… the master branch

…java into davidtorres-master

* merge from master

* make pubsub high perf compile

* make AckDeadlineRenewerTest pass

* use fewer guava classes in public surface

- SubscriberImpl used to expose AbstractService.
  We just make this package-private.
- SubscriberStats defines its own Stats type.

* make AutoValue work in PubSub
Use error_prone_annotations so we don't clobber classpath with Java8 classes.
Migrate SubscriberStats and PublisherStats to AutoValue

…ss to use containment in favor of inheritance

* Update

* Fixing a race condition on shutdown

Some of our tests seem to still be flaking with timeout,
eg https://travis-ci.org/GoogleCloudPlatform/google-cloud-java/jobs/185655553

Making the test not quiet should give us an idea of what's timing out.

* delete pull-related methods from PubSub

Instead, provide a way to create the Subscriber object.

* also delete modifyAckDeadline

* google-java-format changed lines

This commit tries to deflake failures reported in https://ci.appveyor.com/project/garrettjonesgoogle/google-cloud-java/build/57 .

The failure seems to come from `FakeSubscriberServiceImpl::sendError`. When calling `getAndAdvanceCurrentStream`, we waited for there to be a subscription by waiting on the `subscriptionInitialized` variable.

The variable is notified before `addOpenedStream` is called though, and so it's possible that `getAndAdvanceCurrentStream` will see a list without any opened stream. The fix is to wait for at least one stream to be registered first.

* delete publish-related methods from PubSub

* make pubsub work with new gax

update documentation links too

* Delete the handwritten Message type

* delete handwritten PushConfig

* update code samples

* Delete handwritten PubSub layer

* make Publisher/Subscriber builders public

* also delete PubSubRpc

* fix doc annotations

* clean up Publisher and Subscriber surface

- PUBSUB_API_ADDRESS and PUBSUB_API_SCOPE are removed.
  We now use {Publisher,Subscriber}Settings instead.
- Default settings are moved into the Builder class,
  so that they have package-private visibility.
  We can make them public later if we want to.
- Subscriber doc is fixed so that it is not chopped off mid-sentence.
- Subscriber doc mistakenly referred to the Publisher. This is now
  fixed.

* Update version to 0.8.1-SNAPSHOT (#1467)

Also, update versions in README to 0.8.0

* Add link to Maven Central for maven-central badge. (#1468)

Used to link to the image, which wasn't super useful.

* fix more races in pubsub tests

Previously BlockingProcessStreamReader has a terminate() method,
used to tell the Reader to stop reading from the emulator process.

This causes an inter-process race.
If the Reader stops before reading emulator's output,
the emulator process will hang as it tries to write to stdout/stderr
as there's no one to read from the other side of the pipe.

Since there is no way to safely stop the Reader,
this commit deletes the method and its associated test.

Additionally, the timeout for LocalSystemTest is increased to 3 minutes,
since the emulator, somehow, consistently takes just longer than a
minute to shut down.

* Regenerating SPI layer (#1501)

* Converting Error Reporting and Monitoring to use resource name types
* Removing formatX/parseX methods from pubsub, converting usage of the
  same to resource name types
* New methods in Logging and PubSub

* Updating grpc dependency to 1.0.3 (#1504)

* Release 0.8.1 (#1512)

* Fix code snippet (wrong method name) in README.md

Original code snippet in _"Querying data"_ section: 

`..while (!queryResponse.jobComplete()) {..`

This results in a compile error: 

_"Cannot resolve method jobComplete()"_

The correct method is `jobCompleted()`

* Updating version in README files. [ci skip]

* Update version to 0.8.2-alpha-SNAPSHOT

* Allow path in URIs passed to newFileSystem (#1470)

* This makes it easier for users who start with a URI describing a full
path to get a FileSystem that can work with that path, since they no
longer have to needlessly remove the path from the URI.

Note that Oracle's description of newFileSystem [1] puts no restriction
on the passed URI.

[1]
https://docs.oracle.com/javase/7/docs/api/java/nio/file/FileSystems.html#newFileSystem(java.net.URI,%20java.util.Map)

* Preventing logging re-entrance at FINE level (#1523)

* Preventing logging re-entrance at FINE level

Also:
* Reducing the scope of synchronized blocks
* Removing logger exclusions except for Http2FrameLogger

* Add a PathMatcher for CloudStorageFileSystem (#1469)

Add a test, as well.
We reuse the default PathMatcher since it does a fine job of globbing files.

* Set timestamp from LogRecord (#1533)

* Initialize the default MonitoredResource from a GAE environment (#1535)

* BigQuery: Add support to FormatOptions for AVRO

#1441

Added new constant in FormatOptions, and a corresponding factory method. Updated test cases. Confirmed that AVRO does not require special treatment (like CSV does), so no additional changes are required.

* Reverting changed commited by mistake before review.

* BigQuery: Add support to FormatOptions for AVRO

#1441

Added new constant in FormatOptions and a corresponding factory method. Updated test cases. Confirmed that AVRO does not require special treatment (like CSV does), so no additional changes are required.

* use RpcFuture and remove old BundlingSettings (#1572)

* use RpcFuture and remove old BundlingSettings

* Release 0.8.2

Note: This version was accidentally released to Sonatype because of
experimenting with deployment commands, combined with the fact that
autoReleaseAfterClose is set to true. Since releases can't be taken
back, we might as well own up to the release and push the code
forward.

* Updating version in README files. [ci skip]

* Fixing javadoc error in GaeFlexLoggingEnhancer (#1582)

* get tests to compile and pass

This is the same race as
#1495
but in sendStreamResponse rather than sendError.