_group);
}
none { }
}
}
}
fn enlist_in_taskgroup(group_arc: taskgroup_arc,
me: *rust_task) -> option<uint> {
do group_arc.with |_c, state| {
// If 'none', the group was failing. Can't enlist.
let mut newstate = none;
*state <-> newstate;
if newstate.is_some() {
let (tasks,empty_slots) = option::unwrap(newstate);
// Try to find an empty slot.
let slotno = if empty_slots.len() > 0 {
let empty_index = empty_slots.pop();
assert tasks[empty_index] == none;
tasks.set_elt(empty_index, some(me));
empty_index
} else {
tasks.push(some(me));
tasks.len() - 1
};
*state = some((tasks,empty_slots));
some(slotno)
} else {
none
}
}
}
// NB: Runs in destructor/post-exit context. Can't 'fail'.
fn leave_taskgroup(group_arc: taskgroup_arc, me: *rust_task, index: uint) {
do group_arc.with |_c, state| {
let mut newstate = none;
*state <-> newstate;
// If 'none', already failing and we've already gotten a kill signal.
if newstate.is_some() {
let (tasks,empty_slots) = option::unwrap(newstate);
assert tasks[index] == some(me);
tasks.set_elt(index, none);
empty_slots.push(index);
*state = some((tasks,empty_slots));
};
};
}
// NB: Runs in destructor/post-exit context. Can't 'fail'.
fn kill_taskgroup(group_arc: taskgroup_arc, me: *rust_task, index: uint,
is_main: bool) {
// NB: We could do the killing iteration outside of the group arc, by
// having "let mut newstate" here, swapping inside, and iterating after.
// But that would let other exiting tasks fall-through and exit while we
// were trying to kill them, causing potential use-after-free. A task's
// presence in the arc guarantees it's alive only while we hold the lock,
// so if we're failing, all concurrently exiting tasks must wait for us.
// To do it differently, we'd have to use the runtime's task refcounting.
do group_arc.with |_c, state| {
let mut newstate = none;
*state <-> newstate;
// Might already be none, if somebody is failing simultaneously.
// That's ok; only one task needs to do the dirty work. (Might also
// see 'none' if somebody already failed and we got a kill signal.)
if newstate.is_some() {
let (tasks,_empty_slots) = option::unwrap(newstate);
// First remove ourself (killing ourself won't do much good). This
// is duplicated here to avoid having to lock twice.
assert tasks[index] == some(me);
tasks.set_elt(index, none);
// Now send takedown signal.
for tasks.each |entry| {
do entry.map |task| {
rustrt::rust_task_kill_other(task);
};
}
// Only one task should ever do this.
if is_main {
rustrt::rust_task_kill_all(me);
}
// Do NOT restore state to some(..)! It stays none to indicate
// that the whole taskgroup is failing, to forbid new spawns.
}
// (note: multiple tasks may reach this point)
};
}
// FIXME (#2912): Work around core-vs-coretest function duplication. Can't use
// a proper closure because the #[test]s won't understand. Have to fake it.
unsafe fn taskgroup_key() -> local_data_key<taskgroup> {
// Use a "code pointer" value that will never be a real code pointer.
unsafe::transmute((-2 as uint, 0u))
}
fn share_parent_taskgroup() -> (taskgroup_arc, bool) {
let me = rustrt::rust_get_task();
alt unsafe { local_get(me, taskgroup_key()) } {
some(group) {
// Clone the shared state for the child; propagate main-ness.
(group.tasks.clone(), group.is_main)
}
none {
// Main task, doing first spawn ever.
let tasks = arc::exclusive(some((dvec::from_elem(some(me)),
dvec::dvec())));
// Main group has no parent group.
let group = @taskgroup(me, tasks.clone(), 0, none, true);
unsafe { local_set(me, taskgroup_key(), group); }
// Tell child task it's also in the main group.
(tasks, true)
}
}
}
fn spawn_raw(opts: task_opts, +f: fn~()) {
// Decide whether the child needs to be in a new linked failure group.
let ((child_tg, is_main), parent_tg) = if opts.linked {
// It doesn't mean anything for a linked-spawned-task to have a parent
// group. The spawning task is already bidirectionally linked to it.
(share_parent_taskgroup(), none)
} else {
// Detached from the parent group; create a new (non-main) one.
((arc::exclusive(some((dvec::dvec(),dvec::dvec()))), false),
// Allow the parent to unidirectionally fail the child?
if opts.parented {
let (pg,_) = share_parent_taskgroup(); some(pg)
} else {
none
})
};
unsafe {
let child_data_ptr = ~mut some((child_tg, parent_tg, f));
// Being killed with the unsafe task/closure pointers would leak them.
do unkillable {
// Agh. Get move-mode items into the closure. FIXME (#2829)
let mut child_data = none;
*child_data_ptr <-> child_data;
let (child_tg, parent_tg, f) = option::unwrap(child_data);
// Create child task.
let new_task = alt opts.sched {
none { rustrt::new_task() }
some(sched_opts) { new_task_in_new_sched(sched_opts) }
};
assert !new_task.is_null();
// Getting killed after here would leak the task.
let child_wrapper =
make_child_wrapper(new_task, child_tg, parent_tg, is_main, f);
let fptr = ptr::addr_of(child_wrapper);
let closure: *rust_closure = unsafe::reinterpret_cast(fptr);
do option::iter(opts.notify_chan) |c| {
// FIXME (#1087): Would like to do notification in Rust
rustrt::rust_task_config_notify(new_task, c);
}
// Getting killed between these two calls would free the child's
// closure. (Reordering them wouldn't help - then getting killed
// between them would leak.)
rustrt::start_task(new_task, closure);
unsafe::forget(child_wrapper);
}
}
// This function returns a closure-wrapper that we pass to the child task.
// In brief, it does the following:
// if enlist_in_group(child_group) {
// if parent_group {
// if !enlist_in_group(parent_group) {
// leave_group(child_group); // Roll back
// ret; // Parent group failed. Don't run child's f().
// }
// }
// stash_taskgroup_data_in_TLS(child_group, parent_group);
// f();
// } else {
// // My group failed. Don't run chid's f().
// }
fn make_child_wrapper(child: *rust_task, -child_tg: taskgroup_arc,
-parent_tg: option<taskgroup_arc>, is_main: bool,
-f: fn~()) -> fn~() {
let child_tg_ptr = ~mut some((child_tg, parent_tg));
fn~() {
// Agh. Get move-mode items into the closure. FIXME (#2829)
let mut tg_data_opt = none;
*child_tg_ptr <-> tg_data_opt;
let (child_tg, parent_tg) = option::unwrap(tg_data_opt);
// Child task runs this code.
// Set up membership in taskgroup. If this returns none, some
// task was already failing, so don't bother doing anything.
alt enlist_in_taskgroup(child_tg, child) {
some(my_pos) {
// Enlist in parent group too. If enlist returns none, a
// parent was failing: don't spawn; leave this group too.
let (pg, enlist_ok) = if parent_tg.is_some() {
let parent_group = option::unwrap(parent_tg);
alt enlist_in_taskgroup(parent_group, child) {
some(my_p_index) {
// Successful enlist.
(some((parent_group, my_p_index)), true)
}
none {
// Couldn't enlist. Have to quit here too.
leave_taskgroup(child_tg, child, my_pos);
(none, false)
}
}
} else {
// No parent group to enlist in. No worry.
(none, true)
};
if enlist_ok {
let group = @taskgroup(child, child_tg, my_pos,
pg, is_main);
unsafe { local_set(child, taskgroup_key(), group); }
// Run the child's body.
f();
// TLS cleanup code will exit the taskgroup.
}
}
none { }
}
}
}
fn new_task_in_new_sched(opts: sched_opts) -> *rust_task {
if opts.foreign_stack_size != none {
fail ~"foreign_stack_size scheduler option unimplemented";
}
let num_threads = alt opts.mode {
single_threaded { 1u }
thread_per_core {
fail ~"thread_per_core scheduling mode unimplemented"
}
thread_per_task {
fail ~"thread_per_task scheduling mode unimplemented"
}
manual_threads(threads) {
if threads == 0u {
fail ~"can not create a scheduler with no threads";
}
threads
}
osmain { 0u /* Won't be used */ }
};
let sched_id = if opts.mode != osmain {
rustrt::rust_new_sched(num_threads)
} else {
rustrt::rust_osmain_sched_id()
};
rustrt::rust_new_task_in_sched(sched_id)
}
}
/****************************************************************************
* Task local data management
*
* Allows storing boxes with arbitrary types inside, to be accessed anywhere
* within a task, keyed by a pointer to a global finaliser function. Useful
* for task-spawning metadata (tracking linked failure state), dynamic
* variables, and interfacing with foreign code with bad callback interfaces.
*
* To use, declare a monomorphic global function at the type to store, and use
* it as the 'key' when accessing. See the 'tls' tests below for examples.
*
* Casting 'Arcane Sight' reveals an overwhelming aura of Transmutation magic.
****************************************************************************/
/**
* Indexes a task-local data slot. The function's code pointer is used for
* comparison. Recommended use is to write an empty function for each desired
* task-local data slot (and use class destructors, not code inside the
* function, if specific teardown is needed). DO NOT use multiple
* instantiations of a single polymorphic function to index data of different
* types; arbitrary type coercion is possible this way. The interface is safe
* as long as all key functions are monomorphic.
*/
type local_data_key<T: owned> = fn@(+@T);
iface local_data { }
impl<T: owned> of local_data for @T { }
// We use dvec because it's the best data structure in core. If TLS is used
// heavily in future, this could be made more efficient with a proper map.
type task_local_element = (*libc::c_void, *libc::c_void, local_data);
// Has to be a pointer at outermost layer; the foreign call returns void *.
type task_local_map = @dvec::dvec<option<task_local_element>>;
extern fn cleanup_task_local_map(map_ptr: *libc::c_void) unsafe {
assert !map_ptr.is_null();
// Get and keep the single reference that was created at the beginning.
let _map: task_local_map = unsafe::reinterpret_cast(map_ptr);
// All local_data will be destroyed along with the map.
}
// Gets the map from the runtime. Lazily initialises if not done so already.
unsafe fn get_task_local_map(task: *rust_task) -> task_local_map {
// Relies on the runtime initialising the pointer to null.
// NOTE: The map's box lives in TLS invisibly referenced once. Each time
// we retrieve it for get/set, we make another reference, which get/set
// drop when they finish. No "re-storing after modifying" is needed.
let map_ptr = rustrt::rust_get_task_local_data(task);
if map_ptr.is_null() {
let map: task_local_map = @dvec::dvec();
// Use reinterpret_cast -- transmute would take map away from us also.
rustrt::rust_set_task_local_data(task, unsafe::reinterpret_cast(map));
rustrt::rust_task_local_data_atexit(task, cleanup_task_local_map);
// Also need to reference it an extra time to keep it for now.
unsafe::bump_box_refcount(map);
map
} else {
let map = unsafe::transmute(map_ptr);
unsafe::bump_box_refcount(map);
map
}
}
unsafe fn key_to_key_value<T: owned>(
key: local_data_key<T>) -> *libc::c_void {
// Keys are closures, which are (fnptr,envptr) pairs. Use fnptr.
// Use reintepret_cast -- transmute would leak (forget) the closure.
let pair: (*libc::c_void, *libc::c_void) = unsafe::reinterpret_cast(key);
pair.first()
}
// If returning some(..), returns with @T with the map's reference. Careful!
unsafe fn local_data_lookup<T: owned>(
map: task_local_map, key: local_data_key<T>)
-> option<(uint, *libc::c_void)> {
let key_value = key_to_key_value(key);
let map_pos = (*map).position(|entry|
alt entry { some((k,_,_)) { k == key_value } none { false } }
);
do map_pos.map |index| {
// .get() is guaranteed because of "none { false }" above.
let (_, data_ptr, _) = (*map)[index].get();
(index, data_ptr)
}
}
unsafe fn local_get_helper<T: owned>(
task: *rust_task, key: local_data_key<T>,
do_pop: bool) -> option<@T> {
let map = get_task_local_map(task);
// Interpret our findings from the map
do local_data_lookup(map, key).map |result| {
// A reference count magically appears on 'data' out of thin air. It
// was referenced in the local_data box, though, not here, so before
// overwriting the local_data_box we need to give an extra reference.
// We must also give an extra reference when not removing.
let (index, data_ptr) = result;
let data: @T = unsafe::transmute(data_ptr);
unsafe::bump_box_refcount(data);
if do_pop {
(*map).set_elt(index, none);
}
data
}
}
unsafe fn local_pop<T: owned>(
task: *rust_task,
key: local_data_key<T>) -> option<@T> {
local_get_helper(task, key, true)
}
unsafe fn local_get<T: owned>(
task: *rust_task,
key: local_data_key<T>) -> option<@T> {
local_get_helper(task, key, false)
}
unsafe fn local_set<T: owned>(
task: *rust_task, key: local_data_key<T>, +data: @T) {
let map = get_task_local_map(task);
// Store key+data as *voids. Data is invisibly referenced once; key isn't.
let keyval = key_to_key_value(key);
// We keep the data in two forms: one as an unsafe pointer, so we can get
// it back by casting; another in an existential box, so the reference we
// own on it can be dropped when the box is destroyed. The unsafe pointer
// does not have a reference associated with it, so it may become invalid
// when the box is destroyed.
let data_ptr = unsafe::reinterpret_cast(data);
let data_box = data as local_data;
// Construct new entry to store in the map.
let new_entry = some((keyval, data_ptr, data_box));
// Find a place to put it.
alt local_data_lookup(map, key) {
some((index, _old_data_ptr)) {
// Key already had a value set, _old_data_ptr, whose reference
// will get dropped when the local_data box is overwritten.
(*map).set_elt(index, new_entry);
}
none {
// Find an empty slot. If not, grow the vector.
alt (*map).position(|x| x == none) {
some(empty_index) {
(*map).set_elt(empty_index, new_entry);
}
none {
(*map).push(new_entry);
}
}
}
}
}
unsafe fn local_modify<T: owned>(
task: *rust_task, key: local_data_key<T>,
modify_fn: fn(option<@T>) -> option<@T>) {
// Could be more efficient by doing the lookup work, but this is easy.
let newdata = modify_fn(local_pop(task, key));
if newdata.is_some() {
local_set(task, key, option::unwrap(newdata));
}
}
/* Exported interface for task-local data (plus local_data_key above). */
/**
* Remove a task-local data value from the table, returning the
* reference that was originally created to insert it.
*/
unsafe fn local_data_pop<T: owned>(
key: local_data_key<T>) -> option<@T> {
local_pop(rustrt::rust_get_task(), key)
}
/**
* Retrieve a task-local data value. It will also be kept alive in the
* table until explicitly removed.
*/
unsafe fn local_data_get<T: owned>(
key: local_data_key<T>) -> option<@T> {
local_get(rustrt::rust_get_task(), key)
}
/**
* Store a value in task-local data. If this key already has a value,
* that value is overwritten (and its destructor is run).
*/
unsafe fn local_data_set<T: owned>(
key: local_data_key<T>, +data: @T) {
local_set(rustrt::rust_get_task(), key, data)
}
/**
* Modify a task-local data value. If the function returns 'none', the
* data is removed (and its reference dropped).
*/
unsafe fn local_data_modify<T: owned>(
key: local_data_key<T>,
modify_fn: fn(option<@T>) -> option<@T>) {
local_modify(rustrt::rust_get_task(), key, modify_fn)
}
extern mod rustrt {
#[rust_stack]
fn rust_task_yield(task: *rust_task) -> bool;
fn rust_get_sched_id() -> sched_id;
fn rust_new_sched(num_threads: libc::uintptr_t) -> sched_id;
fn get_task_id() -> task_id;
#[rust_stack]
fn rust_get_task() -> *rust_task;
fn new_task() -> *rust_task;
fn rust_new_task_in_sched(id: sched_id) -> *rust_task;
fn rust_task_config_notify(
task: *rust_task, &&chan: comm::chan<notification>);
fn start_task(task: *rust_task, closure: *rust_closure);
fn rust_task_is_unwinding(task: *rust_task) -> bool;
fn rust_osmain_sched_id() -> sched_id;
fn rust_task_inhibit_kill(t: *rust_task);
fn rust_task_allow_kill(t: *rust_task);
fn rust_task_inhibit_yield(t: *rust_task);
fn rust_task_allow_yield(t: *rust_task);
fn rust_task_kill_other(task: *rust_task);
fn rust_task_kill_all(task: *rust_task);
#[rust_stack]
fn rust_get_task_local_data(task: *rust_task) -> *libc::c_void;
#[rust_stack]
fn rust_set_task_local_data(task: *rust_task, map: *libc::c_void);
#[rust_stack]
fn rust_task_local_data_atexit(task: *rust_task, cleanup_fn: *u8);
}
#[test]
fn test_spawn_raw_simple() {
let po = comm::port();
let ch = comm::chan(po);
do spawn_raw(default_task_opts()) {
comm::send(ch, ());
}
comm::recv(po);
}
#[test]
#[ignore(cfg(windows))]
fn test_spawn_raw_unsupervise() {
let opts = {
linked: false
with default_task_opts()
};
do spawn_raw(opts) {
fail;
}
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_cant_dup_task_builder() {
let b = task().unlinked();
do b.spawn { }
// FIXME(#2585): For now, this is a -runtime- failure, because we haven't
// got modes on self. When 2585 is fixed, this test should fail to compile
// instead, and should go in tests/compile-fail.
do b.spawn { } // b should have been consumed by the previous call
}
// The following 8 tests test the following 2^3 combinations:
// {un,}linked {un,}supervised failure propagation {up,down}wards.
// !!! These tests are dangerous. If something is buggy, they will hang, !!!
// !!! instead of exiting cleanly. This might wedge the buildbots. !!!
#[test] #[ignore(cfg(windows))]
fn test_spawn_unlinked_unsup_no_fail_down() { // grandchild sends on a port
let po = comm::port();
let ch = comm::chan(po);
do spawn_unlinked {
do spawn_unlinked {
// Give middle task a chance to fail-but-not-kill-us.
for iter::repeat(8192) { task::yield(); }
comm::send(ch, ()); // If killed first, grandparent hangs.
}
fail; // Shouldn't kill either (grand)parent or (grand)child.
}
comm::recv(po);
}
#[test] #[ignore(cfg(windows))]
fn test_spawn_unlinked_unsup_no_fail_up() { // child unlinked fails
do spawn_unlinked { fail; }
}
#[test] #[ignore(cfg(windows))]
fn test_spawn_unlinked_sup_no_fail_up() { // child unlinked fails
do spawn_supervised { fail; }
// Give child a chance to fail-but-not-kill-us.
for iter::repeat(8192) { task::yield(); }
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_unlinked_sup_fail_down() {
do spawn_supervised { loop { task::yield(); } }
fail; // Shouldn't leave a child hanging around.
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_linked_sup_fail_up() { // child fails; parent fails
let po = comm::port::<()>();
let _ch = comm::chan(po);
// Unidirectional "parenting" shouldn't override bidirectional linked.
// We have to cheat with opts - the interface doesn't support them because
// they don't make sense (redundant with task().supervised()).
let b0 = task();
let b1 = task_builder({
opts: { linked: true, parented: true with b0.opts },
can_not_copy: none,
with *b0
});
do b1.spawn { fail; }
comm::recv(po); // We should get punted awake
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_linked_sup_fail_down() { // parent fails; child fails
// We have to cheat with opts - the interface doesn't support them because
// they don't make sense (redundant with task().supervised()).
let b0 = task();
let b1 = task_builder({
opts: { linked: true, parented: true with b0.opts },
can_not_copy: none,
with *b0
});
do b1.spawn { loop { task::yield(); } }
fail; // *both* mechanisms would be wrong if this didn't kill the child...
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_linked_unsup_fail_up() { // child fails; parent fails
let po = comm::port::<()>();
let _ch = comm::chan(po);
// Default options are to spawn linked & unsupervised.
do spawn { fail; }
comm::recv(po); // We should get punted awake
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_linked_unsup_fail_down() { // parent fails; child fails
// Default options are to spawn linked & unsupervised.
do spawn { loop { task::yield(); } }
fail;
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_spawn_linked_unsup_default_opts() { // parent fails; child fails
// Make sure the above test is the same as this one.
do task().linked().spawn { loop { task::yield(); } }
fail;
}
// A bonus linked failure test
#[test] #[should_fail] // #[ignore(cfg(windows))]
#[ignore] // FIXME (#1868) (bblum) make this work
fn test_spawn_unlinked_sup_propagate_grandchild() {
do spawn_supervised {
do spawn_supervised {
loop { task::yield(); }
}
}
for iter::repeat(8192) { task::yield(); }
fail;
}
#[test]
#[ignore(cfg(windows))]
fn test_spawn_raw_notify() {
let task_po = comm::port();
let task_ch = comm::chan(task_po);
let notify_po = comm::port();
let notify_ch = comm::chan(notify_po);
let opts = {
notify_chan: some(notify_ch)
with default_task_opts()
};
do spawn_raw(opts) {
comm::send(task_ch, get_task());
}
let task_ = comm::recv(task_po);
assert comm::recv(notify_po) == exit(task_, success);
let opts = {
linked: false,
notify_chan: some(notify_ch)
with default_task_opts()
};
do spawn_raw(opts) {
comm::send(task_ch, get_task());
fail;
}
let task_ = comm::recv(task_po);
assert comm::recv(notify_po) == exit(task_, failure);
}
#[test]
fn test_run_basic() {
let po = comm::port();
let ch = comm::chan(po);
do task().spawn {
comm::send(ch, ());
}
comm::recv(po);
}
#[test]
fn test_add_wrapper() {
let po = comm::port();
let ch = comm::chan(po);
let b0 = task();
let b1 = do b0.add_wrapper |body| {
fn~() {
body();
comm::send(ch, ());
}
};
do b1.spawn { }
comm::recv(po);
}
#[test]
#[ignore(cfg(windows))]
fn test_future_result() {
let mut result = none;
do task().future_result(|-r| { result = some(r); }).spawn { }
assert future::get(option::unwrap(result)) == success;
result = none;
do task().future_result(|-r| { result = some(r); }).unlinked().spawn {
fail;
}
assert future::get(option::unwrap(result)) == failure;
}
#[test]
fn test_spawn_listiner_bidi() {
let po = comm::port();
let ch = comm::chan(po);
let ch = do spawn_listener |po| {
// Now the child has a port called 'po' to read from and
// an environment-captured channel called 'ch'.
let res = comm::recv(po);
assert res == ~"ping";
comm::send(ch, ~"pong");
};
// Likewise, the parent has both a 'po' and 'ch'
comm::send(ch, ~"ping");
let res = comm::recv(po);
assert res == ~"pong";
}
#[test]
fn test_try_success() {
alt do try {
~"Success!"
} {
result::ok(~"Success!") { }
_ { fail; }
}
}
#[test]
#[ignore(cfg(windows))]
fn test_try_fail() {
alt do try {
fail
} {
result::err(()) { }
result::ok(()) { fail; }
}
}
#[test]
#[should_fail]
#[ignore(cfg(windows))]
fn test_spawn_sched_no_threads() {
do spawn_sched(manual_threads(0u)) { }
}
#[test]
fn test_spawn_sched() {
let po = comm::port();
let ch = comm::chan(po);
fn f(i: int, ch: comm::chan<()>) {
let parent_sched_id = rustrt::rust_get_sched_id();
do spawn_sched(single_threaded) {
let child_sched_id = rustrt::rust_get_sched_id();
assert parent_sched_id != child_sched_id;
if (i == 0) {
comm::send(ch, ());
} else {
f(i - 1, ch);
}
};
}
f(10, ch);
comm::recv(po);
}
#[test]
fn test_spawn_sched_childs_on_same_sched() {
let po = comm::port();
let ch = comm::chan(po);
do spawn_sched(single_threaded) {
let parent_sched_id = rustrt::rust_get_sched_id();
do spawn {
let child_sched_id = rustrt::rust_get_sched_id();
// This should be on the same scheduler
assert parent_sched_id == child_sched_id;
comm::send(ch, ());
};
};
comm::recv(po);
}
#[nolink]
#[cfg(test)]
extern mod testrt {
fn rust_dbg_lock_create() -> *libc::c_void;
fn rust_dbg_lock_destroy(lock: *libc::c_void);
fn rust_dbg_lock_lock(lock: *libc::c_void);
fn rust_dbg_lock_unlock(lock: *libc::c_void);
fn rust_dbg_lock_wait(lock: *libc::c_void);
fn rust_dbg_lock_signal(lock: *libc::c_void);
}
#[test]
fn test_spawn_sched_blocking() {
// Testing that a task in one scheduler can block in foreign code
// without affecting other schedulers
for iter::repeat(20u) {
let start_po = comm::port();
let start_ch = comm::chan(start_po);
let fin_po = comm::port();
let fin_ch = comm::chan(fin_po);
let lock = testrt::rust_dbg_lock_create();
do spawn_sched(single_threaded) {
testrt::rust_dbg_lock_lock(lock);
comm::send(start_ch, ());
// Block the scheduler thread
testrt::rust_dbg_lock_wait(lock);
testrt::rust_dbg_lock_unlock(lock);
comm::send(fin_ch, ());
};
// Wait until the other task has its lock
comm::recv(start_po);
fn pingpong(po: comm::port<int>, ch: comm::chan<int>) {
let mut val = 20;
while val > 0 {
val = comm::recv(po);
comm::send(ch, val - 1);
}
}
let setup_po = comm::port();
let setup_ch = comm::chan(setup_po);
let parent_po = comm::port();
let parent_ch = comm::chan(parent_po);
do spawn {
let child_po = comm::port();
comm::send(setup_ch, comm::chan(child_po));
pingpong(child_po, parent_ch);
};
let child_ch = comm::recv(setup_po);
comm::send(child_ch, 20);
pingpong(parent_po, child_ch);
testrt::rust_dbg_lock_lock(lock);
testrt::rust_dbg_lock_signal(lock);
testrt::rust_dbg_lock_unlock(lock);
comm::recv(fin_po);
testrt::rust_dbg_lock_destroy(lock);
}
}
#[cfg(test)]
fn avoid_copying_the_body(spawnfn: fn(+fn~())) {
let p = comm::port::<uint>();
let ch = comm::chan(p);
let x = ~1;
let x_in_parent = ptr::addr_of(*x) as uint;
do spawnfn {
let x_in_child = ptr::addr_of(*x) as uint;
comm::send(ch, x_in_child);
}
let x_in_child = comm::recv(p);
assert x_in_parent == x_in_child;
}
#[test]
fn test_avoid_copying_the_body_spawn() {
avoid_copying_the_body(spawn);
}
#[test]
fn test_avoid_copying_the_body_spawn_listener() {
do avoid_copying_the_body |f| {
spawn_listener(fn~(move f, _po: comm::port<int>) {
f();
});
}
}
#[test]
fn test_avoid_copying_the_body_task_spawn() {
do avoid_copying_the_body |f| {
do task().spawn {
f();
}
}
}
#[test]
fn test_avoid_copying_the_body_spawn_listener() {
do avoid_copying_the_body |f| {
task().spawn_listener(fn~(move f, _po: comm::port<int>) {
f();
});
}
}
#[test]
fn test_avoid_copying_the_body_try() {
do avoid_copying_the_body |f| {
do try {
f()
};
}
}
#[test]
fn test_avoid_copying_the_body_unlinked() {
do avoid_copying_the_body |f| {
do spawn_unlinked {
f();
}
}
}
#[test]
fn test_osmain() {
let po = comm::port();
let ch = comm::chan(po);
do task().sched_mode(osmain).spawn {
comm::send(ch, ());
}
comm::recv(po);
}
#[test]
#[ignore(cfg(windows))]
#[should_fail]
fn test_unkillable() {
import comm::methods;
let po = comm::port();
let ch = po.chan();
// We want to do this after failing
do spawn_raw({ linked: false with default_task_opts() }) {
for iter::repeat(10u) { yield() }
ch.send(());
}
do spawn {
yield();
// We want to fail after the unkillable task
// blocks on recv
fail;
}
unsafe {
do unkillable {
let p = ~0;
let pp: *uint = unsafe::transmute(p);
// If we are killed here then the box will leak
po.recv();
let _p: ~int = unsafe::transmute(pp);
}
}
// Now we can be killed
po.recv();
}
#[test]
#[ignore(cfg(windows))]
#[should_fail]
fn test_unkillable_nested() {
import comm::methods;
let po = comm::port();
let ch = po.chan();
// We want to do this after failing
do spawn_raw({ linked: false with default_task_opts() }) {
for iter::repeat(10u) { yield() }
ch.send(());
}
do spawn {
yield();
// We want to fail after the unkillable task
// blocks on recv
fail;
}
unsafe {
do unkillable {
do unkillable {} // Here's the difference from the previous test.
let p = ~0;
let pp: *uint = unsafe::transmute(p);
// If we are killed here then the box will leak
po.recv();
let _p: ~int = unsafe::transmute(pp);
}
}
// Now we can be killed
po.recv();
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_atomically() {
unsafe { do atomically { yield(); } }
}
#[test]
fn test_atomically2() {
unsafe { do atomically { } } yield(); // shouldn't fail
}
#[test] #[should_fail] #[ignore(cfg(windows))]
fn test_atomically_nested() {
unsafe { do atomically { do atomically { } yield(); } }
}
#[test]
fn test_child_doesnt_ref_parent() {
// If the child refcounts the parent task, this will stack overflow when
// climbing the task tree to dereference each ancestor. (See #1789)
const generations: uint = 8192;
fn child_no(x: uint) -> fn~() {
ret || {
if x < generations {
task::spawn(child_no(x+1));
}
}
}
task::spawn(child_no(0));
}
#[test]
fn test_tls_multitask() unsafe {
fn my_key(+_x: @~str) { }
local_data_set(my_key, @~"parent data");
do task::spawn {
assert local_data_get(my_key) == none; // TLS shouldn't carry over.
local_data_set(my_key, @~"child data");
assert *(local_data_get(my_key).get()) == ~"child data";
// should be cleaned up for us
}
// Must work multiple times
assert *(local_data_get(my_key).get()) == ~"parent data";
assert *(local_data_get(my_key).get()) == ~"parent data";
assert *(local_data_get(my_key).get()) == ~"parent data";
}
#[test]
fn test_tls_overwrite() unsafe {
fn my_key(+_x: @~str) { }
local_data_set(my_key, @~"first data");
local_data_set(my_key, @~"next data"); // Shouldn't leak.
assert *(local_data_get(my_key).get()) == ~"next data";
}
#[test]
fn test_tls_pop() unsafe {
fn my_key(+_x: @~str) { }
local_data_set(my_key, @~"weasel");
assert *(local_data_pop(my_key).get()) == ~"weasel";
// Pop must remove the data from the map.
assert local_data_pop(my_key) == none;
}
#[test]
fn test_tls_modify() unsafe {
fn my_key(+_x: @~str) { }
local_data_modify(my_key, |data| {
alt data {
some(@val) { fail ~"unwelcome value: " + val }
none { some(@~"first data") }
}
});
local_data_modify(my_key, |data| {
alt data {
some(@~"first data") { some(@~"next data") }
some(@val) { fail ~"wrong value: " + val }
none { fail ~"missing value" }
}
});
assert *(local_data_pop(my_key).get()) == ~"next data";
}
#[test]
fn test_tls_crust_automorestack_memorial_bug() unsafe {
// This might result in a stack-canary clobber if the runtime fails to set
// sp_limit to 0 when calling the cleanup extern - it might automatically
// jump over to the rust stack, which causes next_c_sp to get recorded as
// something within a rust stack segment. Then a subsequent upcall (esp.
// for logging, think vsnprintf) would run on a stack smaller than 1 MB.
fn my_key(+_x: @~str) { }
do task::spawn {
unsafe { local_data_set(my_key, @~"hax"); }
}
}
#[test]
fn test_tls_multiple_types() unsafe {
fn str_key(+_x: @~str) { }
fn box_key(+_x: @@()) { }
fn int_key(+_x: @int) { }
do task::spawn {
local_data_set(str_key, @~"string data");
local_data_set(box_key, @@());
local_data_set(int_key, @42);
}
}
#[test]
fn test_tls_overwrite_multiple_types() unsafe {
fn str_key(+_x: @~str) { }
fn box_key(+_x: @@()) { }
fn int_key(+_x: @int) { }
do task::spawn {
local_data_set(str_key, @~"string data");
local_data_set(int_key, @42);
// This could cause a segfault if overwriting-destruction is done with
// the crazy polymorphic transmute rather than the provided finaliser.
local_data_set(int_key, @31337);
}
}
#[test]
#[should_fail]
#[ignore(cfg(windows))]
fn test_tls_cleanup_on_failure() unsafe {
fn str_key(+_x: @~str) { }
fn box_key(+_x: @@()) { }
fn int_key(+_x: @int) { }
local_data_set(str_key, @~"parent data");
local_data_set(box_key, @@());
do task::spawn { // spawn_linked
local_data_set(str_key, @~"string data");
local_data_set(box_key, @@());
local_data_set(int_key, @42);
fail;
}
// Not quite nondeterministic.
local_data_set(int_key, @31337);
fail;
}