This reverts commit 279b0c1f6d473d65ed4a449e6fb308ff18ac7c9b.

Conflicts:
	lib/testing/src/async/fake_time.dart
	test/testing/async/fake_time_test.dart

initialTime is only relevant to the Clock, which is not always needed,
so it should not be a constructor parameter.  This allows initialTime to
be a required argument, and thus not be named, and not having to default
it to `new DateTime.now()` which introduces indeterminism to tests.

Add FakeAsync test utility

Add StreamBuffer

…able

Add GeneratingIterable

…ystem-clock

Remove SYSTEM_CLOCK. Use const Clock() instead.

Update .gitignore

…ments

Change signature of classImplements()

…nittest are not scheduled in FakeAsync's custom zone.

…est-fix

FakeAsync: wrap Futures returned to unittest

Remove detailed Metronome examples from README.md

flushMicrotasks and flushTimers

…cleanups

Reformat with dartformat

fixes dart-lang/fake_async#248

Add stats reporting to FakeAsync

No more need for dummy constructors

Fixes dart-lang/fake_async#277

Both the DartVM and dartpad.dartlang.org mark a timer as not-active
while the callback is executing. This can be used in production code
to test some condition - but fails when using quiver-dart/async.dart.

Non-periodic Timers are not active while executing callback()

Dart SDK commit 6255638cd0623b3aa41596b98f4584876f6b8822 introduced
changes to stream subscription cancellation. cancelOnError was
irrelevant to the periodic and timeout tests.

…async#343)

* fix strong mode errors, turn on strong mode analysis
* Push better fix for _id inference

When the function to `runGuarded` throws synchronously, then the uncaught-handler is invoked. Except for very rare cases, that handler can only return `null`. As such, it doesn't make sense to use the result of `runGuarded`. Future versions of the Zone API will thus make `runGuarded` a `void` function.

* rename analysis_options

* replace deprecated throws matcher

* remove unnecessary cast

Fixes dart-lang/fake_async#379

As noted in google/quiver-dart#258, the
current ordering of _getNextTimer() in _FakeAsync is inconsistent with
the native Dart event loop.

In the process of fixing that, I had to fix a few other idiosyncrasies
that were revealed from the existing unit tests which relied on the
incorrect behavior. For instance, if you had a periodic timer occur at
the same time as a delayed timer, whether or not both of them were
"flushed" in flushTimers depended on the order with which they were
queued. Now it does not.

…-lang/fake_async#406)

This brings in a filter-branch view of the history of FakeAsync within
the quiver repository.

This matches the old fake_async package API and cleans up test code
considerably.

Clean up the quiver code

…ync#3)

This brings this package to full API compatibility with fake_async
0.1.

Works around dart-lang/sdkdart-lang/fake_async#33015

Fix uppercase constants and deprecated matcher.

…ke_async#14)

- prefer_collection_literals

Bump minimum SDK to 2.2.0 to allow for Set literals.

- Exposed the `FakeTimer` class as a public class.
  - Added `debugString` getter as well as making duration, creationStackTrace, and isPeriodic public.
- Added `FakeAsync.pendingTimers` which gives access to all pending timers at the time of the call.

The `1.1.0-nullsafety` version is considered _earlier_ than `1.1.0`
which is already published. Use `1.2.0-nullsafety` instead.

Bump the upper bound to allow 2.10 stable and 2.11.0 dev SDK versions.

Use a 2.12.0 lower bound since pub does not understand allowed
experiments for earlier versions.

Use a 3.0.0 upper bound to avoid a warning in pub and to give some
flexibility in publishing for stable.

In a real timer the tick count can go up by more than the number of
times the callback was invoked if a period is missed. In the FakeAsync
case the `tick` matches the number of times the callback was invoked.

…ang/fake_async#40)

I missed a long string in a recent PR. Enable the lint to avoid relying
on human reviewers.

prep to publish 1.3.1