Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
The Machine handler should not really do anything after it has decided
it wants to be destroyed, so it seems better to enforce this by having
the handler return this action.
|
|
This provides two advantages: Machines don't have to be Send/Sync
anymore (but can still be spawned on different threads), and the init()
and new() methods are now merged, allowing the Machine to access its
context on initialization.
I guess Rotor had a point here after all. :)
|
|
TODO: Add a global io-timeout that is inherited by interfaces that don't
set it? Or is that needless complexity?
I've added Sync bounds to Machines and implemented all this using
Arc/Mutex in order for Machines to be runnable on multiple threads later
on.
|
|
|
|
|
|
|