Before:
`InputBufferStream.close()` does not close their buffer so the buffer will leak.
After:
Resolves#4198.
Override the `close` for closing their buffer.
gRPC's protobuf-lite auto-selects between full and lite protobuf based on the
value of crosstool_top. If the user is specifying their own
--android_crosstool_top, then it will not auto-detect correctly. One day,
platforms will fix problems like this, but for the moment it seems we get to
live with it.
The peer socket is read from TRANSPORT_ATTR_REMOTE_ADDR from the
stream attributes. We only log the peer on receive initial metadata.
The call id assumes census is available. The call ID read from the
context via SERVER_CALL_ID_CONTEXT_KEY on server side, and read from
CallOptions via CLIENT_CALL_ID_CALLOPTION_KEY on client side. The
value is copied from CONTEXT_SPAN_KEY which is set by census.
Pass around CallId with two longs, not a byte[].
Server listen sockets differ from normal sockets in that they do not
have a remote address, do not have stats on calls started/failed/etc,
and do not have security info.
They were removed in 137c74d1 since it was believed they were unnecessary.
However, since they are in a macro and not a rule, they are relative to the
caller, not their definition.
Added building the examples to the kokoro CI. Note that this means the examples
are built twice: once in grpc-java's build and once in their own (because it
has a WORKSPACE). Given that the Bazel build is our fastest build, this
slowdown won't probably be an issue.
OkHttpClientInteropServlet and NettyClientInteropServlet both run the
@After method from AbstractInteropTest. Let's make sure we await
termination.
For the long lived channel test, do the cleanup in `destroy`.
If the release branch (i.e., v1.11.x) was not up-to-date with the local
repo, then the push to update the branch will fail. Doing it first
allows you to notice, merge in the updates and fix the tag before
pushing it.
Also document pulling the most up-to-date version of the release branch
before creating the 'release' branch. That would reduce the amount of
time the release branch could be modified.
This fixes a serial warning that breaks the build when building
_without_ ErrorProne. I have no clue why _disabling_ ErrorProne makes
the warning pop up, but whatever, the warning is legit.
I also gave it a name, since the anonymous $1 seems weird if we're
providing a serialVersionUID. Not to say we're actually supporting
serializing this object. But if we ever see a serialized form show up in
the future, having a more clear name would make it obvious what it is.
This allows ProGuard to remove OkHttp's ConnectionSpec in most cases,
saving about 40 methods. The savings won't be realized until
DEFAULT_CONNECTION_SPEC is removed.
Always set the remote address, no reason why this should be a TLS-only
feature. This is needed for channelz, and is especially useful in unit
tests where we are using plaintext.
This PR adds the attr for plaintext.
- To make unit tests easier, added the BinaryLogSinkWriter abstract
class, which allows verifying high level arguments rather than low
level protobufs.
- remove InputStream based marshaller (IDENTITY_MARSHALLER) in favor
of byte[] because retries will need to call serialize multiple
times. This is achieved with a special InputStream that lets us
access the internal byte[] without copying.
- Change ByteBuffer based internal methods to use byte[]
- Update unit tests to reflect the above
We always build the artifacts, but only store them in placer via
kokoro if MVN_ARTIFACTS is set to a non empty value.
The 32 bit script runs tests as well as builds artifacts.
The 64 bit script only builds artifacts.
This is to conform with the GRPCLB spec. The spec doesn't (yet) define
the actual timeout. We choose 10 seconds here arbitrarily. It may be
configurable in the future.
This will fix internal bug b/74410243
Eric Anderson
zpencer
Carl Mastrangelo
Jorg Heymans
Shohei Kamimori
David Brodsky
ZHANG Dapeng
Arnout Engelen
Eric Gribkoff
Amit Desai
Ray Tsang
Jan Tattermusch
Kun Zhang