which raise before first yielding
as can happen if a non-async function is intended to return an...

does not work yet

Note that the imported tests (in trio/tests/aiotest and …/python)
have not been touched: future ...

This boils down to skipping tests for bugs fixed in 3.6

This reverts commit 10e77e2f527811fc77f104b74e25cee53b1dea8d.

The read/write callback may directly close a file descriptor and then
enqueue a callback to term...

insist on using Trio's native methods
no "foreign" worker pool implementation allowed

They hang.

Use a logger for various async results/errors.
They should all be caught somewhere, but we need ...

We can simply cancel / restart the i/o handlers in-place.

to ensure ordering

Oops :-/

The loop's absolute pseudo-time does not pass while it's suspended.

it's benign, and used in testing

Our .cancel() was never called because it was behind
asyncio.Handle.cancel() in the MRO.

asyncio's run_forever() does a lot of things.
Rather than replace it, override the inner single-...

otherwise debug output will be too cryptic
and lack essentioal information

This involves instrumenting _remove_reader/writer to drop them from
._saved_fds when idle.

These are still somewhat untested.

for Trio to wait for an asyncio future
and call an asyncio procedure/method

for doing the same with TimerHandle later

depending on the loop's idle state

a common coding style is a good thing

This change allows common patterns like
t = loop.time()
loop.call_at(t+10, foo)
loop.run_f...

Source: https://bitbucket.org/haypo/aiotest

asyncio semantics says that delayed tasks need to be kept waiting while the
loop is interrupted....

This allows the user to cancel a callback

This enables the user to cancel the task (if async)
after it's started

with support for queuing calls before the loop starts

This allows the loop._ready "array" to accept new handles
which gives us a working ``loop.call_s...

Asyncio compatibility is implemented by heavily hacking a "standard"
Unix asyncio mainloop.

The...