Panic when "forgetting" ReadTask.

[rust-lang]

Panic when "forgetting" ReadTask.

[alexcrichton]

Thanks for the report! The cause here is that a TaskData will panic if accessed with the wrong task. For each connection in this example, there's three tasks in play. One is the task which is driving the acceptance of new sockets. Another is the task for each socket with pair.map(...).forget(). The final one is the read_exact(..).forget().

The TaskIo structure binds the I/O to one particular task, and it will panic if used with another task. Here it's bound to the second task, pair.map(..).forget(), but then when used with read_exact(..).forget() it panics.

The fix here is to change the final map to an and_then as well as removing the .forget() on the read_exact. Does that make sense?

[Antti]

@alexcrichton I managed to make it work (in a similar way to what you suggested), by following the "echo.rs" example more closely.
Is it possible to encode TaskData relations on the type level, so that the given example won't compile, rather then panicking on the runtime?

[alexcrichton]

Unfortunately we don't have a great way of doing that that doesn't hinder the ergonomics to the point of making it unusable :(

[alexcrichton]

For now I'm going to close this as intended behavior, but let me know if you need any help!