The signal-hook library got split into lower-level and higher-level
parts. The tokio-signal uses only API from the lower-level one, so it
can depend on it directly.
The only effect of this change is smaller amount of compiled (and
unused) code during compilation. There's no change in the code actually
used.
* Don't use tokio-core any more for tests. That one brings tokio from
crates.io instead of the current workspace and two versions of that
don't want to cooperate.
* Guard unix-specific examples on windows.
* Leave CI setup to top-level directory.
Replace the sequential counting (which might be exhausted) by an address
of an object (in a box, so it doesn't change). This is also a unique, so
it is acceptable ID.
Run multiple loops (both in parallel and sequentially) to make sure
broadcasting to multiple of them works and we work even after the
initial loop has gone away.
Add the driver task, connecting the signal handler wakeups to the
wakeups of of the streams.
It is a prototype-quality code, a lot of cleanups and similar is needed.
Which is just a wrapper around the futures::sync::mpsc. The sender is in
a global registry.
The part that connects the wakeups to the senders in the registry
doesn't yet exist.
Register the signal handler that wakes up someone through a self-pipe.
That someone doesn't yet exist, though.
Some dependencies (nix, lazy_static) added to speed up the prototyping
process. They are likely to be dropped in some future commits.
Some features (eg. preserving the previous signal handlers) are still
missing.
Michal 'vorner' Vaner