openaphid ( Bo )

        boolean updated = rdq.setUnackTimeout(id, 500);
        assertTrue(updated);
        popped = rdq.pop(1, 1, TimeUnit.SECONDS);  // <-- RedisPipelineQueue.processUnacks() might be scheduled while waiting in pop operation
        assertNotNull(popped);
        assertEquals(0, popped.size());

RedisPipelineQueue.processUnacks() is triggered periodically via a scheduled thread pool. If it happens to run while waiting for pop to finish, pop may return some results and the test fails.

The full console output is as follows:

[INFO ] 2020-08-22 14:28:16,790 RedisPipelineQueue - com.netflix.dyno.queues.redis.v2.RedisPipelineQueue is ready to serve {test_queue.x}, shard=x
[DEBUG] 2020-08-22 14:28:18,801 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,810 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,825 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,846 RedisPipelineQueue - Adding 2 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,890 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,898 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,913 RedisPipelineQueue - Adding 2 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,947 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:19,237 RedisPipelineQueue - Cannot get the message payload some fake id
Message [id=0, payload=Hello World-0, timeout=0, priority=10]
Message [id=1, payload=Hello World-1, timeout=0, priority=9]
Message [id=2, payload=Hello World-2, timeout=0, priority=8]
Message [id=3, payload=Hello World-3, timeout=0, priority=7]
Message [id=4, payload=Hello World-4, timeout=0, priority=6]
Message [id=5, payload=Hello World-5, timeout=0, priority=5]
Message [id=6, payload=Hello World-6, timeout=0, priority=4]
Message [id=7, payload=Hello World-7, timeout=0, priority=3]
Message [id=8, payload=Hello World-8, timeout=0, priority=2]
Message [id=9, payload=Hello World-9, timeout=0, priority=1]
[DEBUG] 2020-08-22 14:28:23,895 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue

expected:<0> but was:<1>
Expected :0
Actual   :1
<Click to see difference>

java.lang.AssertionError: expected:<0> but was:<1>
	at org.junit.Assert.fail(Assert.java:88)
	at org.junit.Assert.failNotEquals(Assert.java:743)
        at com.netflix.dyno.queues.redis.v2.RedisDynoQueueTest.testTimeoutUpdate(RedisDynoQueueTest.java:119)

closed time in 2 months

openaphid

push eventopenaphid/armeria

Armin Hüneburg

commit sha b430f26130d1d08d85d10d6915daba7a89699157

Prettify output of REST calls in DocService if JSON (#2635)

view details

Amos

commit sha 0a7dde3ecf48a31eab986d09c27b73c6c3d1c53e

Fix incorrect Logback settings in documentation (#2658) Motivation: - From Armeria 0.99.2, the prefix for http header of request context is changed (`req.http_headers` -> `req.headers` (https://github.com/line/armeria/commit/6f21290a3e6e3bb2af22d8a9061d472dcabda545#diff-83d81e6e8b60860f5efe62a993aaee06L42) Modifications: - Update Armeria official website manual

view details

Trustin Lee

commit sha c4cb5f353464a4fd31ae062d0ffa0ecbca07b01c

git subrepo pull gradle/scripts subrepo: subdir: "gradle/scripts" merged: "d4b0f2a94" upstream: origin: "https://github.com/line/gradle-scripts.git" branch: "master" commit: "edd58813c" git-subrepo: version: "0.4.0" origin: "[email protected]:ingydotnet/git-subrepo" commit: "5d6aba9"

view details

Trustin Lee

commit sha 358d8ac31ba0b9c93194f7821e5de189ec4b795c

Remove Google Analytics from references (#2660) Motivation: We're going to host our Javadoc in javadoc.io, which means it doesn't make sense anymore to inject Googla Analytics scripts in our Javadoc without explicit user consent. Modifications: - Do not inject Google Analytics script into Javadoc/Xref. Result: - OK to deploy the Javadoc to javadoc.io.

view details

Ikhun Um

commit sha fec9fb833275eba59eb9afaf25fc8b7c28132f80

Add HTTP/1 PING(`OPTIONS *`) (#2636) Motivation: Sometimes, a client that sends a request via a load balancer could get `ClosedSessionException` because the load balancer closes the connection while sending the request. Please see #2576 for detail information. Modifications: - Make HTTP/1 client send `PING` with `OPTIONS *` when a connection is idle. - Introduce `IdleTimeoutScheduler` that schedules a PING interval and an idle timeout. - Fork `IdleTimeoutScheduler` from Netty `IdleTimeoutHandler` and add PING idle. - Remove `IdleTimeoutHandler` in favor of `KeepAliveHandler` - Split PING traffic from normal read and write. - Regardless of PING requests and responses, the connection would be closed on idle. - Add `defaultPingIntervalMillis` flag to configure PING interval. - Breaking - Remove `Flags.defaultHttp2PingTimeoutMillis` in favor of `defaultPingIntervalMillis`. - Remove `Flags.defaultUseHttp2PingWhenNoActiveStreams` without replacement. - HTTP/2 can send PING frame with or without active streams. - HTTP/1 can only send PING on no active streams. - Remove `ClientFactory*.http2PingTimeoutMillis()` and `ServerBuilder.http2PingTimeoutMillis()` in favor of `*.pingIntervalMillis()` - Miscellaneous - Move ClientHttp*ObjectEncoder to `client` package from `internal` and make them package private - Move ServerHttp*ObjectEncoder to `server` package from `internal` and make them package private Result: - Armeria Client can send `OPTION *` when a connection is idle. - You can now prevent an unexpected disconnection from an idle timeout of a load balancer. - Fixes #2576

view details

Trustin Lee

commit sha aa96a87ae69c06c7263c85ffd6bd380dd5cac309

Update the project version to 0.99.4-SNAPSHOT

view details

minux

commit sha 30bd38dcf1466d4e327ea94edf495bf6d509ad40

Do not limit ConcurrencyLimitingClient when maxConcurrency is MAX_VALUE (#2667) Motivation: In `ConcurrencyLimitingClient`, we have separate logic when it's unlimited. We should take the case where the concurrency is `MAX_VALUE` as unlimited. Modifications: - Set the `maxConcurrency` as `0` when it's `Integer.MAX_VALUE`. Result: - Less overhead

view details

Trustin Lee

commit sha ad3ac182d12aaadfa4d1648fffe4ff4e47f7e8cc

Fix a leak in `HttpResponseSubscriber` (#2666) Motivation: `HttpResponseSubscriber` must cancel the subscription of a response stream at any case. Unfortunately, when a service sends a response with content for a `HEAD` request, it calls neither `subscription.request()` nor `subscription.cancel()`, making the stream lingers until the connection is closed. Modifications: Make sure `Subscription.cancel()` is called whenever response stream processing is done. Result: The leak is gone.

view details

Koji Lin

commit sha 7e1e3cbabf4714dd891a25deb762c57a615b0e07

Make DNS resolver handle timeout correctly when querying AAAA & A… (#2665) … records at same time If one of the DNS query timeouts when resolver sends A/AAAA queries together, the resolver will have IllegalStateException internally and return DnsTimeoutException. The resolver should return resolved one instead of return error. Modifications: * Return DNS timeout exception only when all the queries are not completed yet. * Make result follow the preferred order Result: * Fixes #2664

view details

minux

commit sha 6ea17b68bee31795ddb7c1467377009f3b909e9b

Use tryFailure in HttpChannelPool (#2671) Motivations: Found this: ``` java.lang.IllegalStateException: complete already: DefaultPromise@7efa5cc1(failure: com.linecorp.armeria.client.SessionProtocolNegotiationException: expected: http, reason: connection established, but session creation timed out: [id: xxxxx) at io.netty.util.concurrent.DefaultPromise.setFailure(DefaultPromise.java:112) at com.linecorp.armeria.client.HttpChannelPool.lambda$connect$6(HttpChannelPool.java:424) at io.netty.util.concurrent.DefaultPromise.notifyListener0(DefaultPromise.java:577) ... ``` The `sessionPromise` can be already failed in `HttpSessionHandler` so we should call `tryFailure`, not `setFailure`.

view details

Trustin Lee

commit sha 4b3e5432d564115f1e9bfc4aff8ac5a82c7055b8

Update the project version to 0.99.5-SNAPSHOT

view details

Trustin Lee

commit sha 5299bea4212fa3976444c42e21e507a07b2d6dd6

Update the project version to 0.44.10-SNAPSHOT

view details

Trustin Lee

commit sha 2906e674f3a1ad99b6fb7e36afd0881c968a21f9

Revert "Update the project version to 0.44.10-SNAPSHOT" This reverts commit 5299bea4212fa3976444c42e21e507a07b2d6dd6.

view details

Charliocat

commit sha f6df6925eb5e4a641ab14e085f9c8619d2fa94d9

Add `Server.blockUntilShutdown()` (#2659) .. which may be useful shortcut for `whenClosed().get()`. - Closes #2651

view details

Trustin Lee

commit sha 18bf90bbd6c2a9b6fdcf9be938f36a922516f8d0

git subrepo pull gradle/scripts subrepo: subdir: "gradle/scripts" merged: "c5d08f78f" upstream: origin: "https://github.com/line/gradle-scripts.git" branch: "master" commit: "dd65c79c0" git-subrepo: version: "0.4.0" origin: "[email protected]:ingydotnet/git-subrepo" commit: "5d6aba9"

view details

Gihwan Kim

commit sha a9544c21e9a60b9bc689eab4a6c7094b4e67e16f

Distinguish initialization and state changes of circuit breaker (#2675) Motivation: General purpose to use `CircuitBreakerListener` is to be notified when circuit breaker is open or re-close. In that aspect, first `CLOSED` event on `onStateChanged()` seems confused. Modification: Add `onInitialized()` to the `CircuitBreakerListener`.

view details

minux

commit sha 948f38c550a51f695647182160cec52bea550d6c

Fix a bug where two threads access MPSC queue (#2687) Motivation: We use MPSC queue in `DefaultStreamMessage`. If more than one thread consumes the MPSC queue, the index of the queue is not correctly updated so a thread might spin forever in `peek()` as we've seen in #2683. This bug was introduced from 0.98.4 #2539 by me. 😢 Modifications: - Fix to consume the MPSC queue only by the subscription thread. Result: - You no longer see the spinning thread in `StreamMessage`.

view details

minux

commit sha be7786238f86250b4f8950ed90dd7319bbdffb84

Use IP address to register service information in ZooKeeper discovery… (#2679) … by default Motivation: It would be nice if we use an IP address instead of `defaultHostName` in ZooKeeper service discovery when a user didn't specify the `Endpoint`. - There's a high chance that the client cannot connect to the server because the `defaultHostName` might not be registered to the DNS server. - The client does not have to do DNS resolution but just connect to the server directly. Modifications: - Change to use the IP address in ZooKeeper discovery when `Endpoint` is not specified. Result: - Close #2674 - Robustness - You can now get a default IP v4 address using `SystemInfo. defaultNonLoopbackIpV4Address()`

view details

Ikhun Um

commit sha b1decb5d0d33122ab4e42a06a0230cbc2f79249c

Support Kotlin-gRPC client CoroutineStub (#2669) Motivation: gRPC for Kotlin is actively developed and [0.1.1](https://github.com/grpc/grpc-kotlin/releases/tag/v0.1.1) has been released last week. Modifications: - Allow GrpcClientFactory creating client from CoroutineStub - Migrate Kotlin example to gRPC-Kotlin Result: - Partially fixes #2662 - You can now run Armeria gRPC client with `gproto*` and gRPC-Kotlin CoroutinStub Co-authored-by: Gary Lo <[email protected]>

view details

Ikhun Um

commit sha 17ad454c06471ff174c06fabc251a6f0bef8985c

Fix a bug where `HttpStreamReader` calls deframer multiple times with… (#2685) … endOfStream. Motivation: If `ArmeriaMessageDefarmer.deframe(HttpData, endOfStream)` is called with `endOfStream` more than once, it will throws `IlligalStateException`. This happens intermittently due to a race condition like the following: 1) `HttpStreamReader` received all data from publisher and added them to unprocessed of `deframer`. 2) A gRPC client does not request next messages yet, so `deframer` still has unprocessedBytes and is not stalled. 3) `HttpStreamReader` receives onCompleted signal and closes `deframer`. 4) A gRPC client requests a message and the received message contains trailers, so `ArmeriaClientCall` tries to close deframer. Modifications: - Add `isClosing()` method to `ArmeriaMessageDefarmer`, it returns whether or not `closeWhenComplete` is `true` - Do not close `deframer`, if `defarmer` is closing. Result: gRPC client can request unprocessed bytes even if `HttpStreamReader` was completed.

view details

push time in 2 months

issue openedNetflix/dyno-queues

Unit test issue: com.netflix.dyno.queues.redis.v2.RedisDynoQueueTest > testTimeoutUpdate FAILED

One test in testTimeoutUpdate is flawed as its result if not determinated.

        boolean updated = rdq.setUnackTimeout(id, 500);
        assertTrue(updated);
        popped = rdq.pop(1, 1, TimeUnit.SECONDS);  // <-- RedisPipelineQueue.processUnacks() might be scheduled while waiting in pop operation
        assertNotNull(popped);
        assertEquals(0, popped.size());

RedisPipelineQueue.processUnacks() is triggered periodically via a scheduled thread pool. If it happens to run while waiting for pop to finish, pop may return some results and the test fails.

The full console output is as follows:

[INFO ] 2020-08-22 14:28:16,790 RedisPipelineQueue - com.netflix.dyno.queues.redis.v2.RedisPipelineQueue is ready to serve {test_queue.x}, shard=x
[DEBUG] 2020-08-22 14:28:18,801 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,810 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,825 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,846 RedisPipelineQueue - Adding 2 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,890 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,898 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,913 RedisPipelineQueue - Adding 2 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:18,947 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue
[DEBUG] 2020-08-22 14:28:19,237 RedisPipelineQueue - Cannot get the message payload some fake id
Message [id=0, payload=Hello World-0, timeout=0, priority=10]
Message [id=1, payload=Hello World-1, timeout=0, priority=9]
Message [id=2, payload=Hello World-2, timeout=0, priority=8]
Message [id=3, payload=Hello World-3, timeout=0, priority=7]
Message [id=4, payload=Hello World-4, timeout=0, priority=6]
Message [id=5, payload=Hello World-5, timeout=0, priority=5]
Message [id=6, payload=Hello World-6, timeout=0, priority=4]
Message [id=7, payload=Hello World-7, timeout=0, priority=3]
Message [id=8, payload=Hello World-8, timeout=0, priority=2]
Message [id=9, payload=Hello World-9, timeout=0, priority=1]
[DEBUG] 2020-08-22 14:28:23,895 RedisPipelineQueue - Adding 1 messages back to the queue for test_queue

expected:<0> but was:<1>
Expected :0
Actual   :1
<Click to see difference>

java.lang.AssertionError: expected:<0> but was:<1>
	at org.junit.Assert.fail(Assert.java:88)
	at org.junit.Assert.failNotEquals(Assert.java:743)
        at com.netflix.dyno.queues.redis.v2.RedisDynoQueueTest.testTimeoutUpdate(RedisDynoQueueTest.java:119)

created time in 2 months

startedpeer-calls/peer-calls

started time in 2 months

startedAgoraIO-Community/Solo

started time in 2 months

push eventopenaphid/openaphid.github.com

Bo

commit sha 4ff8233f02955723775ba190bef524e3c35cce02

disable disqus and add fav ico

view details

push time in 2 months

create barnchopenaphid/Telegram-iOS

branch : Xcode12-fix

created branch time in 2 months

delete branch openaphid/Telegram-iOS

delete branch : Xcode12-fix

delete time in 2 months

create barnchopenaphid/Telegram-iOS

branch : deprecated-Xcode12-fix

created branch time in 2 months

delete branch openaphid/Telegram-iOS

delete branch : main

delete time in 2 months

create barnchopenaphid/Telegram-iOS

branch : deprecated-main

created branch time in 2 months

create barnchopenaphid/Telegram-iOS

branch : master

created branch time in 2 months

issue openedlotabout/skim

Load default color settings from configure file

I'm using a different color scheme in my terminal, which makes the default colors of skim difficult to use. For example, the highlighted words are hard to read.

Although I can use --color= to customize it, I have to apply it in three places:

.bash_alias for sk command
my irg script file for interactive rg
/usr/local/Cellar/sk/0.8.2/etc/bash_completion.d/key-bindings.bash for shell history search

It would be great if there is a configure file in ~/.skrc that I can put the color settings and make it effective globally.

created time in 3 months

startedcvg/Hierarchical-Localization

started time in 3 months

startedcattystats/Anomaly_Detection

started time in 3 months

push eventopenaphid/Telegram-iOS

Bo

commit sha 2fd8a2588749e48d5e6941c66e28f1b091fe29ae

rm a mistakenly added file while solving a merge conflict

view details

push time in 3 months