Remove linked failure from the runtime

The reasons for doing this are:

* The model on which linked failure is based is inherently complex
* The implementation is also very complex, and there are few remaining who
  fully understand the implementation
* There are existing race conditions in the core context switching function of
  the scheduler, and possibly others.
* It's unclear whether this model of linked failure maps well to a 1:1 threading
  model

Linked failure is often a desired aspect of tasks, but we would like to take a
much more conservative approach in re-implementing linked failure if at all.

Closes rust-lang#8674
Closes rust-lang#8318
Closes rust-lang#8863

doc/rust.md

@@ -2350,9 +2350,9 @@ Indices are zero-based, and may be of any integral type. Vector access
 is bounds-checked at run-time. When the check fails, it will put the
 task in a _failing state_.
 
-~~~~
+~~~~ {.xfail-test}
 # use std::task;
-# do task::spawn_unlinked {
+# do task::spawn {
 
 ([1, 2, 3, 4])[0];
 (["a", "b"])[10]; // fails

doc/tutorial-tasks.md

@@ -402,22 +402,6 @@ freeing memory along the way---and then exits. Unlike exceptions in C++,
 exceptions in Rust are unrecoverable within a single task: once a task fails,
 there is no way to "catch" the exception.
 
-All tasks are, by default, _linked_ to each other. That means that the fates
-of all tasks are intertwined: if one fails, so do all the others.
-
-~~~{.xfail-test .linked-failure}
-# use std::task::spawn;
-# use std::task;
-# fn do_some_work() { loop { task::yield() } }
-# do task::try {
-// Create a child task that fails
-do spawn { fail!() }
-
-// This will also fail because the task we spawned failed
-do_some_work();
-# };
-~~~
-
 While it isn't possible for a task to recover from failure, tasks may notify
 each other of failure. The simplest way of handling task failure is with the
 `try` function, which is similar to `spawn`, but immediately blocks waiting
@@ -464,101 +448,7 @@ it trips, indicates an unrecoverable logic error); in other cases you
 might want to contain the failure at a certain boundary (perhaps a
 small piece of input from the outside world, which you happen to be
 processing in parallel, is malformed and its processing task can't
-proceed). Hence, you will need different _linked failure modes_.
-
-## Failure modes
-
-By default, task failure is _bidirectionally linked_, which means that if
-either task fails, it kills the other one.
-
-~~~{.xfail-test .linked-failure}
-# use std::task;
-# use std::comm::oneshot;
-# fn sleep_forever() { loop { let (p, c) = oneshot::<()>(); p.recv(); } }
-# do task::try {
-do spawn {
-    do spawn {
-        fail!();  // All three tasks will fail.
-    }
-    sleep_forever();  // Will get woken up by force, then fail
-}
-sleep_forever();  // Will get woken up by force, then fail
-# };
-~~~
-
-If you want parent tasks to be able to kill their children, but do not want a
-parent to fail automatically if one of its child task fails, you can call
-`task::spawn_supervised` for _unidirectionally linked_ failure. The
-function `task::try`, which we saw previously, uses `spawn_supervised`
-internally, with additional logic to wait for the child task to finish
-before returning. Hence:
-
-~~~{.xfail-test .linked-failure}
-# use std::comm::{stream, Chan, Port};
-# use std::comm::oneshot;
-# use std::task::{spawn, try};
-# use std::task;
-# fn sleep_forever() { loop { let (p, c) = oneshot::<()>(); p.recv(); } }
-# do task::try {
-let (receiver, sender): (Port<int>, Chan<int>) = stream();
-do spawn {  // Bidirectionally linked
-    // Wait for the supervised child task to exist.
-    let message = receiver.recv();
-    // Kill both it and the parent task.
-    assert!(message != 42);
-}
-do try {  // Unidirectionally linked
-    sender.send(42);
-    sleep_forever();  // Will get woken up by force
-}
-// Flow never reaches here -- parent task was killed too.
-# };
-~~~
-
-Supervised failure is useful in any situation where one task manages
-multiple fallible child tasks, and the parent task can recover
-if any child fails. On the other hand, if the _parent_ (supervisor) fails,
-then there is nothing the children can do to recover, so they should
-also fail.
-
-Supervised task failure propagates across multiple generations even if
-an intermediate generation has already exited:
-
-~~~{.xfail-test .linked-failure}
-# use std::task;
-# use std::comm::oneshot;
-# fn sleep_forever() { loop { let (p, c) = oneshot::<()>(); p.recv(); } }
-# fn wait_for_a_while() { for _ in range(0, 1000u) { task::yield() } }
-# do task::try::<int> {
-do task::spawn_supervised {
-    do task::spawn_supervised {
-        sleep_forever();  // Will get woken up by force, then fail
-    }
-    // Intermediate task immediately exits
-}
-wait_for_a_while();
-fail!();  // Will kill grandchild even if child has already exited
-# };
-~~~
-
-Finally, tasks can be configured to not propagate failure to each
-other at all, using `task::spawn_unlinked` for _isolated failure_.
-
-~~~{.xfail-test .linked-failure}
-# use std::task;
-# fn random() -> uint { 100 }
-# fn sleep_for(i: uint) { for _ in range(0, i) { task::yield() } }
-# do task::try::<()> {
-let (time1, time2) = (random(), random());
-do task::spawn_unlinked {
-    sleep_for(time2);  // Won't get forced awake
-    fail!();
-}
-sleep_for(time1);  // Won't get forced awake
-fail!();
-// It will take MAX(time1,time2) for the program to finish.
-# };
-~~~
+proceed).
 
 ## Creating a task with a bi-directional communication path
 

src/libextra/arc.rs

@@ -655,7 +655,7 @@ mod tests {
         let arc2 = ~arc.clone();
         let (p, c) = comm::stream();
 
-        do task::spawn_unlinked || {
+        do spawn {
             let _ = p.recv();
             do arc2.access_cond |one, cond| {
                 cond.signal();

src/libextra/comm.rs

@@ -137,7 +137,6 @@ pub fn rendezvous<T: Send>() -> (SyncPort<T>, SyncChan<T>) {
 mod test {
     use comm::{DuplexStream, rendezvous};
     use std::rt::test::run_in_uv_task;
-    use std::task::spawn_unlinked;
 
 
     #[test]
@@ -177,7 +176,7 @@ mod test {
     #[test]
     fn send_and_fail_and_try_recv() {
         let (port, chan) = rendezvous();
-        do spawn_unlinked {
+        do spawn {
             chan.duplex_stream.send(()); // Can't access this field outside this module
             fail!()
         }
@@ -187,7 +186,7 @@ mod test {
     #[test]
     fn try_send_and_recv_then_fail_before_ack() {
         let (port, chan) = rendezvous();
-        do spawn_unlinked {
+        do spawn {
             port.duplex_stream.recv();
             fail!()
         }
@@ -198,7 +197,7 @@ mod test {
     #[should_fail]
     fn send_and_recv_then_fail_before_ack() {
         let (port, chan) = rendezvous();
-        do spawn_unlinked {
+        do spawn {
             port.duplex_stream.recv();
             fail!()
         }

src/libextra/future.rs

@@ -27,7 +27,6 @@
 
 use std::cell::Cell;
 use std::comm::{PortOne, oneshot};
-use std::task;
 use std::util::replace;
 
 /// A type encapsulating the result of a computation which may not be complete
@@ -130,29 +129,12 @@ impl<A:Send> Future<A> {
 
         let (port, chan) = oneshot();
 
-        do task::spawn_with(chan) |chan| {
+        do spawn {
             chan.send(blk());
         }
 
         Future::from_port(port)
     }
-
-    pub fn spawn_with<B: Send>(v: B, blk: proc(B) -> A) -> Future<A> {
-        /*!
-         * Create a future from a unique closure taking one argument.
-         *
-         * The closure and its argument will be moved into a new task. The
-         * closure will be run and its result used as the value of the future.
-         */
-
-         let (port, chan) = oneshot();
-
-         do task::spawn_with((v, chan)) |(v, chan)| {
-            chan.send(blk(v));
-         }
-
-         Future::from_port(port)
-    }
 }
 
 #[cfg(test)]
@@ -207,12 +189,6 @@ mod test {
         assert_eq!(f.get(), ~"bale");
     }
 
-    #[test]
-    fn test_spawn_with() {
-        let mut f = Future::spawn_with(~"gale", |s| { s });
-        assert_eq!(f.get(), ~"gale");
-    }
-
     #[test]
     #[should_fail]
     fn test_futurefail() {