Why is it safe for old rust_port.cpp to store a pointer to its owning task?

[bblum]

rust_kernel has a facility for looking up a port by id and returning it refcounted so the port doesn't go away. But what stops the task from going away, making port->task invalid to somebody attempting to send on a channel associated with that port?

[bblum]

Likely related to #1923.

[brson]

Destroying a port involves 'detaching' it, after which there are supposed to be no references to it and it's not accessible via the kernel's port table, and a task can't die without destroying its ports.

There's an assert in rust_port::end_detach that checks that the port is not referenced. I have seen that assert fail in the past, though not recently.

[bblum]

@eholk should close this when #2851 closes.

[bblum]

Ah, I see now. It looks good to me.