Datasets:

---

feedback-to-code

/

mozzarella

like 1

Follow

Feedback-2-Code 7

[ "timestamp(timedelta=11.0, similarity=0.9110794132522602)" ]

[ "@blucas makes sense. We love PRs as you know ;)\n", "+1 to PRs :)\n", "Let me fix this.\n", "Fixed by https://github.com/netty/netty/pull/3948\n" ]

[]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[]

[ "timestamp(timedelta=78683.0, similarity=0.8417185682114167)" ]

[ "I misread the code, I thought it was adding the handlers after registering with the event loop, but it's actually adding the handlers before registering with the event loop, which is clearly not allowed.\n", "@jroper can you show me the code you use ?\n", "Sorry, I just noticed... the problem is that my `handlerAdded` callback is throwing an exception, and firing that exception is failing due to no event loop. Up to you if you think this should be handled better, if not, just close this issue.\n", "@jroper it's a bug... Let me fix it.\n", "Fixed by https://github.com/netty/netty/pull/3935\n" ]

[ "Can you do this with real javadocs style ?\n\n``` java\n/**\n * ...\n * ...\n */\n```\n", "same as above \n", "also could you open an openjdk issue and link it ?\n", "@normanmaurer Sure but what does this buy us?\n", "Add CRLF after `**` \n", "Add CRLF after `**`\n", "That you can click the link ;)\n\n> On 07 Jul 2015, at 22:54, Scott Mitchell [email protected] wrote:\n> \n> In transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java https://github.com/netty/netty/pull/3951#discussion_r34089345:\n> \n> > @@ -284,7 +284,19 @@ protected void doWrite(ChannelOutboundBuffer in) throws Exception {\n> > // Only one ByteBuf so use non-gathering write\n> > ByteBuffer nioBuffer = nioBuffers[0];\n> > for (int i = config().getWriteSpinCount() - 1; i >= 0; i --) {\n> > - final int localWrittenBytes = ch.write(nioBuffer);\n> > - final int localWrittenBytes;\n> > - try {\n> > - localWrittenBytes = ch.write(nioBuffer);\n> > - } catch (IllegalArgumentException e) {\n> > - // There appears to be a JDK bug that when a write operation occurs on a direct buffer,\n> > @normanmaurer https://github.com/normanmaurer Sure but what does this buy us?\n> \n> —\n> Reply to this email directly or view it on GitHub https://github.com/netty/netty/pull/3951/files#r34089345.\n", "@normanmaurer - O. eclipse only has this capability when you are declaring a method/variable/calls (places where javadocs typically are) and not in code.\n" ]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java" ]

[ "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java" ]

[]

[ "timestamp(timedelta=1325.0, similarity=0.8820260019713634)" ]

[ "/cc @Scottmitch \n\nLooks like `HttpToHttp2ConnectionHandler` currently requires `FullHttpMessage`. I don't recall whether or not there was a reason that it couldn't take `HttpMessage` or `HttpContent`. Contributions are always welcome :) \n", "I think the main driver was HTTP generally has no proper framing and no stream concept. Without building higher level abstractions it is difficult to correlate one `HttpContent` with a particular stream. `FullHttpMessage` provides the accumulation of all data and definitively associates all HTTP objects with a stream ID.\n\nWe may be able to keep state in `HttpToHttp2ConnectionHandler` as to the current stream ID or create some kind of interface which combines `HttpMessage` and something that provides a stream id (i.e `int http2StreamId()`), but I haven't thought about it too much.\n\nI can think more when I get some more cycles...but as @nmittler said contributions are welcome :)\n", "I'm more than willing to help out with this. As I've stated previously, I have already implemented a rough draft that caters for my scenario.\n\n@Scottmitch addressing the lack of a streaming concept, could we take the same approach used in the SPDY handlers by tracking the current stream ID? (see [here](https://github.com/netty/netty/blob/master/codec-http/src/main/java/io/netty/handler/codec/spdy/SpdyHttpEncoder.java#L121) [here](https://github.com/netty/netty/blob/master/codec-http/src/main/java/io/netty/handler/codec/spdy/SpdyHttpEncoder.java#L163) and [here](https://github.com/netty/netty/blob/master/codec-http/src/main/java/io/netty/handler/codec/spdy/SpdyHttpEncoder.java#L278))\n\nOther question I have is what do we do about trailing headers? It doesn't seem like the current HTTP/2 implementation converts `FullHttpMessage` HTTP/1 trailing headers into some HTTP/2 Frame (I assume we'd use a HEADERS frame as specified in [HTTP Request/Response Exchange](https://http2.github.io/http2-spec/#rfc.section.8.1) point 4.).\n", "> @Scottmitch addressing the lack of a streaming concept, could we take the same approach used in the SPDY handlers by tracking the current stream ID? (see here here and here)\n\nYes I think that is inline with what I was thinking.\n\n> Other question I have is what do we do about trailing headers? It doesn't seem like the current HTTP/2 implementation converts FullHttpMessage HTTP/1 trailing headers into some HTTP/2 Frame (I assume we'd use a HEADERS frame as specified in HTTP Request/Response Exchange point 4.).\n\nYes if we are ignoring trailing headers then that is a bug. The HTTP/2 codec should support this just by doing a `writeHeaders(..)` call. We should just take care to end the stream at the appropriate time.\n", "Sounds good. Any takers?\n\n> Am 09.07.2015 um 03:14 schrieb Scott Mitchell [email protected]:\n> \n> @Scottmitch addressing the lack of a streaming concept, could we take the same approach used in the SPDY handlers by tracking the current stream ID? (see here here and here)\n> \n> Yes I think that is inline with what I was thinking.\n> \n> Other question I have is what do we do about trailing headers? It doesn't seem like the current HTTP/2 implementation converts FullHttpMessage HTTP/1 trailing headers into some HTTP/2 Frame (I assume we'd use a HEADERS frame as specified in HTTP Request/Response Exchange point 4.).\n> \n> Yes if we are ignoring trailing headers then that is a bug. The HTTP/2 codec should support this just by doing a writeHeaders(..) call. We should just take care to end the stream at the appropriate time.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "I can give this a shot tonight. I'll let you know how I get on.\n", "Hey guys,\n\nSorry its taken so long... [here it is](https://github.com/netty/netty/pull/3975)\n", "Closing this issue now that the pull request has been cherry-picked into master (3ec02bf) and 4.1 (57d28dd)\n" ]

[ "Should probably set `release = false` just before the write.\n", "Just to aid in readability, can you separate these blocks out into separate methods (e.g. `writeHttpMessage` and `writeHttpContent`)?\n", "I think we should be able to simplify the logic a bit. Could we do something like this:\n\n``` java\nHttpContent content = null;\n// Just always use the promise aggregator.\nSimpleChannelPromiseAggregator promiseAggregator = new SimpleChannelPromiseAggregator(promise, ctx.channel, ctx.executor);\n\nif (msg instanceof HttpMessage) {\n HttpMessage httpMsg = (HttpMessage) msg;\n\n // Write headers using the promise aggregator.\n}\n\nif (msg instanceof HttpContent) {\n HttpContent httpContent = (HttpContent) msg;\n\n // Write the data.\n\n if (msg instanceof LastHttpContent) {\n LastHttpContent lastHttpContent = (LastHttpContent) msg;\n\n // Write Data\n\n // Write trailers if provided.\n }\n}\n\npromiseAggregator.doneAllocatingPromises();\n```\n\nWDYT?\n", "+1\n", "I think after is good so as writeData() may throw an exception before write any data.\n", "I think the cast is not needed\n", "yeah maybe not a bad idea\n", "toLowerCase(Locale.US) ?\n", "Change signature to accept HttpHeaders ?\n", "+1. Less duplication and we will only have to worry about releasing in the `HttpContent` block where you would expect it.\n", "I'm not sure we should check this here (or worry about resetting `currentStreamId = 0`). The HTTP/2 codec is responsible for enforcing this.\n", "+1 to @nmittler. the writeData is responsible for releasing the data (even in the presence of exceptions) per the interface contract. We should avoid double releasing here and elsewhere in this PR.\n", "I think @nmittler's suggestion (https://github.com/netty/netty/pull/3975#discussion_r34686595) to consolidate the code will help ensure we are releasing correctly.\n", "Lets remove this per https://github.com/netty/netty/pull/3975#discussion-diff-34731790\n", "+1. This code should also be shared with the other method `toHttp2Headers` that also has this code.\n", "I'm not sure this is necessary. It is converted to an `AsciiString` which implies a certain encoding already. WDYT?\n", "Need to call `promiseAggregator.donAllocatingPromises()` here too ... above `setFailure`.\n", "Maybe further simplify the logic here to something like:\n\n``` java\nif (msg instanceof HttpMessage) {\n // Write headers.\n}\n\nif (!endStream && msg instanceof HttpContent) {\n // If this is the last content, extract the trailers.\n boolean lastContent = false;\n Http2Headers trailers = null;\n if (msg instanceof LastHttpContent) {\n lastContent = true;\n trailers = HttpUtil.toHttp2Headers(lastContent.trailingHeaders());\n // TODO: check to see if trailers can really be null here.\n if (trailers == null || trailers.isEmpty()) {\n trailers = null;\n }\n }\n\n release = false;\n encoder.writeData(..., lastContent && trailers == null, ...);\n if (trailers != null) {\n encoder.writeHeaders(..., true, ...);\n }\n\n endStream = lastContent;\n}\n```\n", "@nmittler - I don't think we should have to call `donAllocatingPromises()` in the event we are able to detect an immediate failure. I think the `SimpleChannelPromiseAggregator` should support termination by calling `setFailure(..)`.\n", "+1. I think we can drop the `null` check for `trailers` too.\n", "Ah ok ... wasn't entirely sure. Haven't looked at that class in a while :)\n", "hmm... should this always end the stream if a FullHttpMessage is used ?\n", "I would move the release directly before writeData(...) call. Just in case\n", "As far as I'm aware, yes. If the FullHttpMessage don't not contain any data, the only HTTP/2 frame to write is the HEADERS frame. You can see the old code did the same thing.\n", "You are right. I'll fix that.\n", "nit: just return here.\n\n``` java\nreturn out.add(toHttp2Headers(inHeaders));\n```\n" ]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandler.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpUtil.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandler.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpUtil.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandlerTest.java" ]

[ "timestamp(timedelta=57.0, similarity=0.8544424543669463)" ]

[ "Just do something like this:\n\n``` java\nclass IntObjectHashMap implements Map<Integer, Object> {\n public Object put(Integer key, Object value) {\n put(key.intValue(), value);\n }\n\n public Object put(int key, Object value) {\n // implementation goes here.\n }\n}\n```\n\nThat way one can use this as a normal `Map` and all those `IntObjectMap`, `LongObjectMap` interfaces can be deleted.\n\n@nmittler @normanmaurer \n", "@buchgr the problem is that this will add extra boxing. \n", "I think it would be fine if there were boxing methods in addition to the existing non-boxing methods.\n", "@normanmaurer No it doesn't in my example above ?\n", "@nmittler @buchgr right if we keep both it would be fine\n", "@normanmaurer @nmittler I think we should do it before 4.1 is released. Map implementation would be much more elegant and useful. We should also get rid of the custom interfaces. WDYT?\n", "@buchgr Absolutely, I can't wait to review the PR that you put together without any help from us whatsoever! :stuck_out_tongue_winking_eye:\n", "lol\n", "@nmittler well I would really love to do that, but most unfortunately my hands are tied due to the nice folks at legal :-). \n\nHowever, to our rescue I do know a great and highly motivated dev at Google who would be perfect for the job _wink_ _wink_! And you know ... I am a nice guy, so I am happy to with extreme care and in most detail review and comment on every line of that sweet PR :laughing: \n", "_sigh_ ... I guess that would be \"googly\" of me :/\n", "@buchgr I've thrown together #3984 ... I await my free beer! ;)\n", "@nmittler thanks! I ll make it a pitcher boss :laughing: \n", "+1 :)\n" ]

[ "~~@nmittler those cast's aren't necessary no?~~\n", "@nmittler I would be in favor of getting rid of `PEntry`. Don't think it's really worth it. WDYT? // @normanmaurer @Scottmitch \n", "~~@nmittler can we do something like `containsKey(@primitiveType@)` i.e. `containsKey(int key)`?~~\n", "Yes they are necessary ... see my previous comment. If I don't cast, this triggers autoboxing on the method and leads to ambiguity in the method prototype.\n", "@nmittler that shouldn't be necessary. if it's not of correct type the cast will fail anyway. `get, put, etc.` are quite critical to performance, I think we shouldn't do `instanceof` checks there?\n", "On the fence. The code would certainly be simpler. Having it does provide a lower cost means of iteration, however.\n", "@nmittler yeah I think we should get rid of it ... cause ultimately we need to convert to `Entry<K, V>` anyway as there are no primitive generics. So that doesn't buy us anything?\n", "@nmittler oh you have a `entries()` method ... hmm my guess would be that escape analysis will get rid of the and it won't have any impact, but I have nothing to back that up. So If you wanna keep it, let's keep it :-).\n", "@nmittler s/pPut/put/\n", "@nmittler s/pRemove/remove/\n", "@nmittler s/pGet/get/\n", "@nmittler I have to read up on auto boxing rules T_T\n", "@nmittler alternatively have the primitive map `implement Iterable<PEntry<V>>` and make this an Iterator?\n", "@buchgr you may be right ... I don't feel strongly one way or the other. It would certainly clean up the code to get rid of it.\n\n@normanmaurer @Scottmitch any thoughts?\n", "@buchgr yeah I was considering that as well ... I guess we can live with a `ClassCastException`.\n", "done.\n", "done.\n", "done.\n", "@nmittler remove `.entries()`. \n", "why this double cast ?\n", "I guess remove it should be fine\n", "@nmittler I'm not sure I understand why the casts are needed. Can you give more details what exactly happens if you not cast ? Does it not compile ? \n", "@normanmaurer yes it fails to compile. My understanding of what happens is: the compiler searches for `put(int, int)`. Can't find it. Now it falls back to auto boxing. Both `put(Integer, Integer)` and `put(int, Integer)` are possible due to auto boxing - thus the call is ambiguous. If one casts to `Integer` however, the compiler can immediately find `put(int, Integer)` and doesn't use auto boxing at all.\n", "I think this should be `Long.valueOf(value)`?\n", "Should we make this a static final variable? \n", "@buchgr @normanmaurer yeah that's the behavior I was seeing. Didn't see anything online describing that behavior, but it does fail to compile.\n", "Yes it should be `Long.valueOf(value)` ... done.\n", "done.\n", "done.\n", "`iterator` -> `EMPTY_ITERATOR`?\n\nAlso is this needed anymore? If so should we expose it via a `public static` method of some kind?\n", "if this is `Unmodifiable`...should we `.clone()`?\n", "if this is `Unmodifiable`...should we `.clone()`?\n", "Should we cache and lazy initialize this like `UnmodifiableMap` does?\n\n``` java\npublic Set<K> keySet() {\n if (keySet==null)\n keySet = unmodifiableSet(m.keySet());\n return keySet;\n }\n```\n", "Should we cache and lazy initialize this like `UnmodifiableMap` does?\n", "Should this be unmodifiable? Also we should consider caching this.\n", "Should `EmptyIntObjectMap` be \"script templated\" like the other classes? Also the associated accessors in this class?\n", "Should `UnmodifiableIntObjectMap` be \"script templated\" like the other classes? Also the associated accessors in this class?\n", "Sry I'm a bit late to the discussion on this one. Do we have any benchmarking to determine if getting rid of this is \"acceptable\"? I agree it makes things simpler/cleaner, but wondering what getting rid of the \"native\" entry will cost us.\n", "add `@Override`?\n", "Is this cast safe?\n", "`objectToKey` currently assumes the object will be of the correct type to translate to a native type. For example in `IntObjectHashMap` we have:\n\n``` java\n private int objectToKey(Object key) {\n return (int) ((Integer) key).intValue();\n }\n```\n\nHowever what if the `Object` is not of the correct type?\n", "Can we unroll this `sb.length() == 0` check? I think we just want to unconditionally append a `'{'` before the loop anyways right?\n", "`size == 0` -> `isEmpty()`?\n", "@Scottmitch I don't believe we're using this. Should we just get rid of it?\n", "No ... do we care? One way or another the call will get an exception.\n", "I had previously had a check for `instanceof`, but @buchgr asked that it be removed for performance. I don't think we care since it will throw an exception one way or another.\n", "I think it's fair to assume that iteration over a map is slow :). With that said, I think the places where it's done aren't frequently called or terribly time-sensitive.\n", "Is this used?\n", "@nmittler agree.\n", "Do we need this anymore? I wonder if it is better to just use the `Collection<V>` from `Map`? This way you are not forced to allocate the entire structure in memory.\n", "Now that we no longer have `entries()` this option seems like the only way to \"iterate\" over a native array/collection. This method may be a bit intensive and require a bunch of memory allocation. Is it worth supporting this at the interface level?\n", "I see map allows a `ClassCastException`. Fine as is.\n", "Fine as is.\n\nYes this makes sense. Map allows `ClassCastException` to be thrown. I was originally assuming we should not throw in all cases...but it seems OK where we are using this method currently.\n", "Hum not sure if we are using it...but it may be good for completeness to have this. We are using the empty map...and the unmodifiable should follow the same pattern...so shouldn't be too much extra work?\n", "> I think it's fair to assume that iteration over a map is slow :)\n\nAre we adding boxing penalties on top of the iteration though?\n\n> With that said, I think the places where it's done aren't frequently called or terribly time-sensitive.\n\nOne interesting place where we iterate over the entries is in `DefaultHttp2RemoteFlowController` every time we allocate bytes. It would be good to understand the impact in this case.\n", "+1\n", "Yes, but let's address that in a follow-on PR?\n", "No it's not used ... removed.\n", "Agreed. https://github.com/netty/netty/issues/4001\n", "See https://github.com/netty/netty/issues/4001\n", "+1 thanks!\n", "done.\n", "done.\n", "It is already a copy so it's not a problem if they change it.\n", "It is already a copy so it's not a problem if they change it.\n", "done.\n", "done.\n", "done.\n", "done.\n", "Yes, good catch. Done.\n", "sgtm. thanks!\n", "sgtm. thanks!\n", "Sounds reasonable. Done.\n", "Done.\n", "final ?\n", "done.\n", "@nmittler - Can we leave the `collection` terminology out of this comment? Maybe something like...\n\n```\nGet an iterator to traverse over the primitive entries contained in this map.\n```\n", "nit: Maybe `entryIterator()` to be more consistent with the `Map` interface of `entrySet()`?\n", "add `@SuppressWarnings(\"unchecked\")`\n", "@nmittler - Add a `@SuppressWarnings(\"unchecked\")`?\n", "add `private` on all of these members?\n", "@nmittler - I'm wondering if we should just be verbose and rename `PEntry` to `PrimitiveEntry`?\n", "I don't think this is used anymore.\n", "Do we have to have this as part of the exposed interface? Consider making `private`.\n", "should you do a `toExternal(value)` here?\n", "line 467: `values[index] = toInternal(oldVal);` do you need the `toInternal` here?\n", "@nmittler - Since this involves iterator over the map....and we have seen there may be benefits to using the \"primitive\" iterator... should we keep this in the interface?\n", "`final`?\n", "The class is already private.\n", "@Scottmitch I don't have a strong opinion here. Anyone else have an opinion?\n\n@normanmaurer / @buchgr ?\n", "done.\n", "Not needed ... the cast seems to be fine.\n", "@Scottmitch this shouldn't be needed ... are you seeing an issue in your IDE?\n", "Just to close the loop ... we've reverted to using primitive entries.\n", "@Scottmitch what line are you referring to?\n", "done.\n", "done.\n", "done.\n", "@Scottmitch the other `putAll` method is already optimized for this.\n", "done.\n", "It's not returning an `Iterator`. This method is more analogous to `entrySet`, which returns a `Set` ... this returns an `Iterable`. \n", "@nmittler - Yip. Eclipse doesn't like it.\n", "You are right...fine as is. Read too quickly...thought we were returning an iterator.\n", "The class visibility and the member visibility are independent. Unless you are using these members outside the scope of the inner class (intentionally package private to avoid the potential of synthetic method generate) I think we should set the accessor level like we would for \"regular\" classes.\n", "@nmittler - Sorry I could not leave a comment on the line but I was talking about the `toInternal` call in the `rehash(..)` method.\n", "@nmittler - I'm not sure what we are gaining by abbreviating here. Is `Primitive` overly verbose?\n", "@nmittler no strong opinion. IMO `PrimitiveEntry` sounds better though.\n", "@nmittler - Aren't we casting without checking? Seems like this is an appropriate warning.\n", "@nmittler - Disregard my comment :) I see the `instanceof` check now.\n", "@nmittler - Eclipse doesn't like it :)\n\n```\nType safety: Unchecked cast from Map<capture#4-of ? extends Integer,capture#5-of ? extends V> to IntObjectHashMap<V>\n```\n", "@Scottmitch do you have an example of how one can access a package private field of a private class outside of the private scope? It's not obvious to me that this is possible?\n", "+1 ... this seems strange to me. An example (or a link) would be helpful.\n", "Good enough ... `PrimitiveEntry` it is :)\n", "done.\n", "Ah .. no, shouldn't need that. Done.\n", "done.\n", "The point I am trying to focus on is just because this is an inner class doesn't mean we should use different rules for the internals (unless there is a reason to... i.e. avoid synthetic method generation).\n\n[DefaultHttp2Connection.java#L839](https://github.com/netty/netty/blob/master/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java#L839). The access level does not have to be package private to access them outside the scope, but this is typically done if the member is used outside the scope of the inner class to avoid the potential of synthetic member generation to access an inner private member of the class.\n\nHere is another example:\n\n``` java\npublic class AccessLevelTest {\n private AccessLevelTest() {\n InnerClass inner = new InnerClass();\n int x = inner.a + inner.b + inner.c + inner.d;\n System.out.println(\"x=\" + x);\n }\n\n private static final class InnerClass {\n private int a = 1;\n int b = 2;\n protected int c = 3;\n public int d = 4;\n }\n\n public static void main(String[] args) {\n new AccessLevelTest();\n }\n}\n```\n", "@Scottmitch are you confident that the example you gave does not cause synthetic member generation for `numActiveStreams`? I would expect it would since the class is still private, but perhaps there is some hackery going on in the compiler for this case?\n\nRegardless, I'm fine making the change ... done.\n", "@Scottmitch I don't see why we should care about synthetic method generation. It's come up in a couple of code reviews and I would like to understand why we spend time discussing it? Like what good comes from trying to avoid it? :-)\n" ]

[ "improvement" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ClientUpgradeCodec.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Settings.java", "common/src/main/java/io/netty/util/collection/PrimitiveCollections.java", "common/src/main/script/codegen.groovy", "common/src/main/templates/io/netty/util/collection/KObjectHashMap.template", "common/src/main/templates/io/netty/util/collection/KObjectMap.template", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollEventLoop.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ClientUpgradeCodec.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Settings.java", "common/src/main/java/io/netty/util/collection/PrimitiveCollections.java", "common/src/main/script/codegen.groovy", "common/src/main/templates/io/netty/util/collection/KObjectHashMap.template", "common/src/main/templates/io/netty/util/collection/KObjectMap.template", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollEventLoop.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2SettingsTest.java", "common/src/test/templates/io.netty.util.collection/KObjectHashMapTest.template" ]

[ "timestamp(timedelta=17.0, similarity=0.9038943285140661)" ]

[ "That's strange... let me check\n", "@mingyu89 thanks for reporting.. I think I know what the problem is and I'm working on a fix.\n", "@normanmaurer Thank you!\n" ]

[]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/PendingWriteQueue.java" ]

[ "transport/src/main/java/io/netty/channel/PendingWriteQueue.java" ]

[ "transport/src/test/java/io/netty/channel/PendingWriteQueueTest.java" ]

[ "timestamp(timedelta=32.0, similarity=0.8461993405517902)" ]

[ "@justinjhendrick - Thanks for pointing this out. Also feel free to submit a PR and get credit for fixing :)\n", "@Scottmitch, I would but my company wants to review any open source before we contribute. So, something small like this wouldn't be worth the time necessary for a review.\n", "No worries I will fix it\n\n> Am 18.07.2015 um 00:07 schrieb Justin Hendrick [email protected]:\n> \n> @Scottmitch, I would but my company wants to review any open source before we contribute. So, something small like this wouldn't be worth the time necessary for a review.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Fixed by https://github.com/netty/netty/pull/4003\n" ]

[]

[ "documentation" ]

[ "transport/src/main/java/io/netty/channel/ChannelConfig.java" ]

[ "transport/src/main/java/io/netty/channel/ChannelConfig.java" ]

[]

[ "timestamp(timedelta=64.0, similarity=0.8452350038604813)" ]

[ "@nmittler - FYI :)\n\nhttps://github.com/netty/netty/pull/3984#discussion_r34911766\n" ]

[]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java", "common/src/main/java/io/netty/util/collection/package-info.java", "common/src/main/java/io/netty/util/collection/PrimitiveCollections.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java", "common/src/main/templates/io/netty/util/collection/KCollections.template" ]

[]

[ "timestamp(timedelta=4790.0, similarity=0.8761306193745646)" ]

[ "@fratboy - Thanks for reporting!\n\n@normanmaurer - Assigned to you for now.\n", "@fratboy thanks... fix in the works.\n", "@normanmaurer My pleasure.\n", "@fratboy could you verify https://github.com/netty/netty/pull/4016 ?\n", "@normanmaurer It's exactly same with my patch in my project. It works perfect. \n", "@fratboy - Thanks for the report and verification!\n", "Why `pendingAcquireCount` use int not AtomicInteger ? @normanmaurer @fratboy ", "@wyzssw because everything is done from the same Thread (EventLoop)", "@normanmaurer I get it, thanks " ]

[]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java" ]

[]

[ "timestamp(timedelta=1035.0, similarity=0.889983997942325)" ]

[ "Please review pull request: https://github.com/netty/netty/pull/4027\n", "Will do tomorrow!\n\n> Am 26.07.2015 um 20:30 schrieb ioanbsu [email protected]:\n> \n> Please review pull request: #4027\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Applied all the changes to satisfy #4027 pull requests comments.\nUnfortunately I somehow closed my original pull request. Probably because I tried to do force push so my commits are squashed.. \n\nI had to create new pull request that contains all my original changes and changes that were suggested by @normanmaurer in my original pull request.\nPlease review https://github.com/netty/netty/pull/4029\n", "Fixed by #4076 \n" ]

[ "channelWasNotAcuired -> channelWasNotAcquired (typo)\n", "final\n", "Shouldn't we call also \"channelClosePromise.setFailure(...)\" if not success full ?\n", "You can use FutureListener\n", "You can use FutureListener\n", "Don't we have to execute this with `FixedChannelPool.executor` ?\n", "Hm. should I just do\n\n``` java\nchannelClosePromise.setFailure(new IllegalStateException(\"Failed to close unhealthy channel.\"));\n```\n\n?\n", "Sure.\n" ]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java", "transport/src/main/java/io/netty/channel/pool/SimpleChannelPool.java" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java", "transport/src/main/java/io/netty/channel/pool/SimpleChannelPool.java" ]

[]

[ "timestamp(timedelta=14.0, similarity=0.8902631432145431)" ]

[]

[ "I think the old name was a better fit.\n", "I got it, thanks.\n", "change this to `assert acquiredChannelCount >= 0`\n" ]

[]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java" ]

[]

[ "timestamp(timedelta=14.0, similarity=0.8524517148954559)" ]

[ "@burtonator - Thanks for reaching out! Sounds like a useful feature. I'll put this on the todo list but if you get around to it before we do please feel free to submit a PR.\n", "Thanks will check once i have free cycles\n\n> Am 25.07.2015 um 00:20 schrieb Kevin Burton [email protected]:\n> \n> It would be really nice to have netty support IP_FREEBIND and Any-IP.\n> \n> Essentially it allows you to bind() to any IP address, even if it's not associated with an Interface.\n> \n> I was thinking this could be really helpful in ipv6 setups whereby you just setup your webserver on a subnet with 16k IPs and then when you want to deploy a new webapp, you just give it one of those IPs.\n> \n> The problem is that your box wouldn't want to manually add every IP as your ifconfig would then list 16k IPs.\n> \n> With Any-IP you can add a whole subnet in one command.\n> \n> Then you can bind sockets to it..\n> \n> The PROBLEM is I can't find out how to do this with Java . Using netty would allow you to build really awesome apps this way but I don't think you can bind sockets this way in Java.\n> \n> The C programmers always seem to get the cool toys before we do!\n> \n> http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=ab79ad14a2d51e95f0ac3cef7cd116a57089ba82\n> \n> IP_FREEBIND (since Linux 2.4)\n> If enabled, this boolean option allows binding to an IP\n> address that is nonlocal or does not (yet) exist. This\n> permits listening on a socket, without requiring the\n> underlying network interface or the specified dynamic IP\n> address to be up at the time that the application is trying to\n> bind to it. This option is the per-socket equivalent of the\n> ip_nonlocal_bind /proc interface described below.\n> \n> ... I may be looking at this in the next few weeks/months and if so will just implement it myself.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@burtonator please check https://github.com/netty/netty/pull/4034\n", "Fixed by https://github.com/netty/netty/pull/4034\n" ]

[]

[ "feature" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.h", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollServerSocketChannelConfig.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.h", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollServerSocketChannelConfig.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[]

[ "timestamp(timedelta=78.0, similarity=0.9239012526455588)" ]

[ "We love contributions. Maybe care to submit a PR?\n\n> Am 29.07.2015 um 11:23 schrieb Jorge Castellanos Solaz [email protected]:\n> \n> Netty version: 4.1.0.Beta5\n> \n> Context:\n> I'm implementing an MQTT broker based on netty using the classes from the package io.netty.handler.codec.mqtt, in the version 3.1.1 of the MQTT protocol it has been added a new flag in the variable header of the CONNACK message called \"session present\": http://docs.oasis-open.org/mqtt/mqtt/v3.1.1/os/mqtt-v3.1.1-os.html#_Toc398718035, it would be great if that flag was supported in io.netty.handler.codec.mqtt.MqttConnAckVariableHeader.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@normanmaurer I tried to do it in PR. https://github.com/netty/netty/pull/4042\n\nbut I am not sure if it's OK to do it against 4.1 (the branch we're using) or if I should do it against master and you guys (or me?) take care of the cherry-pick\n\nAlso I just noticed now that I didn't format the commit properly, but it's so trivial I'm not sure it's worth it...\n", "@flozano - Thanks for the PR! Doing against 4.1 should be fine. We will take care of cherry-picking into other branches.\n", "@flozano thanks.. Adjusted commit message and cherry-picked into 4.1(a91e64850505938cfe3a2079e1ac2c247df6532b) and master (6ff66fe967ed1aa45f2e87d263463f35c959b5cf).\n" ]

[]

[ "improvement" ]

[ "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttConnAckVariableHeader.java", "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttDecoder.java", "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttEncoder.java" ]

[ "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttConnAckVariableHeader.java", "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttDecoder.java", "codec-mqtt/src/main/java/io/netty/handler/codec/mqtt/MqttEncoder.java" ]

[ "codec-mqtt/src/test/java/io/netty/handler/codec/mqtt/MqttCodecTest.java" ]

[ "timestamp(timedelta=1386.0, similarity=0.9335526844307486)" ]

[ "Can you also test with 4.1.0.Beta6-SNAPSHOT just to be sure it is not fixed already ?\n", "I've just checked newest commit on branch 4.1 (94f65ed7ff214de3fc9680d80a5ec2d2fe2aaf48) and the problem still exists. \n\nThe problem is in line 115, below comment _'found </, decrementing openBracketsCount'_. IMHO additional validation should be added there to check if closing bracket '>' is available in ByteBuf. I'll have to fix it, my application is broken beacouse of this bug - I'll create pull request ASAP.\n", "@normanmaurer BTW - last version of netty on Maven central is 4.1.0.Beta4, isn't this some oversight?\n", "@tomaszc I can see Beta5 in maven central as well:\nhttp://repo1.maven.org/maven2/io/netty/netty-all/4.1.0.Beta5/\n", "@normanmaurer you're right, it is there, my mistake (morning without coffe ;) )\n\nMeanwhile I've found additional error in XmlFrameDecoder:\n\nStepsToReproduce:\nConfigure XmlFrameDecoder in your Channel Pipeline and send XML message with more than two root XML elements: `<a><b/><a><a><b/><a><a><b/><a>`. \nFirst decoder output will be proper: `<a><b/><a>`. \nSecond decoder output will be corrupt: `<a><b/><a><a><b/><a>` instead of `<a><b/><a>`.\n\nCause:\nIn XmlFrameDecoder class, final part of decode(....) method is corrupt. When creating output frame (line 169) length parameter is invalid. It is based on input ByteBuff readerIndex, but it is not taken to account that it's value can be higher than 0. Therefore if XmlFrameDecoder will process byteBuff which readerIndex was already moved further, it's output will be corrupt.\n", "Pull request with bugfix was created, hope it pass your verification\n" ]

[]

[ "defect" ]

[ "codec/src/main/java/io/netty/handler/codec/xml/XmlFrameDecoder.java" ]

[ "codec/src/main/java/io/netty/handler/codec/xml/XmlFrameDecoder.java" ]

[ "codec/src/test/java/io/netty/handler/codec/xml/XmlFrameDecoderTest.java" ]

[ "timestamp(timedelta=32.0, similarity=0.9286218771515881)" ]

[ "Problematic jar example: https://search.maven.org/remotecontent?filepath=io/netty/netty-transport/5.0.0.Alpha2/netty-transport-5.0.0.Alpha2-sources.jar\n", "@FeiWongReed We love contributions. Maybe submit a PR ?\n", "Ok, will try.\n", "It's easy to fix this, but it was not accidentally mistake. Look at #2058. So I don't know what to do.\n", "Anybody? \n", "@FeiWongReed unfortunaly I'm not really an OSGI user and so have no idea what the right behaviour is. @trustin @Scottmitch maybe you ?\n", "So, which is the right thing? @ChristianSchwarz and @FeiWongReed ?\n", "Which = OSGi manifest in the source/javadoc JAR or not\n", "I'm sure that OSGi manifests must not be in javadocs/jars. Such approach breaks OSGi container from correct dependency resolution and leads to undescriptive NoClassDefFound exceptions. So, I think you should turn off such manifests (or, maybe, move them into maven profile for some strange freaks).\n", "Right, the manifest must not be in the javadoc.jar but in the source bundle\nto be able to compile it in your own.\nAm 19.06.2015 10:33 nachm. schrieb \"Fei Wong Reed\" <[email protected]\n\n> :\n> \n> I'm sure that OSGi manifests must not be in javadocs/jars. Such approach\n> breaks OSGi container from correct dependency resolution and leads to\n> undescriptive NoClassDefFound exceptions. So, I think you should turn off\n> such manifests (or, maybe, move them into maven profile for some strange\n> freaks).\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/3886#issuecomment-113633268.\n", "Thanks ... Will see if I can fix it or if someone else will beat me\n\n> Am 20.06.2015 um 12:15 schrieb Christian Schwarz [email protected]:\n> \n> Right, the manifest must not be in the javadoc.jar but in the source bundle\n> to be able to compile it in your own.\n> Am 19.06.2015 10:33 nachm. schrieb \"Fei Wong Reed\" <[email protected]\n> \n> > :\n> > \n> > I'm sure that OSGi manifests must not be in javadocs/jars. Such approach\n> > breaks OSGi container from correct dependency resolution and leads to\n> > undescriptive NoClassDefFound exceptions. So, I think you should turn off\n> > such manifests (or, maybe, move them into maven profile for some strange\n> > freaks).\n> > \n> > —\n> > Reply to this email directly or view it on GitHub\n> > https://github.com/netty/netty/issues/3886#issuecomment-113633268.\n> > \n> > —\n> > Reply to this email directly or view it on GitHub.\n", "So, does that mean, it's fine to have OSGi manifest in the source JARs?\n", "> So, does that mean, it's fine to have OSGi manifest in the source JARs?\n> I'm sure that NO. Christian, could you show me any artifact in Central, having manifest in source bundle?\n\nChristian, are you understand that manifests in source bundles ruins dependency resolution when we are using OBR repositories?\n\nTrustin, I think you should not add manifests into any jars, which are not containing classes, this is incorrect- you say, that your jar \"provides\" some package, but there is no any class related to that package. If somebody has such specific needs, he should build Netty manually. I think you should leave special profile (like mvn -Pwith-osgi-manifests-in-source-jars), but do not publish broken jars to central\n", "Also, please read this again: https://stackoverflow.com/questions/23149966/classnotfoundexception-for-a-type-that-is-available-to-the-osgi-runtime-io-net\n", "All OSGi Bundes at eclipse orbit with source jars contain a manifest, see\nhttp://download.eclipse.org/tools/orbit/downloads/drops/R20150519210750/.\nIt is also valid to have jars without classes, eclipse does this eg for\nnative libs and images. Can you link a OSGi spec that define that this is\nillegal? I guess you face a special problem with your build enviroment.\nAm 21.06.2015 2:41 nachm. schrieb \"Fei Wong Reed\" <[email protected]\n\n> :\n> \n> Also, please read this again:\n> https://stackoverflow.com/questions/23149966/classnotfoundexception-for-a-type-that-is-available-to-the-osgi-runtime-io-net\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/3886#issuecomment-113894729.\n", "Eclipse != OSGi. In case of OBR repository and OBR dependency resolving such manifests will break it.\n\nOk, I've followed your link. I've downloaded sources jar: http://download.eclipse.org/tools/orbit/downloads/drops/R20150519210750/repository/plugins/ch.qos.logback.core.source_1.0.7.v20121108-1250.jar\n\nHere is manifest.mf from it:\n\n```\nManifest-Version: 1.0\nEclipse-SourceBundle: ch.qos.logback.core;version=\"1.0.7.v20121108-125\n 0\"\nBundle-Vendor: %Bundle-Vendor.0\nBundle-Localization: plugin\nBundle-Name: %Bundle-Name.0\nBundle-SymbolicName: ch.qos.logback.core.source\nEclipse-SourceReferences: scm:cvs:pserver:dev.eclipse.org:/cvsroot/too\n ls:org.eclipse.orbit/ch.qos.logback.core/source-bundle;project=\"ch.qo\n s.logback.core.source\";tag=v20121108-1250\nBundle-Version: 1.0.7.v20121108-1250\nBundle-ManifestVersion: 2\n\nName: ch/qos/logback/core/joran/spi/Pattern.java\nSHA1-Digest: sHNPdRJB7FVHRYmEdjvX/rRt1mo=\n\nName: ch/qos/logback/core/recovery/ResilientSyslogOutputStream.java\nSHA1-Digest: l6xMKJO0mQRqEbSlc18Z0HmM8G8=\n\nName: ch/qos/logback/core/net/SMTPAppenderBase.java\nSHA1-Digest: JIeJrOQyo9RkM3qp2vxbXBbxEos=\n\nName: ch/qos/logback/core/joran/action/AppenderRefAction.java\nSHA1-Digest: G4U1IW6jz2E2iZJ+IQ+I+5j8W8o=\n\nName: ch/qos/logback/core/recovery/ResilientOutputStreamBase.java\nSHA1-Digest: kfXjky4i9bwaw58nKxMOWFclMbI=\n\nName: ch/qos/logback/core/LayoutBase.java\nSHA1-Digest: 1peaXZWI1Yywyt7kVLdpc/EjGBY=\n\nName: ch/qos/logback/core/spi/PreSerializationTransformer.java\nSHA1-Digest: pD7QFRZ6nNyFiB1PwqJzJn4Xh9A=\n\nName: ch/qos/logback/core/spi/ContextAware.java\nSHA1-Digest: l8rvEk9fTZKmQkh08aGSxT4PGNk=\n\nName: ch/qos/logback/core/joran/conditional/ThenAction.java\nSHA1-Digest: VOYwyhJVtkTrV1EUjiqbGhq9bro=\n\nName: ch/qos/logback/core/joran/conditional/PropertyEvalScriptBuilder.\n java\nSHA1-Digest: y4lp1A8Q9mUkPVNGxOpXlNyEGBQ=\n\nName: ch/qos/logback/core/pattern/Converter.java\nSHA1-Digest: vNhpputskGgMH6klE3PXElF8+DE=\n\nName: ch/qos/logback/core/joran/action/StatusListenerAction.java\nSHA1-Digest: Rh1xOYKDaes9FlFWW1G9qTgsjR4=\n\nName: ch/qos/logback/core/rolling/DefaultTimeBasedFileNamingAndTrigger\n ingPolicy.java\nSHA1-Digest: teeeTU1+9XNxujQF+DLOWqQNl3s=\n\nName: ch/qos/logback/core/rolling/RollingPolicyBase.java\nSHA1-Digest: 753cT1XGe9spECLRDTajA2ZyD3Y=\n\nName: ch/qos/logback/core/rolling/helper/Compressor.java\nSHA1-Digest: bTfz2L8sWVJCjd5h7psWFOE2pdI=\n\nName: ch/qos/logback/core/pattern/FormattingConverter.java\nSHA1-Digest: GVTVQUvquTAkbfO/9vR8bqEnIWI=\n\nName: ch/qos/logback/core/db/dialect/DBUtil.java\nSHA1-Digest: tqfXn2Azkq15iAPV8gIYlMRPhU8=\n\nName: ch/qos/logback/core/rolling/FixedWindowRollingPolicy.java\nSHA1-Digest: j+zg+Kh41YtA9CGx7IiaeKUDuLs=\n\nName: ch/qos/logback/core/joran/action/ConversionRuleAction.java\nSHA1-Digest: y6bJbrLCjVKXVQ305ekqG4e0y2c=\n\nName: ch/qos/logback/core/helpers/NOPAppender.java\nSHA1-Digest: hYjPjD+M8aytSroOYFPdZuEuHE0=\n\nName: ch/qos/logback/core/util/CachingDateFormatter.java\nSHA1-Digest: BZUBcEYM1ZBcAeQdGoI5Boz7g3Y=\n\nName: ch/qos/logback/core/joran/action/ContextPropertyAction.java\nSHA1-Digest: J9mSjgj4n3oBPAvdvbl9H7sXDfY=\n\nName: ch/qos/logback/core/helpers/Transform.java\nSHA1-Digest: epTLhXSh08uiHDwh23l81QwE6oQ=\n\nName: ch/qos/logback/core/filter/AbstractMatcherFilter.java\nSHA1-Digest: O18jGQmiisRL/+8LDTTYMeqY4ks=\n\nName: ch/qos/logback/core/ContextBase.java\nSHA1-Digest: NDg3FYubdWZS07UeDbm+uBKCldc=\n\nName: ch/qos/logback/core/db/dialect/SQLDialectCode.java\nSHA1-Digest: PaRVPF7DZmCLv5mLNPlLWO1nMR0=\n\nName: ch/qos/logback/core/html/IThrowableRenderer.java\nSHA1-Digest: YoB8Sg9ZaJJ2WHcN4S2YMJCNvXo=\n\nName: ch/qos/logback/core/joran/conditional/Condition.java\nSHA1-Digest: N0gzcbNKtESMyF41paeS8wMFtUw=\n\nName: ch/qos/logback/core/package.html\nSHA1-Digest: 8+6ySXqVNteK7ZwtgpR5iePJj8k=\n\nName: ch/qos/logback/core/joran/action/ActionConst.java\nSHA1-Digest: iqlRnJlCSA0lXPeL1i4fLCqJAf0=\n\nName: ch/qos/logback/core/db/dialect/HSQLDBDialect.java\nSHA1-Digest: Ulxp4amMaoYk+bfCdYC4yrirGCE=\n\nName: ch/qos/logback/core/status/StatusChecker.java\nSHA1-Digest: 1l4AuYHWVCv3WEBp1QPidK7zhuI=\n\nName: ch/qos/logback/core/joran/spi/HostClassAndPropertyDouble.java\nSHA1-Digest: d1A9lQQK0ErQUIb0y4UeHkfOhJs=\n\nName: ch/qos/logback/core/util/EnvUtil.java\nSHA1-Digest: zcukyMJEpi8Muo4MT7FJWuXp29o=\n\nName: ch/qos/logback/core/recovery/RecoveryCoordinator.java\nSHA1-Digest: zim+aFWhiF8up9EMDphkACo5z0s=\n\nName: ch/qos/logback/core/rolling/helper/MonoTypedConverter.java\nSHA1-Digest: V8soiVkqymqCDTpX8Q3BAR3zIe4=\n\nName: ch/qos/logback/core/boolex/Matcher.java\nSHA1-Digest: dQLtZCqtm8Fvf5/QRKkVbrdYAXQ=\n\nName: ch/qos/logback/core/spi/FilterReply.java\nSHA1-Digest: KZtGChUPxQTZXIQw2y4pMes+eYg=\n\nName: ch/qos/logback/core/db/BindDataSourceToJNDIAction.java\nSHA1-Digest: 4UdTaIdVmBhJvPaudu6eVy2NNDo=\n\nName: ch/qos/logback/core/pattern/color/BoldYellowCompositeConverter.j\n ava\nSHA1-Digest: rst4Ay/EO9q1MZZUGImSl9I/tK0=\n\nName: ch/qos/logback/core/pattern/util/AsIsEscapeUtil.java\nSHA1-Digest: TbXVCDgtht0T81W74akKZER0S1Q=\n\nName: ch/qos/logback/core/boolex/JaninoEventEvaluatorBase.java\nSHA1-Digest: Uo9gjC1i685cyV8ME8LnKOWdyPI=\n\nName: ch/qos/logback/core/util/FileSize.java\nSHA1-Digest: LBFVjtVYuylGouZNRZsy1dNXvzM=\n\nName: ch/qos/logback/core/pattern/color/ForegroundCompositeConverterBa\n se.java\nSHA1-Digest: b/4JwVk1IQCdDzDnaIM8YahR3ek=\n\nName: ch/qos/logback/core/status/ViewStatusMessagesServletBase.java\nSHA1-Digest: 3+TB5zIbjhNqtrB0XcVdQDkz3Dw=\n\nName: ch/qos/logback/core/joran/spi/EventPlayer.java\nSHA1-Digest: PrqQwAn4pHlxl6OxPVEjr0g0hso=\n\nName: ch/qos/logback/core/util/package.html\nSHA1-Digest: 1dtJQn1ZAkxu7rs95ELfXFm4J7E=\n\nName: ch/qos/logback/core/util/Loader.java\nSHA1-Digest: K1cHVIy14AUj40KEEgVfexTsRHw=\n\nName: ch/qos/logback/core/util/PropertySetterException.java\nSHA1-Digest: 0/hwKj1+JdZrnt8byq106ERWua8=\n\nName: ch/qos/logback/core/ConsoleAppender.java\nSHA1-Digest: 578ovg8hfeUTMawrdZcK9MckC/M=\n\nName: ch/qos/logback/core/db/dialect/MySQLDialect.java\nSHA1-Digest: Ru8GOSTLrL8rEji5cTkO7SRpbaY=\n\nName: ch/qos/logback/core/pattern/package.html\nSHA1-Digest: xFtYv+Gz1C4mMcsCX/OVu/Zd0Eo=\n\nName: ch/qos/logback/core/pattern/parser/FormattingNode.java\nSHA1-Digest: 03gX1rr35hgVQzhrqdWAq9MO2uI=\n\nName: ch/qos/logback/core/rolling/TimeBasedFileNamingAndTriggeringPoli\n cyBase.java\nSHA1-Digest: FZzRwPJVXFVu0NFfJFzv1Uti4xY=\n\nName: ch/qos/logback/core/LogbackException.java\nSHA1-Digest: pnavpqP/jlBkD6yzmwQ1fEB+sj0=\n\nName: ch/qos/logback/core/joran/action/NestedComplexPropertyIA.java\nSHA1-Digest: DGBJJIkdGGCzwwtsYnqC37S5Bwg=\n\nName: ch/qos/logback/core/status/StatusUtil.java\nSHA1-Digest: KwzgQH7kka+/cqFGI0HZfcNpQU8=\n\nName: ch/qos/logback/core/status/package.html\nSHA1-Digest: 3cOCFjILSEGEttKgfDMgM8UwwZQ=\n\nName: ch/qos/logback/core/spi/PropertyContainer.java\nSHA1-Digest: okumGtGxl8aXMmJaLJmQmhUTjmY=\n\nName: ch/qos/logback/core/rolling/helper/DefaultArchiveRemover.java\nSHA1-Digest: sBxwSauPnQVU6Gzgxo0V5FWoxPs=\n\nName: ch/qos/logback/core/status/StatusListener.java\nSHA1-Digest: IUkm8ArgKwntxTvjh/VTbpNzlmU=\n\nName: ch/qos/logback/core/joran/spi/ActionException.java\nSHA1-Digest: OuvFBt567mwq+fscjOUtIBKtQGo=\n\nName: ch/qos/logback/core/joran/spi/XMLUtil.java\nSHA1-Digest: CWe0KOq5YNOtBX6VEbo0NcbsPsc=\n\nName: ch/qos/logback/core/pattern/color/MagentaCompositeConverter.java\nSHA1-Digest: Q59Xu1wdW0JGaFKpsdkuQRUQeiI=\n\nName: ch/qos/logback/core/rolling/TriggeringPolicy.java\nSHA1-Digest: Y0DNZr1B/e4YeBtuWpQjiiUTGPg=\n\nName: ch/qos/logback/core/AsyncAppenderBase.java\nSHA1-Digest: jKO5i1J448mIoZq7UnnCVi30rdw=\n\nName: ch/qos/logback/core/helpers/CyclicBuffer.java\nSHA1-Digest: fqAx7X8xPfRv0DU5Er7nnM46nAM=\n\nName: ch/qos/logback/core/joran/util/ConfigurationWatchListUtil.java\nSHA1-Digest: bYGq6T0dFuW7T/e0uBlhZmQX9Ns=\n\nName: ch/qos/logback/core/filter/Filter.java\nSHA1-Digest: lDSeOOlS/FbYuRUHcWp9RvO99CY=\n\nName: ch/qos/logback/core/joran/event/BodyEvent.java\nSHA1-Digest: sMFlBJftQZLbQmJ6sJQDV34TgBE=\n\nName: ch/qos/logback/core/db/dialect/SybaseSqlAnywhereDialect.java\nSHA1-Digest: EEZ/oEjSgm7aEabVCjNWlQ9vJmo=\n\nName: ch/qos/logback/core/spi/FilterAttachableImpl.java\nSHA1-Digest: LHdV8aWRgNC4bh2GzMt3lxS6go4=\n\nName: ch/qos/logback/core/encoder/ObjectStreamEncoder.java\nSHA1-Digest: YZR0ME4Cfab8Ac0FbsSuBbrOsR0=\n\nName: ch/qos/logback/core/AppenderBase.java\nSHA1-Digest: a6tO7MbBVGauTsqb+1HxjKREPSM=\n\nName: ch/qos/logback/core/joran/event/InPlayListener.java\nSHA1-Digest: qKokwLfNyvASbTyc9mqepBh0JJo=\n\nName: ch/qos/logback/core/filter/EvaluatorFilter.java\nSHA1-Digest: yppxSAqJA2PwhrIhHRxIjbrB6n0=\n\nName: ch/qos/logback/core/spi/PropertyDefiner.java\nSHA1-Digest: J3PbO1RgXQW6saajnFFcu+Wr1HU=\n\nName: ch/qos/logback/core/joran/action/ImplicitAction.java\nSHA1-Digest: Y+Bbb6VyW4CJX+hFsJN23S/UjS4=\n\nName: ch/qos/logback/core/BasicStatusManager.java\nSHA1-Digest: Td0BVtaQPoy+qVGZhoS3FgwI9DU=\n\nName: ch/qos/logback/core/filter/package.html\nSHA1-Digest: aAioE9ut4fjo6aqitW4XkBLbAk4=\n\nName: META-INF/eclipse.inf\nSHA1-Digest: KyT9FF7C7t86NoBoa2kZT3ZJBfw=\n\nName: ch/qos/logback/core/status/StatusListenerAsList.java\nSHA1-Digest: MH49WCtRucf/XCgvUWW7O0YCpfM=\n\nName: ch/qos/logback/core/db/JNDIConnectionSource.java\nSHA1-Digest: U8YoR9i1u3HB4JZibsiwwq4RVuU=\n\nName: ch/qos/logback/core/boolex/EvaluationException.java\nSHA1-Digest: qsQXELmV7nRIUetBiZxXlwdAPtQ=\n\nName: ch/qos/logback/core/joran/action/PropertyAction.java\nSHA1-Digest: 7quqwIxgz1DdeLgvf/ahK4lMsXk=\n\nName: ch/qos/logback/core/util/Duration.java\nSHA1-Digest: 1WkBtSKNejuaWaUG2ZngDjBuoag=\n\nName: ch/qos/logback/core/util/AggregationType.java\nSHA1-Digest: T9OpjW5f6kwGeMvlGAy5yPCj7yI=\n\nName: ch/qos/logback/core/status/InfoStatus.java\nSHA1-Digest: DbWEPAYUJnnGPS4s/ZFCl1xDtMg=\n\nName: ch/qos/logback/core/util/TimeUtil.java\nSHA1-Digest: mTVjC/ULm2PK0FdouR54IMykajc=\n\nName: ch/qos/logback/core/rolling/helper/RollingCalendar.java\nSHA1-Digest: wx6IpUpgrAOKWUiWP63Gr7kRmow=\n\nName: ch/qos/logback/core/pattern/color/BoldCyanCompositeConverter.jav\n a\nSHA1-Digest: W+9f7HNEZJL/prvim/wpUisPqYc=\n\nName: ch/qos/logback/core/sift/SiftingAppenderBase.java\nSHA1-Digest: bpcl+kN3J426s/Ns9u9nGqIEw9g=\n\nName: ch/qos/logback/core/util/DynamicClassLoadingException.java\nSHA1-Digest: wIStsISSjy/DChx9d8hm3sb3Xns=\n\nName: ch/qos/logback/core/util/DatePatternToRegexUtil.java\nSHA1-Digest: y1KawRY6tHgvmvlhnN/aPPJb/qM=\n\nName: ch/qos/logback/core/boolex/EventEvaluatorBase.java\nSHA1-Digest: vn4orj5LFuaK67m2A+JdO6pMlmM=\n\nName: ch/qos/logback/core/sift/AppenderTracker.java\nSHA1-Digest: wQumGBDr144Skpm3yt2CG2Rb9PQ=\n\nName: ch/qos/logback/core/encoder/EventObjectInputStream.java\nSHA1-Digest: zQ58cujG4+GYLnMCMNtPBKd3b3U=\n\nName: ch/qos/logback/core/spi/ContextAwareBase.java\nSHA1-Digest: aVMKbueOs0NZ4Ov5hCGE4sx3hdw=\n\nName: ch/qos/logback/core/PropertyDefinerBase.java\nSHA1-Digest: HHYlXsxjjUrEQRsXmiDpfp5f1Ck=\n\nName: ch/qos/logback/core/joran/action/AbstractEventEvaluatorAction.ja\n va\nSHA1-Digest: IeSmLZw2zt2Jbb3HUkJxlNxpC+o=\n\nName: ch/qos/logback/core/joran/spi/ConsoleTarget.java\nSHA1-Digest: UPgEyFGnRaPbTkRGcxUKgpYY494=\n\nName: ch/qos/logback/core/rolling/helper/FileFilterUtil.java\nSHA1-Digest: mg4s0px1cC6/RUDuqUcpkjVit/I=\n\nName: ch/qos/logback/core/pattern/PatternLayoutBase.java\nSHA1-Digest: 0rEOyz31Kn1r6wDBG+XVCULz+ZA=\n\nName: plugin.properties\nSHA1-Digest: MPvQwfEZ2Ta16BKiIhpeXf7finI=\n\nName: ch/qos/logback/core/spi/ContextAwareImpl.java\nSHA1-Digest: iIYJQxt53DzahpN706ia1/eOSqQ=\n\nName: ch/qos/logback/core/rolling/helper/IntegerTokenConverter.java\nSHA1-Digest: xC/Q7ofmiVw5h2nu71uQpgxferw=\n\nName: ch/qos/logback/core/spi/CyclicBufferTrackerImpl.java\nSHA1-Digest: ei46lQrdMqQ7Nu7Q4vkCuCL3ijc=\n\nName: ch/qos/logback/core/spi/AppenderAttachable.java\nSHA1-Digest: GF/2bIoWlf7ncxA4NDHpNlDwh+A=\n\nName: ch/qos/logback/core/joran/event/StartEvent.java\nSHA1-Digest: hG9xCba3ttsO0zdgwRiVj6kzIM4=\n\nName: ch/qos/logback/core/pattern/color/BoldMagentaCompositeConverter.\n java\nSHA1-Digest: XASeC3L3Ik3/d9SJgvRdueIfZZM=\n\nName: ch/qos/logback/core/status/OnConsoleStatusListener.java\nSHA1-Digest: sirJLxju2Fh6BEHSlkDMnt8T3Sc=\n\nName: ch/qos/logback/core/CoreConstants.java\nSHA1-Digest: Gkp1n7rwFqmR2kuB3sZSKdtJXqk=\n\nName: ch/qos/logback/core/net/SocketAppenderBase.java\nSHA1-Digest: a5R5kbnMyPibbQaoChlshKb6ha0=\n\nName: ch/qos/logback/core/rolling/helper/DateTokenConverter.java\nSHA1-Digest: KhGlL1JbFuX1TI7NyBdtEdAwaw0=\n\nName: ch/qos/logback/core/util/StatusPrinter.java\nSHA1-Digest: XrgbERGhfvctjBwA2Xcj76Q5oN4=\n\nName: ch/qos/logback/core/util/FileUtil.java\nSHA1-Digest: L4dpqFja1RLUYXz2N9giqMEKDiw=\n\nName: ch/qos/logback/core/util/InvocationGate.java\nSHA1-Digest: jDBFfkKHeQYdcsB7Tcf9MkxxpJQ=\n\nName: ch/qos/logback/core/rolling/helper/CompressionMode.java\nSHA1-Digest: wC8RearGZyD04yvIgnlG8HTL2Qs=\n\nName: ch/qos/logback/core/joran/action/IADataForComplexProperty.java\nSHA1-Digest: gV9K5EeJAW8bLmHAdf9so77n9r8=\n\nName: ch/qos/logback/core/status/StatusBase.java\nSHA1-Digest: vd6faOOrcnwZ/bDF4SWmDmdFln4=\n\nName: ch/qos/logback/core/joran/action/IADataForBasicProperty.java\nSHA1-Digest: 6VtahKWICX4c/ii/YJUnUhqvMBI=\n\nName: ch/qos/logback/core/joran/conditional/PropertyWrapperForScripts.\n java\nSHA1-Digest: fz+TwAwTNY5DukrLMUPfyzCY6Mk=\n\nName: ch/qos/logback/core/OutputStreamAppender.java\nSHA1-Digest: 6lWghwyLrlVJlZMlBoPL+Gfi8Dw=\n\nName: ch/qos/logback/core/pattern/color/CyanCompositeConverter.java\nSHA1-Digest: 8YpWPxA6j2woSnFrQAiNLDnrmyU=\n\nName: ch/qos/logback/core/pattern/parser/Node.java\nSHA1-Digest: 9LJwrAKuPkQSkioRd1iGf3Bj7vw=\n\nName: ch/qos/logback/core/db/dialect/SQLiteDialect.java\nSHA1-Digest: R1BMUL1SkJS9U6e8zNjmgbsAXrw=\n\nName: ch/qos/logback/core/pattern/parser/Compiler.java\nSHA1-Digest: jwPK4nLp8nrTgye5spCMlQhRPJs=\n\nName: ch/qos/logback/core/Context.java\nSHA1-Digest: W+onX5fpzWXj34su6qgdY+6dXvk=\n\nName: ch/qos/logback/core/rolling/RolloverFailure.java\nSHA1-Digest: yf3nlvcvIJmB5Fr/y2jeuyxX8QE=\n\nName: ch/qos/logback/core/joran/util/PropertySetter.java\nSHA1-Digest: ZTofSoqz3Em6eLJHSARJqqZ3YSk=\n\nName: ch/qos/logback/core/pattern/ReplacingCompositeConverter.java\nSHA1-Digest: eTYBnRwFwT5gTpEs3zG403p7o3w=\n\nName: ch/qos/logback/core/spi/LifeCycle.java\nSHA1-Digest: ZlgRrZIr2miDOn1wPMNZ5ewtFnc=\n\nName: ch/qos/logback/core/encoder/EchoEncoder.java\nSHA1-Digest: UgG9wygMI4BRRxir88tDDkoO80c=\n\nName: ch/qos/logback/core/pattern/PatternLayoutEncoderBase.java\nSHA1-Digest: ZKAQeR4MOPx7CUWO0PSM9SPIBSc=\n\nName: ch/qos/logback/core/pattern/color/RedCompositeConverter.java\nSHA1-Digest: gqS55ljcnrrb8Y8OJzC7/YMwFPw=\n\nName: ch/qos/logback/core/joran/action/NewRuleAction.java\nSHA1-Digest: /c+CxeYq9WpoH66GIiE9rnEyoVM=\n\nName: ch/qos/logback/core/rolling/RollingFileAppender.java\nSHA1-Digest: hlGm+x7nmrxvF124GVXqoZN21s4=\n\nName: ch/qos/logback/core/subst/Parser.java\nSHA1-Digest: NC9lQc0ZJX8fsiG4nV18uVNE1uE=\n\nName: ch/qos/logback/core/db/DBHelper.java\nSHA1-Digest: NfiXIgR5l8J4lDBRtmAFCo/WaTc=\n\nName: ch/qos/logback/core/spi/CyclicBufferTracker.java\nSHA1-Digest: OSKHaXJnGQub0LT2/VFw5X0b9PY=\n\nName: ch/qos/logback/core/boolex/package.html\nSHA1-Digest: CwXbedL7omfMx8gxMwfm/jUZqKw=\n\nName: ch/qos/logback/core/spi/AppenderAttachableImpl.java\nSHA1-Digest: 4PavjCnbo+j0M0HNZ9mnZ16dcck=\n\nName: ch/qos/logback/core/encoder/NonClosableInputStream.java\nSHA1-Digest: lOrFMVpRag2bKzUziL+Az3gKZsI=\n\nName: ch/qos/logback/core/read/CyclicBufferAppender.java\nSHA1-Digest: TzJKdZVbq8etIomeSgDFxgiyEXk=\n\nName: ch/qos/logback/core/html/NOPThrowableRenderer.java\nSHA1-Digest: 3K4lQ93jMp9eaNUm6h7nMAyXGMk=\n\nName: ch/qos/logback/core/db/dialect/MsSQLDialect.java\nSHA1-Digest: 1gXzLLHobeJZBhWlSLjhkHPhZSA=\n\nName: ch/qos/logback/core/net/JMSAppenderBase.java\nSHA1-Digest: PKZXlAIEzPLpgrZKrSQAHR3GsTk=\n\nName: ch/qos/logback/core/spi/LogbackLock.java\nSHA1-Digest: werNSXtufm5IA/P+3pzznOszis8=\n\nName: ch/qos/logback/core/db/dialect/SQLDialect.java\nSHA1-Digest: BtWcgnk3qlsqcmUITbnuw0ybBiU=\n\nName: ch/qos/logback/core/joran/action/NOPAction.java\nSHA1-Digest: dyHXrkthld0T/Ds4iISbM2R64fk=\n\nName: ch/qos/logback/core/db/DataSourceConnectionSource.java\nSHA1-Digest: otmrShBY1GEH9k78KsoaN8DAQc0=\n\nName: ch/qos/logback/core/joran/conditional/ThenOrElseActionBase.java\nSHA1-Digest: S9qVQjoNkwVykgWnWHjihBvR3qA=\n\nName: ch/qos/logback/core/status/ErrorStatus.java\nSHA1-Digest: 1uxABECJF31xpM/O1kAUr/keEoo=\n\nName: ch/qos/logback/core/pattern/color/BlackCompositeConverter.java\nSHA1-Digest: foqg/LIsLt3jXgcLr/+3aeWObXs=\n\nName: ch/qos/logback/core/subst/Token.java\nSHA1-Digest: bAGKsMt3+tLx7e8lU/HRDbnqYXg=\n\nName: ch/qos/logback/core/spi/DeferredProcessingAware.java\nSHA1-Digest: Au+28B3mgY7/A8+DenojmSIG8qw=\n\nName: ch/qos/logback/core/db/dialect/PostgreSQLDialect.java\nSHA1-Digest: Riq9GSsRVp0PMkU3K8TFzbAhDSY=\n\nName: ch/qos/logback/core/pattern/parser/CompositeNode.java\nSHA1-Digest: xjOb+d37Q7eRmhFWEtNAd7WIhaA=\n\nName: ch/qos/logback/core/rolling/TriggeringPolicyBase.java\nSHA1-Digest: x9h5kDa0AOJx9t2BZvyA7LIhheA=\n\nName: ch/qos/logback/core/pattern/SpacePadder.java\nSHA1-Digest: ytIL4TGUo3H/qjNjgTerUXObGvA=\n\nName: ch/qos/logback/core/rolling/helper/AsynchronousCompressor.java\nSHA1-Digest: d2C/qzI+5JJBXX/jTApNkUH5oJI=\n\nName: ch/qos/logback/core/rolling/TimeBasedFileNamingAndTriggeringPoli\n cy.java\nSHA1-Digest: UGF48SMIFzRERFYEEwxxQZdM8yE=\n\nName: ch/qos/logback/core/pattern/color/ANSIConstants.java\nSHA1-Digest: 1JXLqn5pN4AdTU9PSdIRmwln5N4=\n\nName: ch/qos/logback/core/encoder/ByteArrayUtil.java\nSHA1-Digest: 08o/ojOhFgYbf0anOyfDnk1Nzng=\n\nName: ch/qos/logback/core/util/IncompatibleClassException.java\nSHA1-Digest: fybBUECVKTMyrrGbZr7OeAmbx2g=\n\nName: ch/qos/logback/core/subst/NodeToStringTransformer.java\nSHA1-Digest: u9OjQJdfiRs7vtAxDg+yP1HoYY0=\n\nName: ch/qos/logback/core/rolling/helper/CompressionRunnable.java\nSHA1-Digest: wdElup4L+TbSH1bo56BA5A/Wybg=\n\nName: ch/qos/logback/core/rolling/SizeAndTimeBasedFNATP.java\nSHA1-Digest: QrHYuQEARP+BxASa+cRk8sxKXwA=\n\nName: ch/qos/logback/core/util/ContextUtil.java\nSHA1-Digest: 30LHqxnHeq/1JmIfMaPJQQlc2to=\n\nName: ch/qos/logback/core/FileAppender.java\nSHA1-Digest: jMN+nyoVI1M7Ot37lNu7gEFSprM=\n\nName: ch/qos/logback/core/pattern/parser/Parser.java\nSHA1-Digest: u9hU/XJCWy/9nicCnxcHNAYO2ew=\n\nName: ch/qos/logback/core/status/StatusManager.java\nSHA1-Digest: 45sbRR7rIzEEOSCQIISB7c0KcBc=\n\nName: ch/qos/logback/core/sift/SiftingJoranConfiguratorBase.java\nSHA1-Digest: BimiloBJBflAfh3AjKGdaqPFs50=\n\nName: ch/qos/logback/core/pattern/color/BoldGreenCompositeConverter.ja\n va\nSHA1-Digest: QZY5PnRYUoFcBTGbSDq4SOm56Fk=\n\nName: ch/qos/logback/core/joran/JoranConfiguratorBase.java\nSHA1-Digest: 3L+YISaEHK6O22hsol6orm2kC8Q=\n\nName: ch/qos/logback/core/db/ConnectionSource.java\nSHA1-Digest: lrQxqf6jxOvo69IjiYEb8b6sdlI=\n\nName: ch/qos/logback/core/html/package.html\nSHA1-Digest: hrY9FdHpLTUk42rleabwDRY79yo=\n\nName: ch/qos/logback/core/pattern/color/YellowCompositeConverter.java\nSHA1-Digest: E2c/joRm0dhNdlv4jy3O9sGcl2k=\n\nName: ch/qos/logback/core/joran/spi/Interpreter.java\nSHA1-Digest: LX3FxVtxnIt/hhXtCEtS0nRNwLg=\n\nName: ch/qos/logback/core/db/DriverManagerConnectionSource.java\nSHA1-Digest: pzsKicZzVo5Psd9jNMPKTc1ziGg=\n\nName: ch/qos/logback/core/db/package.html\nSHA1-Digest: jfW/GhbFFQA6+uGczSPV/3lLP5U=\n\nName: ch/qos/logback/core/Layout.java\nSHA1-Digest: ygVxXGf9G6HK2qtwFBRCJM30nxo=\n\nName: ch/qos/logback/core/pattern/parser/SimpleKeywordNode.java\nSHA1-Digest: 0mFIn69YW8X91CRcrOKz9/0p3a0=\n\nName: ch/qos/logback/core/pattern/parser/Token.java\nSHA1-Digest: uok2O7s6GKY19sf2UmSAwX1SY8E=\n\nName: ch/qos/logback/core/rolling/helper/RenameUtil.java\nSHA1-Digest: pEmLEW/0of6WTdmv15eMdnYCGHQ=\n\nName: ch/qos/logback/core/Appender.java\nSHA1-Digest: x68woIji4s7w+twWFToy5MZcrgE=\n\nName: ch/qos/logback/core/rolling/helper/package.html\nSHA1-Digest: ueqVvqTrMl4kRlLardlW49rrGng=\n\nName: ch/qos/logback/core/pattern/color/BoldWhiteCompositeConverter.ja\n va\nSHA1-Digest: 9FPr9AV7tmgYmcAW+zL3ETCDW4I=\n\nName: ch/qos/logback/core/joran/spi/InterpretationContext.java\nSHA1-Digest: WkNSxskNCgux1CaiHz0t8xj2TEQ=\n\nName: ch/qos/logback/core/joran/spi/RuleStore.java\nSHA1-Digest: +hbjNx2Z/47q9cqWlXpvCj8rDkQ=\n\nName: ch/qos/logback/core/joran/util/StringToObjectConverter.java\nSHA1-Digest: rfQrg1VCq/gAUg4tDhv0NJrfy10=\n\nName: about.html\nSHA1-Digest: oXBJd6QAQkK3IFu1yqF1bVxYwzA=\n\nName: ch/qos/logback/core/net/package.html\nSHA1-Digest: ZBbiCQSwRZGXLrv3llwql0fHf3I=\n\nName: ch/qos/logback/core/html/HTMLLayoutBase.java\nSHA1-Digest: ZhE5vAu/DRheEjkzWdzMObEaMk8=\n\nName: ch/qos/logback/core/joran/package.html\nSHA1-Digest: QAGuLuzX2/sz0kuX9cSTTIGFpDU=\n\nName: ch/qos/logback/core/sift/DefaultDiscriminator.java\nSHA1-Digest: QgwuLcUNid5w4RElA8rkDw2yW8o=\n\nName: ch/qos/logback/core/rolling/helper/ArchiveRemover.java\nSHA1-Digest: jfOcJVVOAJo4dvR8MYdkvRJu3AY=\n\nName: ch/qos/logback/core/db/dialect/OracleDialect.java\nSHA1-Digest: VAGDm5YfyJeEMolvqX59PT1+kYg=\n\nName: ch/qos/logback/core/joran/conditional/ElseAction.java\nSHA1-Digest: 8tzSuFpq9PUdTQibtf2yedxy+8E=\n\nName: ch/qos/logback/core/joran/spi/JoranException.java\nSHA1-Digest: eHRYRexSgyq3EFYdk2LG7i/TWqM=\n\nName: ch/qos/logback/core/joran/action/TimestampAction.java\nSHA1-Digest: CrzkRWxE4HkfU8E0VLyY8KrvOmo=\n\nName: ch/qos/logback/core/sift/AppenderTrackerImpl.java\nSHA1-Digest: tDfwsynGJSDgBPeMnsVyyJFktvg=\n\nName: ch/qos/logback/core/helpers/package.html\nSHA1-Digest: 76FRxUzqQFPfbE46M2L47IMOjQg=\n\nName: ch/qos/logback/core/net/SyslogAppenderBase.java\nSHA1-Digest: 2epTG4IO/E33TA6Q2W8p3pnzd/U=\n\nName: ch/qos/logback/core/pattern/DynamicConverter.java\nSHA1-Digest: 5q+3OrFoey0WTSnMTZUNDk9Szxo=\n\nName: ch/qos/logback/core/spi/ScanException.java\nSHA1-Digest: J/Yqzczb0SRAETBs+EA7ClUs3Gk=\n\nName: ch/qos/logback/core/pattern/PostCompileProcessor.java\nSHA1-Digest: xqVp0qj3CV/yp751OvPukrk0iXY=\n\nName: ch/qos/logback/core/net/LoginAuthenticator.java\nSHA1-Digest: QZyZjSEC2N4laMN3S+dWHkRra4g=\n\nName: ch/qos/logback/core/pattern/util/RestrictedEscapeUtil.java\nSHA1-Digest: OCIUUnikPGSwZdkA7tx+Y1XHjWQ=\n\nName: ch/qos/logback/core/rolling/TimeBasedRollingPolicy.java\nSHA1-Digest: h9zrbz94uhzrNbsQNiu2dA3dJTc=\n\nName: ch/qos/logback/core/pattern/CompositeConverter.java\nSHA1-Digest: QVO88JXv18IXTiViVBU+Sh2OOAU=\n\nName: ch/qos/logback/core/boolex/EventEvaluator.java\nSHA1-Digest: U4r2M8UEY9OUP70vP1slKNewSao=\n\nName: ch/qos/logback/core/db/DBAppenderBase.java\nSHA1-Digest: miOobjYq+b/VmKlFIYBTDA/QCl8=\n\nName: ch/qos/logback/core/rolling/helper/PeriodicityType.java\nSHA1-Digest: 8Xph+1NhgngfL+ozd16Ljd4oTkA=\n\nName: ch/qos/logback/core/joran/conditional/IfAction.java\nSHA1-Digest: b5QiHEdoiDoe9R1nMwPGMTjQSuo=\n\nName: ch/qos/logback/core/pattern/parser/OptionTokenizer.java\nSHA1-Digest: x0sLSlfiisJKyqlJ5AaHcSJ73B0=\n\nName: ch/qos/logback/core/subst/Node.java\nSHA1-Digest: QbLjY7V8F2DYdeqOWZyvwovYSYc=\n\nName: ch/qos/logback/core/html/CssBuilder.java\nSHA1-Digest: 1qzLZuP2Js3+rKU1dE5wU/CKCUE=\n\nName: ch/qos/logback/core/joran/spi/NoAutoStart.java\nSHA1-Digest: eBWrHSZCPVvenK2PsNemc/gFUCI=\n\nName: ch/qos/logback/core/joran/event/EndEvent.java\nSHA1-Digest: O5on3MRqIwq2pTl8c95pdzS1/4A=\n\nName: ch/qos/logback/core/util/ContentTypeUtil.java\nSHA1-Digest: 1JC5EjdjYbs4+1Kt1ph4ufMmbjI=\n\nName: ch/qos/logback/core/pattern/FormatInfo.java\nSHA1-Digest: nVGYLyct4zX0i38x2FLQ/aFxHJw=\n\nName: ch/qos/logback/core/db/dialect/package.html\nSHA1-Digest: 6gsqlEAzu/vRCsYiP3+zOqF7HZ0=\n\nName: ch/qos/logback/core/pattern/color/BoldBlueCompositeConverter.jav\n a\nSHA1-Digest: cTu9b8ecwh5vgm50LvMMIEZfrMY=\n\nName: ch/qos/logback/core/pattern/color/GreenCompositeConverter.java\nSHA1-Digest: gwwo7CdOVHFpwru03vD/iafvyh4=\n\nName: ch/qos/logback/core/util/OptionHelper.java\nSHA1-Digest: iiIgzw/BO793soPKonwTPLeXMjQ=\n\nName: ch/qos/logback/core/joran/action/ParamAction.java\nSHA1-Digest: 2JTWRMezvHQSu5ZxIGR9JftwhZ0=\n\nName: ch/qos/logback/core/util/SystemInfo.java\nSHA1-Digest: vz6Bh68BnSxo4k6fCItlRxIKCWI=\n\nName: ch/qos/logback/core/joran/action/IncludeAction.java\nSHA1-Digest: CfjVp5ZWD54M6PeNgUl1FbPz3rk=\n\nName: ch/qos/logback/core/joran/spi/DefaultNestedComponentRegistry.jav\n a\nSHA1-Digest: nFt39iAYwYZqtfHSgDOGriaSfNw=\n\nName: ch/qos/logback/core/encoder/LayoutWrappingEncoder.java\nSHA1-Digest: si8gkRgprpC39AztKgeQ9hFltAA=\n\nName: ch/qos/logback/core/pattern/ConverterUtil.java\nSHA1-Digest: Z1fo3MV/8tUMn5rOUYGGsJ6LsqQ=\n\nName: ch/qos/logback/core/rolling/helper/FileNamePattern.java\nSHA1-Digest: 8ghUApdJaJD+aXUqR5NPDnDF94Q=\n\nName: ch/qos/logback/core/sift/Discriminator.java\nSHA1-Digest: BWUWQesDtdaot7A4H2orf+XtVPA=\n\nName: ch/qos/logback/core/pattern/util/RegularEscapeUtil.java\nSHA1-Digest: JgBkrzaHZD2UwMs5AOzntafjHTQ=\n\nName: ch/qos/logback/core/joran/spi/ConfigurationWatchList.java\nSHA1-Digest: Nm5eDunUC0VEoe0M5bIJnHqwe3I=\n\nName: ch/qos/logback/core/joran/spi/DefaultClass.java\nSHA1-Digest: Hev0GR3/ScknPd6kAg8rVv6WoEk=\n\nName: ch/qos/logback/core/joran/spi/NoAutoStartUtil.java\nSHA1-Digest: 9A0nkJz3Ftr8io3PrcN3DtCtSmM=\n\nName: ch/qos/logback/core/joran/spi/SimpleRuleStore.java\nSHA1-Digest: QkgazK7YtUj/4lP66ih8n+eTsaY=\n\nName: ch/qos/logback/core/pattern/util/AlmostAsIsEscapeUtil.java\nSHA1-Digest: pLTxJnL2E2pDynLwgrEY9e42TX8=\n\nName: ch/qos/logback/core/spi/package.html\nSHA1-Digest: eprPoiEf1rkdx0lTTN5liuLrrws=\n\nName: ch/qos/logback/core/joran/action/Action.java\nSHA1-Digest: 4lGEuGo04GVf6SP7GJSYLqL0XWw=\n\nName: ch/qos/logback/core/pattern/parser/TokenStream.java\nSHA1-Digest: 3XcJjMaqLbUXeXSL5hMDbDnkkaw=\n\nName: ch/qos/logback/core/recovery/ResilientFileOutputStream.java\nSHA1-Digest: K62vIjZT4ZI0yuQiaxFUkVyCyCc=\n\nName: ch/qos/logback/core/pattern/color/WhiteCompositeConverter.java\nSHA1-Digest: ekBdnb8Qt3uyDv/76ViuZNarB28=\n\nName: ch/qos/logback/core/rolling/package.html\nSHA1-Digest: ch1Pvz2/QHGNULM+tuWarTr+rjE=\n\nName: ch/qos/logback/core/db/dialect/H2Dialect.java\nSHA1-Digest: YgO5aP9NbK2Y5kElx8b2+PtW5vo=\n\nName: ch/qos/logback/core/status/Status.java\nSHA1-Digest: R39ux17CA6Z1zdbUypHOXzemt3E=\n\nName: ch/qos/logback/core/pattern/util/IEscapeUtil.java\nSHA1-Digest: oV9iSi/xUj6qD6ylLG4mMElOAmE=\n\nName: ch/qos/logback/core/spi/FilterAttachable.java\nSHA1-Digest: 7T15eTgp2a3jS8b8kdMrDCb8k8I=\n\nName: ch/qos/logback/core/subst/Tokenizer.java\nSHA1-Digest: q1KGylC2xUS70Qr+48PLl7EPtYA=\n\nName: ch/qos/logback/core/encoder/Encoder.java\nSHA1-Digest: c66ZnqwgxwlrrbI8xfBJMr7R0rM=\n\nName: ch/qos/logback/core/joran/event/SaxEventRecorder.java\nSHA1-Digest: 0O00/lYnSXnWQTADscuRsnYWVao=\n\nName: ch/qos/logback/core/net/SyslogConstants.java\nSHA1-Digest: 4MoskN0Ur8m25y+LEMQxj9Eegxs=\n\nName: ch/qos/logback/core/rolling/SizeBasedTriggeringPolicy.java\nSHA1-Digest: CX562TFKm9cqAujvgzTleRVXf2c=\n\nName: ch/qos/logback/core/property/FileExistsPropertyDefiner.java\nSHA1-Digest: ONNsiokOHkYzqRMF7vE+ingrsqU=\n\nName: ch/qos/logback/core/util/CharSequenceState.java\nSHA1-Digest: 6uIawZMiypbWmBBmGg4HUgzFIRw=\n\nName: ch/qos/logback/core/joran/action/ActionUtil.java\nSHA1-Digest: +mBUb74Jkm8rhR/tB3p2t02bYw4=\n\nName: ch/qos/logback/core/sift/AppenderFactoryBase.java\nSHA1-Digest: B7JGGX3HNURfNDoiMyE2n++XbS8=\n\nName: ch/qos/logback/core/joran/GenericConfigurator.java\nSHA1-Digest: 2hAMADa/spoKc9RWjQy1/A2m/w0=\n\nName: ch/qos/logback/core/status/WarnStatus.java\nSHA1-Digest: dl5pCmrwP/gwQgHU+iQ+7bflQ4M=\n\nName: ch/qos/logback/core/UnsynchronizedAppenderBase.java\nSHA1-Digest: 5jtRt4TcynUqSpy2sbqlL9+snKw=\n\nName: ch/qos/logback/core/layout/EchoLayout.java\nSHA1-Digest: gFQG3qwEGqDIglyQEDU/BAX75RA=\n\nName: ch/qos/logback/core/rolling/helper/SizeAndTimeBasedArchiveRemove\n r.java\nSHA1-Digest: jr44KPCeOxiCJbt/SObTKdRCark=\n\nName: ch/qos/logback/core/joran/action/NestedBasicPropertyIA.java\nSHA1-Digest: PfhZ1a4LOfIiNZ9Xb12c4UPS6AY=\n\nName: ch/qos/logback/core/pattern/color/BlueCompositeConverter.java\nSHA1-Digest: JABbdew6dFpsRpsRwubP+iVouSQ=\n\nName: ch/qos/logback/core/rolling/RollingPolicy.java\nSHA1-Digest: G/DbDkYMqTSPg9KY04GoMXK2MX8=\n\nName: ch/qos/logback/core/rolling/helper/TimeBasedArchiveRemover.java\nSHA1-Digest: aAZd/UxnqwPzB0RpRLCFAQ3sQRw=\n\nName: ch/qos/logback/core/joran/action/DefinePropertyAction.java\nSHA1-Digest: 0G2Fpb00kCpHRTq081IDdiUdkVg=\n\nName: ch/qos/logback/core/joran/event/SaxEvent.java\nSHA1-Digest: 8jsRV06P4O0b4Oa/78hwOQkysBc=\n\nName: ch/qos/logback/core/helpers/ThrowableToStringArray.java\nSHA1-Digest: /kr1tifsOoSt1Xg9fMPyc5SEgqg=\n\nName: ch/qos/logback/core/pattern/IdentityCompositeConverter.java\nSHA1-Digest: UYpVI6WB2s/WjGksvbrxv0o2pm4=\n\nName: ch/qos/logback/core/joran/action/AppenderAction.java\nSHA1-Digest: y1Qw2C8sZZyIkfIHk7WYsMd6IEs=\n\nName: ch/qos/logback/core/pattern/LiteralConverter.java\nSHA1-Digest: rL/BlouGFBzH0D+1A5tQqjqatY4=\n\nName: ch/qos/logback/core/rolling/helper/TokenConverter.java\nSHA1-Digest: LBdX5HV7TYS4DOJoWGaf9JL16rA=\n\nName: ch/qos/logback/core/read/ListAppender.java\nSHA1-Digest: wNh/onbjQaeAHJg5mZP8I1TqmYA=\n\nName: ch/qos/logback/core/db/ConnectionSourceBase.java\nSHA1-Digest: O3O4cvc1OZqrtal1F/85rzoMpug=\n\nName: ch/qos/logback/core/encoder/EncoderBase.java\nSHA1-Digest: 3e32ZzqPSRoyVX8K++8TIRhTe6E=\n\nName: ch/qos/logback/core/net/SyslogOutputStream.java\nSHA1-Digest: h/8d3eh4ME2/jZ4Xj086CNFSLbA=\n\nName: ch/qos/logback/core/pattern/color/BoldRedCompositeConverter.java\nSHA1-Digest: 02y9KR0wHyNdo95NgWPDGnZ7qno=\n\nName: ch/qos/logback/core/util/CharSequenceToRegexMapper.java\nSHA1-Digest: 5o/XSCT9EV/Zhg3dPdN8U3tfGeg=\n```\n\nCould you show me any Export-Package or Import-Package directive?\n\nSo, I'm insisting that almost all artifacts in Central has no manifests with Exports or Imports in source jars. I'm insisting that it's incorrect and breaks dependency resolution. I'm insisting that you should not include manifests with exports/imports into source jars.\n\nOk, here is javadoc for export-package: https://osgi.org/javadoc/r5/core/org/osgi/framework/Constants.html#EXPORT_PACKAGE\n\n> Manifest header identifying the packages that the bundle offers to the Framework for export. \n\nThere is nothing to export from source jar! There are no classes inside!\n\nPlease stop breaking our fragile world.\n", "Guys, could you just disable manifest copying?..\n", "@normanmaurer - I'm also not familiar with OSGi.\n\n@FeiWongReed - Would you mind submitting a PR to remedy the issue (I don't think many Netty devs have OSGi experience)? Maybe it will be easier for you and @ChristianSchwarz to discuss the pain points.\n", "Ok, here is pull request: https://github.com/netty/netty/pull/4045\n\nI think that I solved problem at all. For now main jars have correct osgi manifests and source jars have eclipse-specific manifests.\n", "@FeiWongReed - Thanks! Lets move the detailed discussion to the PR.\n" ]

[]

[ "defect" ]

[ "pom.xml" ]

[ "pom.xml" ]

[]

[ "timestamp(timedelta=54.0, similarity=1.0000000000000002)" ]

[]

[]

[]

[ "buffer/src/main/java/io/netty/buffer/PoolThreadCache.java" ]

[ "buffer/src/main/java/io/netty/buffer/PoolThreadCache.java" ]

[]

[ "timestamp(timedelta=37.0, similarity=0.8739797762840745)" ]

[ "@normanmaurer - I recall discussing this when we originally reviewed the pool. I think your rational was since we have to check the health anyways when we pull out of the pool, and the health check may be expensive (may include RTT), we wanted to minimize the number of health checks done. WDYT?\n", "@Scottmitch yep that was my argument, I may been wrong here and we should kind of make this configurable. WDYT ?\n", "@normanmaurer yes. As this issue describes I think there is value in having the option to check up front too.\n", "Fixed by https://github.com/netty/netty/pull/4078\n", "@Scottmitch Please make sure not to introduce a build breakage when you cherry-pick into 4.0 and 4.1\n", "@trustin - I'm sorry. Will be more careful in the future. Thanks for cleaning up after my mess.\n", "@Scottmitch No worries. I make the same mistake over and over. :-)\n" ]

[ "As discussed in https://github.com/netty/netty/issues/4077#issuecomment-130461684. Perhaps we should make this check configurable (some boolean flag to indicate we should check health when putting back into pool).\n", "@Scottmitch sure, will do. What do you think the default behavior should be? Check health on release by default or do not check by default?\n", "My vote is default should check the health. If this hurts performance in specific use cases it can be switched off. Our default health checker `ChannelHealthChecker.ACTIVE` shouldn't involve any network traffic so should be OK. @normanmaurer - WDYT?\n", "I'm not sure we need to verify / require this text is in the message.\n\nAlso do we need the `ExpectedException` `@Rule`? Is this not something we can catch in normal code flow? I think we typically catch expected exceptions and `fail()` if we don't get the exception.\n", "@Scottmitch, so that I know for sure that expected exception(not any random) had been thrown. \n", "@ioanbsu - Understood, but I don't think we need to couple the tests so tightly to the implementation that the message in the exception is verified. I think just verifying the correct type of exception is thrown at the correct time should be sufficient.\n", "@Scottmitch, got it. removed message validation.\n", "You also need to empty the stack trace as done for \"FULL_EXCEPTION\" below.\n", "`true` -> `{@code true}`.\n", "`true` -> `{@code true}`.\n", "Is it necessary to have `offerHealthyOnly` as an argument? Seems like we are just passing in the member variable when ever this is called.\n", "Consider renaming to `releaseHeathCheck`....and also the same for the publicly visible in constructors.\n", "remove `final`.\n", "should this just be `<p>`, or do you want to wrap the above in `<p>...</p>`?\n", "You're right. will fix.\n", "Why?\n", "This is not part of my changes actually. This part got auto-reformatted by IDEA based on the rules that are used in netty project. I can certainly change it to <p>\n", "just a nit, but it is not needed.\n", "Why this change?\n", "This is general code style practice that I try to follow in my projects. It helps to prevent possible bugs. If there's no need to change the value of a variable and someone does, the IDE and/or compiler will provide a warning.\nIt may also be a helping hint to compiler to perform some optimizations for the code.\n\nI don't mind removing it if in netty projects you don't have such rules.\n", "Lets revert for this PR. I understand the rational, and do not necessarily disagree, but lets keep the set of changes focused on what is required to resolve the issue. If need be we can open a followup PR focused on using `final` more consistently.\n", "Sounds good.\n", "Done.\n" ]

[ "feature" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java", "transport/src/main/java/io/netty/channel/pool/SimpleChannelPool.java" ]

[ "transport/src/main/java/io/netty/channel/pool/FixedChannelPool.java", "transport/src/main/java/io/netty/channel/pool/SimpleChannelPool.java" ]

[ "transport/src/test/java/io/netty/channel/pool/SimpleChannelPoolTest.java" ]

[ "timestamp(timedelta=29.0, similarity=0.8904482250405097)" ]

[ "@yawkat can you tell me what you mean with \"abstract domain socket\" ?\n", "See [this man page](http://man7.org/linux/man-pages/man7/unix.7.html). Basically, domain socket names allow `\\0` bytes in them, and on linux, a domain socket starting with `\\0` does not appear in the file system and is called abstract. \n\n[socat](http://linux.die.net/man/1/socat) supports these through a special `ABSTRACT` address family. [dbus](http://dbus.freedesktop.org/doc/dbus-specification.html) supports them through an `abstract` property in their address system. \n\nSorry for not linking the exact paragraphs, I'm on mobile. \n", "@yawkat so I guess using 'GetStringChars' would be right here... I'm correct ?\n", "Wouldn't that return an array of 16-bit numbers? Casting that to 8 bit would just give you lots of 0s.\n\nI'm guessing the right way would either be decoding it as ascii or passing through the byte array directly instead of the string. Also need to replace the string operations because they rely on a trailing 0.\n", "@yawkat good point. \n", "Okay, had nothing else to do so I tried my hand on a fix: #4092\n", "Fixed by https://github.com/netty/netty/pull/4092\n" ]

[ "Use : getBytes(CharsetUtil.UTF8)\n", "Use : getBytes(CharsetUtil.UTF8)\n" ]

[ "feature" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[ "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollAbstractDomainSocketEchoTest.java" ]

[ "timestamp(timedelta=9.0, similarity=0.8442024319863696)" ]

[ "@rkapsi thx will check\n", "@normanmaurer - I got this one.\n", "Interesting, if I run that test from the CLI it's eventually starting to spew these ones...\n\n```\njava.nio.channels.ClosedChannelException\n java.nio.channels.ClosedChannelException\n java.nio.channels.ClosedChannelException\n java.io.IOException: readAddress() failed: Invalid argument\n at io.netty.channel.epoll.Native.newIOException(Native.java:129)\n at io.netty.channel.epoll.Native.ioResult(Native.java:145)\n at io.netty.channel.epoll.Native.readAddress(Native.java:286)\n at io.netty.channel.epoll.AbstractEpollChannel.doReadBytes(AbstractEpollChannel.java:244)\n at io.netty.channel.epoll.AbstractEpollStreamChannel$EpollStreamUnsafe.epollInReady(AbstractEpollStreamChannel.java:790)\n at io.netty.channel.epoll.EpollEventLoop.processReady(EpollEventLoop.java:345)\n at io.netty.channel.epoll.EpollEventLoop.run(EpollEventLoop.java:253)\n at io.netty.util.concurrent.SingleThreadEventExecutor$4.run(SingleThreadEventExecutor.java:699)\n at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:137)\n at java.lang.Thread.run(Thread.java:745)\n java.io.IOException: readAddress() failed: Invalid argument\n at io.netty.channel.epoll.Native.newIOException(Native.java:129)\n at io.netty.channel.epoll.Native.ioResult(Native.java:145)\n at io.netty.channel.epoll.Native.readAddress(Native.java:286)\n at io.netty.channel.epoll.AbstractEpollChannel.doReadBytes(AbstractEpollChannel.java:244)\n at io.netty.channel.epoll.AbstractEpollStreamChannel$EpollStreamUnsafe.epollInReady(AbstractEpollStreamChannel.java:790)\n at io.netty.channel.epoll.EpollEventLoop.processReady(EpollEventLoop.java:345)\n at io.netty.channel.epoll.EpollEventLoop.run(EpollEventLoop.java:253)\n at io.netty.util.concurrent.SingleThreadEventExecutor$4.run(SingleThreadEventExecutor.java:699)\n at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:137)\n at java.lang.Thread.run(Thread.java:745)\n java.io.IOException: readAddress() failed: Invalid argument\n at io.netty.channel.epoll.Native.newIOException(Native.java:129)\n at io.netty.channel.epoll.Native.ioResult(Native.java:145)\n at io.netty.channel.epoll.Native.readAddress(Native.java:286)\n at io.netty.channel.epoll.AbstractEpollChannel.doReadBytes(AbstractEpollChannel.java:244)\n at io.netty.channel.epoll.AbstractEpollStreamChannel$EpollStreamUnsafe.epollInReady(AbstractEpollStreamChannel.java:790)\n at io.netty.channel.epoll.EpollEventLoop.processReady(EpollEventLoop.java:345)\n at io.netty.channel.epoll.EpollEventLoop.run(EpollEventLoop.java:253)\n at io.netty.util.concurrent.SingleThreadEventExecutor$4.run(SingleThreadEventExecutor.java:699)\n at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:137)\n at java.lang.Thread.run(Thread.java:745)\n java.io.IOException: readAddress() failed: Invalid argument\n at io.netty.channel.epoll.Native.newIOException(Native.java:129)\n at io.netty.channel.epoll.Native.ioResult(Native.java:145)\n at io.netty.channel.epoll.Native.readAddress(Native.java:286)\n at io.netty.channel.epoll.AbstractEpollChannel.doReadBytes(AbstractEpollChannel.java:244)\n at io.netty.channel.epoll.AbstractEpollStreamChannel$EpollStreamUnsafe.epollInReady(AbstractEpollStreamChannel.java:790)\n at io.netty.channel.epoll.EpollEventLoop.processReady(EpollEventLoop.java:345)\n at io.netty.channel.epoll.EpollEventLoop.run(EpollEventLoop.java:253)\n at io.netty.util.concurrent.SingleThreadEventExecutor$4.run(SingleThreadEventExecutor.java:699)\n at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:137)\n at java.lang.Thread.run(Thread.java:745)\n\n... many more of the same\n```\n", "@rkapsi - Thanks. Will check. Can I take some of your code you posted and morph it into another form of unit test?\n", "@Scottmitch yes, feel free.\n\nThe readAddress() exceptions are possibly a result of me adding a finally block to that test (because I've been turning it also into an unit test on my end).\n\n```\n} finally {\n server.close().syncUninterruptibly();\n }\n```\n", "@rkapsi - Thanks. I have a PR pending approval. I will ping you when I push it upstream and have you verify the changes.\n" ]

[]

[ "defect" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollServerChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollDatagramChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollDomainSocketChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollServerChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollDatagramChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollDomainSocketChannel.java" ]

[ "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollSocketChannelTest.java" ]

[ "timestamp(timedelta=126215.0, similarity=0.8547458327797111)" ]

[ "The only solution to this bug would be to dispose the ChannelGroup, so once a CG is closed, it remains closed and every new add()-invocation results in closing the channel.\n\nThe only other solution is to synchronize on the CG which would be bad.\n\nEDIT: Is there nobody who can verify if I'm right or wrong ?\n", "Can really nobody prove or disprove this statement ? \n", "@chrisprobst I'm not sure I understand the actual problem... Can you show a unit test that demonstrate the issue ?\n", "I'm actually on holiday (small laptop, no ide, no real internet etc.), but I will try to explain the problem.\n- Status-quo:\n - Doc suggests closing ServerChannel + ChildChannels by using a ChannelGroup\n - Doc suggests adding ChildChannels to the ChannelGroup in their corresponding channelActive method\n - Docs says, this is enough to close everything\n- Problem lies between accepting a channel and invoking ChildChannel.channelActive\n - Time between ServerChannel.close() and ChildChannel.channelActive is _not_ atomic\n- Flow to reproduce the problem\n - Let the ServerChannel \"read\" (accept) the incoming ChildChannel\n - ChildChannel.channelActive is scheduled for later invocation \n - THEN: Call ChannelGroup.close()\n - ServerChannel gets close()'d first, then all _registered_ ChildChannels\n - BUT: The previously accepted ChildChannel is _not_ yet registered\n - The previously started and scheduled ChildChannel will work just fine and launch, but it shouldn't =(\n - Boom: Race-Condition, which can lead to never closed ChildChannels\n\nI hope, this made it clear. It's actually very easy to reproduce, my initial source example (I know, not a testable unit) shows how.\n\nThis hit me quite hard, I simply couldn't find the bug, but it was the evil ChannelGroup xD.\nI know, usually this is a very short time frame and unlikely to happen. But it _is_ a race condition, IMHO.\n", "@chrisprobst I agree this is a race condition. I didn't think deep on how we can fix this, but the easiest workaround is to shut down the event loop, which closes every single channel it manages.\n", "Hi Trustin,\n\nmaybe we should consider creating a ChannelGroup, which remains closed. So every channel, which adds itself after the ChannelGroup has been closed, will also be closed during adding.\n\nEDIT: And of course, changing the doc =).\n", "@chrisprobst That sounds like a good idea IMHO. Care to sign the [ICLA](http://netty.io/s/icla) and sending us a PR? What do you think @normanmaurer ? \n", "I believe, I've already done die signing, but I'm not sure. Can I check this ?\n\nAbout the PR: I could do the change when I'm back from my holiday ;) But maybe it's too close, after holiday I'm in Russia for a few days. Mid September sounds realistic.\n\nThe only problem is, that using this technique, the channel group will not be Lock-Free anymore, because IHMO we would need a Lock, since adding to the group depends on the state of the flag, which seems to be unsolvable with Atomics\\* or volatiles. I have to refresh my knowledge about concurrent stuff I think.\n\nMaybe Doug Lea can help :+1: \n", "Hey @trustin @normanmaurer ,\n\nI've just thought about it a bit deeper. I think the ChannelGroup is not the best spot for solving this race condition. A closed ChannelGroup is messier than I first thought, how should other methods react to a closed state ?\n\nMaybe we should follow the path of less resistance - just change the doc, so nobody will do the same mistake as me.\n\nWe could just enhance the doc, which suggests using it as Server/Child-Channel manager.\nSo instead of saying, that a ChannelGroup on its own is enough, we could suggest using a simple bool flag, which should be very fine and should guard against this nasty race condition.\nThe following shows by idea:\n\n```\n// In the ChildHandler we can do:\n\n @Override\n public void channelActive(ChannelHandlerContext ctx) throws Exception {\n channelGroup.add(ctx.channel());\n if(volatileClosedFlag == true) { ctx.channel().close(); } // Must be checked AFTER adding!!\n\n super.channelActive(ctx);\n }\n\n// If you want to close ALL channels:\n// [...]\nvolatileClosedFlag = true; // Set this BEFORE closing\ngroup.close();\n// [...]\n```\n\nMy Java-Memory-Model knowledge is a bit rusty, but I think this should do it.\n\nIMHO we need at this point no locks, as long as we check the flag for being set AFTER adding child channels to it.\n\nSo the time-flow is something like: \n- set flag to true -> close channels async -> add child to channel group -> check for flag, close() if true\n\nConcurrency is always a bit of a brain twister, maybe I'm wrong. But I believe this should be safe.\n\nBest Regards,\nChris\n", "Ok, to keep it simple:\n\n```\npublic final class CloseableChannelGroup extends DefaultChannelGroup {\n\n private volatile boolean closed = false;\n\n public CloseableChannelGroup(EventExecutor executor) {\n super(executor);\n }\n\n public CloseableChannelGroup(String name, EventExecutor executor) {\n super(name, executor);\n }\n\n @Override\n public boolean add(Channel channel) {\n boolean addResult = super.add(channel);\n\n // First add channel, than check if closed\n // Seems inefficient at first, but this way a volatile\n // gives us enough synchronization to be thread-safe\n // If true: Close right away...\n // If false: Channel will definitely be closed by the group (closed=true always happens-before close();)\n if (closed) {\n channel.close();\n }\n return addResult;\n }\n\n @Override\n public ChannelGroupFuture close(ChannelMatcher matcher) {\n // It is important to set the flag to true, before closing channels\n // Our invariants are: closed=true -> closed() -> add() -> closed==true?\n closed = true;\n return super.close(matcher);\n }\n\n @Override\n public ChannelGroupFuture close() {\n // It is important to set the flag to true, before closing channels\n // Our invariants are: closed=true -> closed() -> add() -> closed==true?\n closed = true;\n return super.close();\n }\n}\n```\n\nThis should be enough. Don't know if you @normanmaurer @trustin agree with style and stuff..\n\nRegards,\nChris\n", "@chrisprobst Yeah, you've signed the ICLA on 4/30/2015 0:56:11. Thanks for the heads up.\n\nRe: The proposed change - LGTM. You are gonna incorporate the change into DefaultChannelFutureGroup, right?\n", "@trustin Hi, I'm confused. Why the DefaultChannelFutureGroup? I'm not sure how to do fix the RC there.\nIMHO, we have 2 choices. Change the DefaultChannelGroup (bad, because for compatibility), or create a new ChannelGroup based on the DefaultChannelGroup and let it be closable only once.\n\nRe: Yes, I could do the change.\n", "Ugh, I meant incorporating the change into `DefaultChannelGroup`. \n", "I certainly would, but it would change the semantics of the DefaultChannelGroup. You couldn't reuse a ChannelGroup after it is closed.\n\n_Is this expected ?_\n", "If that's the only way to solve the race condition, I think that's fine. To be honest, I haven't thought people would reuse it.\n", "IMHO its the only option, if you want to be lock-free. Something like a reset button would definitely require locks. And a locked channel group... meh. We could add a specification parameter at construction time like (stay closed) other similar.\n\nI will prepare a pull request with the changes (including doc).\n" ]

[]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/group/DefaultChannelGroup.java" ]

[ "transport/src/main/java/io/netty/channel/group/DefaultChannelGroup.java" ]

[]

[ "timestamp(timedelta=11.0, similarity=0.8811317741985843)" ]

[ "@rkapsi - Is `PlatformDependent.throwException(null)` called from Netty code, and if so do you know where from?\n", "@Scottmitch no, not from Netty code. It was some code I wrote and one can argue it was my own fault (given the nature of that class) but I wish it'd have failed more gracefully.\n", "@rkapsi - Humm I guess at least in the case of unsafe we should ensure the language specification is honored:\n\nhttp://docs.oracle.com/javase/specs/jls/se8/html/jls-14.html#jls-14.18\n\n> If evaluation of the Expression completes normally, producing a null value, then an instance V' of class NullPointerException is created and thrown instead of null. The throw statement then completes abruptly, the reason being a throw with value V'.\n\n@trustin @normanmaurer - WDYT?\n", "I have a PR pending for this.\n", "+1. Thanks for taking care.\n" ]

[ "Maybe renaming t to cause might make the exception message more sense?\n", "Will do.\n" ]

[ "defect" ]

[ "common/src/main/java/io/netty/util/internal/PlatformDependent0.java" ]

[ "common/src/main/java/io/netty/util/internal/PlatformDependent0.java" ]

[]

[ "timestamp(timedelta=8.0, similarity=0.9799484004742384)" ]

[ "@benevans @trustin - Any thoughts?\n", "/cc @jestan \n", "@jestan - Any objections to me removing this code?\n", "@Scottmitch Please go ahead. We can't wait forever.\n", "@trustin - Will do :)\n" ]

[]

[ "defect" ]

[ "transport-sctp/src/main/java/io/netty/channel/sctp/oio/OioSctpChannel.java" ]

[ "transport-sctp/src/main/java/io/netty/channel/sctp/oio/OioSctpChannel.java" ]

[]

[ "timestamp(timedelta=19.0, similarity=0.9035731286287193)" ]

[ "@trustin should I take care?\n\n> Am 27.08.2015 um 05:11 schrieb Trustin Lee [email protected]:\n> \n> It seems like some users are observing an edge case where the WeakOrderQueue and Stack grow beyond its max capacity. Although we should fix it, there should be a way for a user to disable recycling completely so that he or she can perform with-recycler vs. without-recycler comparison.\n> \n> /cc @ninja-\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Do you have spare cycles? Then please feel free to take it. :-)\n", "@trustin spare cycles not really but let me fix it anyway =P\n", "Given https://github.com/netty/netty/issues/4134#issuecomment-135484445 do we feel this is still worth it? I think we should push to diagnose / solve issues rather than disable features.\n", "@Scottmitch I think it still make sense to allow it for people that have limited memory constraints etc. \n", "@normanmaurer @Scottmitch looking at Recycler, isn't setting the size to 0 giving that effect? \n", "@ninja- - We currently don't allow the property to be set to 0 (see https://github.com/netty/netty/blob/master/common/src/main/java/io/netty/util/Recycler.java#L51). However doesn't seem like it would hurt if we did allow 0... @normanmaurer - Is this what you had in mind?\n", "@Scottmitch it's fine when set to 1 [per thread]. I think we can keep it like that. I am using it on production since today.\n\nRegarding the Recycler's logic...I think it's a wrong way to approach this problem :) the \"max number\" should be used for the case when there are _really that many objects_ used _at the same time_. And not just increasing on every recycle()! That makes no sense even for heavy http whatnot apps!\nBased on that, the cache should be allowed to grow to the _max_.\n\nI would think about an another condition to deny caching around here, https://github.com/netty/netty/blob/4.0/common/src/main/java/io/netty/util/Recycler.java#L383 @trustin \n\nFor example: let's track the balance between .get() and .recycle() :)?\n\nget: balance++\nrecycle: balance--\npush: if (size >= maxCapacity || size >= balance) return; (something like that)\n\nthat seems sane to me because I wouldn't trust current recycler with _any_ app, not only my proxy\n", "@ninja- - I don't fully understand your request. Are you asking for a way to be able to \"purge\" entries and reduce the size of the stack? Maybe submit a PR so we can get a better idea?\n", "@Scottmitch not purging but the logic should include shrinking also. I am saying that the _current logic_ doesn't suit _any_ type of the app. I proposed a new type of logic for Recycler that would cache the objects, but to a reasonable extent. \n\nI got rid of recycler by setting max to 1, tested on production, everything great, happy with it.\nI am discussing this just to improve netty.\n", "Fixed\n", "@ninja- - Please open another issue for a feature request for your \"purging\" / \"shrinking\" Recycler entries\n" ]

[ "Should we stick to this default value? /cc @normanmaurer @tea-dragon \n", "@trustin for now yes. Let us use a new pr if we want to adjust it\n" ]

[ "feature" ]

[ "common/src/main/java/io/netty/util/Recycler.java" ]

[ "common/src/main/java/io/netty/util/Recycler.java" ]

[ "common/src/test/java/io/netty/util/RecyclerTest.java" ]

[ "timestamp(timedelta=7884.0, similarity=0.8773752337812172)" ]

[]

[ "+1\n", "would we need to check if the peer.finishReadFuture == peerReadFuture and only set to null in this case ? A.k.a an atomic operation.\n", "Good catch!\n", "Please try to use `PlatformDependent.newAtomicReferenceFieldUpdater` first in a static block like we do in other places like:\n\nhttps://github.com/netty/netty/blob/master/transport/src/main/java/io/netty/channel/ChannelOutboundBuffer.java#L90\n", "Cool...wasn't aware of this.\n" ]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/local/LocalChannel.java" ]

[ "transport/src/main/java/io/netty/channel/local/LocalChannel.java" ]

[ "transport/src/test/java/io/netty/channel/local/LocalChannelTest.java" ]

[ "timestamp(timedelta=52.0, similarity=0.8607263387581197)" ]

[ "@jpinner I may miss something but I think in netty 4+ we don't need the synchronized stuff at all because of the new thread-model. Can you confirm ?\n", "@jpinner ping :)\n", "@jpinner could you please check so we can fix this... Thanks !\n", "Let me fix this\n" ]

[]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/spdy/SpdySessionHandler.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/spdy/SpdySessionHandler.java" ]

[]

[ "timestamp(timedelta=60.0, similarity=0.8817071786875241)" ]

[]

[ "I think this a copy and paste error. \n\nTCP_USER_TIMEOUT.\n", "Remoce extra empty line\n", "Remove extra empty line\n" ]

[ "feature" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollSocketChannelConfig.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_epoll_Native.c", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollSocketChannelConfig.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/Native.java" ]

[]

[ "timestamp(timedelta=37.0, similarity=0.958631723650472)" ]

[ "@blucas - Sorry for breaking you :(. I guess we have at least 2 options:\n1. Add a constructor to `DefaultSpdyHeaders` which takes a `boolean validateHeaders` argument which can make header name validation optional. We would then need to carry this through the SPDY codec to pass along the `boolean` where ever the `DefaultSpdyHeaders` are created.\n2. Your suggestion, plus a `boolean convertHeaderNamesToLowerCase` (or something similar) to control whether conversion should be done or not.\n\nFeel free to submit a PR for one/both of the above :)\n", "@Scottmitch - no problem, it happens to all of us :).\n\nI have made the necessary changes for 1. above (although I've followed the way it was done for HTTP such that if `validateHeaders` is set to `true` (default), header names **and** values will be validated)\n\nDo we really need to add a `boolean convertHeaderNamesToLowerCase` to `SpdyHttpEncoder`? We don't do this for HTTP to HTTP/2 conversion (see [here](https://github.com/netty/netty/blob/master/codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java#L318)). I think we should just do the same. What do you think?\n", "> Do we really need to add a boolean convertHeaderNamesToLowerCase to SpdyHttpEncoder? \n\nIt is not strictly necessary. Just a potential fast path or optimization if your environment is controlled. Feel free to leave it out...we can always add it later.\n", "Fixed by https://github.com/netty/netty/pull/4190\n", "Just a heads up: SpdyHttpHeaders.Names are still uppercase, is that correct?\n", "@CodingFabian - No that is a bug. Fixed with https://github.com/netty/netty/pull/4226\n" ]

[]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/spdy/SpdyHttpEncoder.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/spdy/SpdyHttpEncoder.java" ]

[]

[ "timestamp(timedelta=11.0, similarity=0.859242328286967)" ]

[ "@ejona86 @Scottmitch @nmittler @louiscryan is this something we need to fix before 4.1.0.Final ?\n", "I don't think so. We seem to be conforming to the spec since we're checking the size on the inbound frame.\n", "@nmittler so closing ?\n", "@ejona86 thoughts on closing?\n", "wouldn't hurt to check the write size...make it harder for people to put bad stuff out on the network.\n", "@Scottmitch care to submit a pr ;) ?\n", "@normanmaurer - Will do.\n\n@nmittler - Any objections?\n", "@Scottmitch :heart_eyes: ;)\n", "pr pending\n", "@Scottmitch no objections ... thanks for taking care\n" ]

[]

[ "improvement" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java" ]

[]

[ "timestamp(timedelta=102.0, similarity=0.867344004592763)" ]

[ "@Scottmitch @nmittler can you check?\n", "fixed by https://github.com/netty/netty/pull/4245\n" ]

[]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandlerTest.java" ]

[ "timestamp(timedelta=19.0, similarity=0.867944925222094)" ]

[ "@blucas Thanks for the report!\n\nRegarding the [writeBytes method](https://github.com/netty/netty/blob/master/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java#L633), I believe the comment is wrong. I suspect it should be re-worded to something like \"The frame has data but none of it can be written\". @Scottmitch, does that sound right to you?\n", "@Scottmitch So it sounds like the problem may be that we're using an estimate of the channel writable bytes, and not actually writing until the channel is not writable. With the current code, if our write doesn't actually cause an `unwritable event`, then data will get stuck since we'll never get the `writable event` to cause us to flush. Does that sound right to you?\n", "> Regarding the writeBytes method, I believe the comment is wrong. I suspect it should be re-worded to something like \"The frame has data but none of it can be written\".\n\n+1. The same comment appears 2 times in that method, and is wrong both times.\n", "@Scottmitch I think we just need to loop while the channel is writable. WDYT?\n", "@nmittler - Looking now. The intention is to have the allocation algorithm be allowed to use the fewest amount of bytes as necessary to transition `isWritable()` to false...then `Http2ConnectionHandler` should be notified when writibiliity changes back to true and we should finish writing (or continue in this cycle until nothing left to write)... So the \"looping\" should be done by getting channel writability change notifications. \n", "@nmittler - I agree with looping until the channel is no longer writable. The issue with my above assumption was that there may be `ChannelHandlers` that manipulate the size of the data (`SslHandler`, compression, etc...) so that the number of bytes written from the flow controller is not necessarily the number of bytes written to the channel, and thus the writabilty change is not guaranteed to come unless we loop.\n", "@blucas - I have verified with https://github.com/netty/netty/pull/4227 that FireFox is able to download a 2Mb file. Before this PR I was also observing the issue you reported where the entire file would not be downloaded. Can you verify with this PR?\n", "@Scottmitch would you mind separating this change out from #4227? I think this is a separate issue and would make reviewing a bit easier. If it's too much of a pain then don't worry about it.\n", "@Scottmitch - I checked out your branch for #4227 and can confirm that it does not suffer from this issue. I assume it is because of the `do...while` loop you have added to `writePendingBytes`.\n\nI agree with @nmittler, if you could separate that change from the bulk of the work done in #4227 that would be great, as I assume it will take some time before #4227 is merged.\n", "@blucas @nmittler - Yes I will pull out the relevant pieces from PR #4227 to fix this issue.\n\n@blucas - Thanks for verifying...I'll ping you when the new PR is ready.\n", "@Scottmitch SGTM ... thanks!\n" ]

[]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[]

[ "timestamp(timedelta=14.0, similarity=0.8620227261918824)" ]

[ "...why would you even do that\n", "@willisblackburn thanks for the report. Will add a guard against this\n", "@willisblackburn - I'm assuming you are using Netty 4.0?\n" ]

[ "In 4.1 we are not consistent in that this method `returns this;` and `add` throws an exception. I think we should be consistent in how we behave across versions (and maybe how we react to these 2 situations). So do we want to throw an exception, or just `return this;`?\n", "hmm. when adding this check my thought process was that this should only be triggered by a coding error (i.e. couldn't think of a reason why one would need to do this).\n\neither way is fine for me. I don't care deeply about this :-). Should be documented either way though.\n", "+1. I lean toward only throw exception if the semantics of the interface cannot be preserved. If `return this;` preserves semantics then no need for exception.\n", "Ok so only for `add(...)` fine by me as well. Not that I think calling set with itself makes any sense though\n", "I'm leaning toward throwing an exception in both cases, since it is a programming error and you want to be explicit. Not throwing may lead to buggy application code. That said, I don't feel strongly.\n", "@Scottmitch so what ? @nmittler agree's with my original patch :)\n", "LOL ... yeah, what he said!!!! (grrrrrr)\n", "@normanmaurer @nmittler - Well now things get interesting :)\n\nI'm not strong on this either and it seems like no one is so @normanmaurer why don't you decide and just make sure we handle it consistently across versions (and documented).\n\nJust to clarify the programming error statement ... the following code (and any code that calls `set` where it may take the same object) has a programming error:\n\n``` java\nHttpHeaders a = ...;\ndoSomething(a, someMethodWhichMayReturnParamter(a));\n\nvoid doSomething(HttpHeaders a, HttpHeaders b) {\n if (something happens) {\n // Programming error because no check to a != b?\n a.set(b);\n }\n}\n```\n\nSeems strange if we could just handle it with no side effects and not require the check (both internally to `set` and in the callers of `set`)\n", "@Scottmitch yes that was the idea... \n" ]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpHeaders.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpHeaders.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/HttpHeadersTest.java" ]

[ "timestamp(timedelta=80.0, similarity=0.8921734592383369)" ]

[ "@nmittler - FYI\n", "This is also related to https://github.com/netty/netty/issues/4242\n", "@Scottmitch - Could you give an example for when this would happen? I haven't had this happen to me yet when testing HTTP/2 with Chrome/FF.\n", "+1\n\n@Scottmitch as discussed offline, the first thing we need is a new test in [DefaultHttp2RemoteFlowControllerTest](https://github.com/netty/netty/blob/master/codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java) that demonstrates this.\n", "@nmittler - On it.\n", "@blucas - This issue was really the root cause of the issue https://github.com/netty/netty/issues/4242 you reported. However the fix for https://github.com/netty/netty/pull/4258 (which is also necessary for other reasons) should mask the negative effects which may result from this issue. I think the only \"visible\" effects from this should be extra tree traversal(s) to make sure all bytes are written.\n", "Fixed by https://github.com/netty/netty/pull/4282\n" ]

[ "revert this? It was more readable with the carriage return.\n", "will do :)\n", "Don't you still need `nextTotalWeight += child.weight`?\n", "Oh you updated `stillHungry` to do that ... much better :)\n", "I already moved into `stillHungry`. Seemed like it belonged there.\n", "> Oh you updated stillHungry to do that ... much better :)\n\n+1. Good suggestion :wink: \n", "attach a link to the bug? Might be helpful in the future to provide context.\n", "Maybe something like `unstreamableParentsShouldFeedHungryChildren`?\n", "Maybe update the diagram to indicate that A, B, and C have zero bytes and D has bytes?\n", "Done\n", "Not sure we need this in the test. It should be enough to see that D is written to.\n", "sgtm\n", "How do we indicate this in the diagram? I think the `[B]` notation indicates the flow control window is collapsed?\n", "Start a new notation? Maybe something like:\n\n``` java\n 0\n / | \\\n A(0) B(0) C(0)\n |\n D(>0)\n```\n", "I added some writabilty change stuff at the last minute that actually wrecked the test. Check updated test.\n\nThe key is that we need the streamableBytes to be greater than the amount will will attempt to send. The amount we will be allowed to send is determined by this method.\n", "Sure I can invent something new :) Wasn't sure if there was an existing notation we were using.\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java" ]

[ "timestamp(timedelta=388.0, similarity=0.8750055050435257)" ]

[ "@trustin @normanmaurer - FYI\n", "Already working on it :)\n\n> Am 27.09.2015 um 07:42 schrieb Scott Mitchell [email protected]:\n> \n> @trustin @normanmaurer - FYI\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Thanks @normanmaurer !\n", "Thank you guys!\n" ]

[ "Does it make sense to keep this to be able to test against \"real\" servers during development? Enabled via some explicit command line flag?\n", "not sure... I think it is fine to just use the mock.\n", "Does it make sense to have some variation in the responses (and number of responses)?\n", "sgtm. lets keep it simple.\n", "sure I can do this\n", "Is this a PITA to add diversity for bcz the dns server doesn't really support AAAA records?\n", "nit: newline?\n", "yep... sniff :/\n", "ok...if we pull in now as is...should we create a followup issue to improve this later?\n", "yes!\n", "yep... let me think about if I maybe can also support it now with a bit work\n", "k thx. I would prefer to get it in now...but I have no problem doing it in increments.\n" ]

[ "improvement" ]

[ "pom.xml", "resolver-dns/pom.xml", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsQueryContext.java" ]

[ "pom.xml", "resolver-dns/pom.xml", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsQueryContext.java" ]

[ "resolver-dns/src/test/java/io/netty/resolver/dns/DnsNameResolverTest.java", "resolver-dns/src/test/resources/logback-test.xml" ]

[ "timestamp(timedelta=14.0, similarity=0.861430330228934)" ]

[ "@jroper sounds like a bug... let me fix it.\n", "Fixed by https://github.com/netty/netty/pull/4341\n" ]

[]

[ "defect" ]

[ "buffer/src/main/java/io/netty/buffer/DefaultByteBufHolder.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketFrame.java", "transport-sctp/src/main/java/io/netty/channel/sctp/SctpMessage.java" ]

[ "buffer/src/main/java/io/netty/buffer/DefaultByteBufHolder.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketFrame.java", "transport-sctp/src/main/java/io/netty/channel/sctp/SctpMessage.java" ]

[ "buffer/src/test/java/io/netty/buffer/DefaultByteBufHolderTest.java" ]

[ "timestamp(timedelta=31.0, similarity=0.9040483003970246)" ]

[ "Will check\n\n> Am 03.10.2015 um 13:26 schrieb scf37 [email protected]:\n> \n> There is interesting comment:\n> \n> // Maybe we could also check if we can unwrap() to access the wrapped buffer which\n> // may be an AbstractByteBuf. But this may be overkill so let us keep it simple for now.\n> But - simple HTTP server under load (like 5k req/sec) uses 38% time on String.getBytes. I use ByteBufUtil.writeUtf8 to write HTTP response (about 10Kb).\n> \n> Please consider adding that check. Additional instanceof is cheap those days and do matter in some cases.\n> \n> Netty 4.0.32.Final, ResourceLeakDetector is on SIMPLE.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@scf37 can you show me your code and how this change would help you ?\n", "Code is qute complex, but profiling via jmc states that `writeUtf8` is the bottleneck.\n\nHow 20-30% improved performance would help me? Good question. Not even sure what to answer. May be I should stick with unwrap() hack, at least it is fast.\n\n```\noverride def read(): Future[HttpContent] = {\n //reading by blocks of default socket buffer size\n val len = str.length\n\n if (readDone) {\n throw new IllegalStateException(\"Response.read() after EOF\")\n }\n\n val size = Math.min(len - off, 65536)\n\n //netty internals use size * 3 to ensure buffer for utf-8 strings.\n //Set it as expected to avoid later resizing\n val buf = PooledByteBufAllocator.DEFAULT.ioBuffer(size * 3)\n\n //ByteBufUtil.writeUtf8 has fast path for WrappedByteBuf\n //unfortunately resource leak detector wraps it in WrappedByteBuf which is not detected\n //by writeUtf8\n //so, trying to unwraaaaap\n //see https://github.com/netty/netty/issues/4313\n val fastBuf = buf match {\n case buf: AbstractByteBuf => buf\n case buf if buf.unwrap().isInstanceOf[AbstractByteBuf] => buf.unwrap()\n case buf => buf\n }\n\n if (off == 0 && size == len) {\n ByteBufUtil.writeUtf8(fastBuf, str)\n } else {\n ByteBufUtil.writeUtf8(fastBuf, new CharSequenceView(str, off, size))\n }\n\n off += size\n\n if (off == len) {\n readDone = true\n Future successful new DefaultLastHttpContent(buf)\n } else {\n Future successful new DefaultHttpContent(buf)\n }\n}\n```\n", "@scf37 ah got it! May I ask what level for the leak detector you use ?\n", "Default level. ResourceLeakDetector is on SIMPLE\n", "@scf37 thanks... working on a patch.\n", "@scf37 see #4345\n", "@normanmaurer Assuming you are asking me to review. Done.\n", "@scf37 replied to your comments\n" ]

[]

[ "improvement" ]

[ "buffer/src/main/java/io/netty/buffer/ByteBufUtil.java", "buffer/src/main/java/io/netty/buffer/WrappedByteBuf.java" ]

[ "buffer/src/main/java/io/netty/buffer/ByteBufUtil.java", "buffer/src/main/java/io/netty/buffer/WrappedByteBuf.java" ]

[ "buffer/src/test/java/io/netty/buffer/ByteBufUtilTest.java" ]

[ "timestamp(timedelta=15.0, similarity=0.8927650762146052)" ]

[ "Will check\n", "@luengnat ok I think i know where the problem is... working on a fix now :+1: \n", "Fixed\n", "Oddly I started encountering this issue while upgrading from jdk 8u60 to 8u65. Not really sure why that would explain this. I was just curious @luengnat do you happen to remember what JDK you were on when you were encountering this? We've been on 4.0.26/4.0.27 for about 6 months or and I hadn't seen this issue once until we recently upgraded JDKs in our CI environment and it seems to happen occasionally.\n" ]

[]

[ "defect" ]

[ "common/src/main/java/io/netty/util/concurrent/GlobalEventExecutor.java" ]

[ "common/src/main/java/io/netty/util/concurrent/GlobalEventExecutor.java" ]

[]

[ "timestamp(timedelta=11.0, similarity=0.8850444964418077)" ]

[ "@freels hmmm... s you would argue that either the user provides a full initialised TrustManagerFactory and not certs or none and the certs ?\n", "Yes. This would match the behavior of JdkSslServerContext as well.\n\nIn my case, I am trying to provide a trust manager initialized with a root CA cert that a server context will use to validate client certs. However, because OpenSslServerContext reinitializes the provided trust manager w/ the server's cert, it drops the root cert and trusts only its own server cert.\n", "Also, if neither a trust manager or trusted cert are provided, I'm pretty sure it should be able to use a default trust manager (w/o the call to init) in order to use the system provided trust store. (I'm looking at https://github.com/netty/netty/blob/4.0/handler/src/main/java/io/netty/handler/ssl/OpenSslServerContext.java#L370-L380 )\n", "Alright... Looking or if you like you could also do a pr ;) ?\n\n> Am 14.10.2015 um 17:07 schrieb Matt Freels [email protected]:\n> \n> Also, if neither a trust manager or trusted cert are provided, I'm pretty sure it should be able to use a default trust manager (w/o the call to init) in order to use the system provided trust store. (I'm looking at https://github.com/netty/netty/blob/4.0/handler/src/main/java/io/netty/handler/ssl/OpenSslServerContext.java#L370-L380 )\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "I'd be happy to provide a PR sometime in the next few days... just wanted to make sure this was something to fix before spending more time on it. :)\n", "If it is how it is done in Jdk one yes :)\n\n> Am 14.10.2015 um 17:13 schrieb Matt Freels [email protected]:\n> \n> I'd be happy to provide a PR sometime in the next few days... just wanted to make sure this was something to fix before spending more time on it. :)\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@freels nevermind... working on a fix, stay tuned :)\n", "Awesome, thanks!\n", "@freels please check https://github.com/netty/netty/pull/4387\n" ]

[]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslServerContext.java" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslServerContext.java" ]

[]

[ "timestamp(timedelta=112.0, similarity=0.9327689825970706)" ]

[ "I guess all the native transport related stuff should only be built/triggered when a Linux only maven profile is enabled.\n", "yes, that's also how the pom file seems to be set up, but idk how this \"conditional dependency\" stuff works in maven.\n", "aww i think I have a fix.\n", "@buchgr @normanmaurer CI build was successful and I just squashed the 2 commits.\n", "Fixed\n" ]

[ "this is just cause I also got warnings, that referring to props without a `project` prefix is deprecated.\n" ]

[]

[ "microbench/pom.xml", "pom.xml" ]

[ "microbench/pom.xml", "pom.xml" ]

[]

[ "timestamp(timedelta=26.0, similarity=0.8603650312112185)" ]

[ "thanks for reporting @maxwindiff!\n", "@Scottmitch Does this also fix https://github.com/netty/netty/issues/3450?\n", "@buchgr yes I think so.\n", "@buchgr - Based upon your description (same scenario that is described in this issue), I would think this should take care of #3450. Is it possible to verify?\n", "@Scottmitch I don't have the code anymore, but yeah it looks like it fixes both issues. I think it's a perfectly valid fix as it makes it work with all Netty provided Executors. However, it doesn't work with Guava's `DirectExecutor` or Developer Joes `SelfBakedDirectExecutor` :). I couldn't think of fix for those scenarios myself either ... \n", "@buchgr - I also couldn't think of a good way to generalize the behavior of `the executor will always (or on the next execute call) use the same thread to execute tasks.` ... and just punted for the near term :(\n" ]

[ "Static?\n", "done\n", "just a nit but do we want to also prefix the new methods with \"test\" as the others in the class ?\n", "done\n" ]

[ "duplicate" ]

[ "common/src/main/java/io/netty/util/concurrent/DefaultPromise.java" ]

[ "common/src/main/java/io/netty/util/concurrent/DefaultPromise.java" ]

[ "common/src/test/java/io/netty/util/concurrent/DefaultPromiseTest.java" ]

[ "timestamp(timedelta=432.0, similarity=0.8530380492232348)" ]

[ "Thanks for reporting @ninja- !\n\n@normanmaurer - assigned to you.\n", "@Scottmitch :smiley: I am _squeezing_ every last bit of top hot methods and found lots of interesting things already.\nbtw. recycler.get() was something like 6% for the work and 94% for the thread local madness...I'd use something separate for write-once locals like recycler stack references. array[currentThread.getId()] is what I am going to try first. It's not that simple to do considering that thread ids could grow biggg. But it might be worth hacking hottest references like recycler stacks into FastThreadLocalThread as fields. But then it's not easy to get the fields out of FastThreadLocalThread as instanceof+cast isn't acceptable.\n\nInstead of removing fast thread local I would rather try to fix it's problems even if it's hard.\n(and then I would inject my recycler stuff to UnpaddedInternalThreadLocalMap...)\n", "I have a commit that removes the \"fast\" internals(but keeps the \"fast\" threads, they may be useful later) while keeping the API and wrapping around java's ThreadLocal...can PR.\nThe thread local map is also kept for internal usecases just uses java's ThreadLocal for get and set.\n\n```\nBenchmark Mode Cnt Score Error Units\nFastThreadLocalBenchmark.fastThreadLocal thrpt 4 1704783,743 ± 376972,859 ops/s\nFastThreadLocalBenchmark.jdkThreadLocalGet thrpt 4 1596349,202 ± 1182866,832 ops/s\n\n^ on par (modified benchmark a bit, it was taking too long)\n```\n\n```\nBenchmark Mode Cnt Score Error Units\nFastThreadLocalBenchmark.fastThreadLocal thrpt 20 66182,502 ± 2268,703 ops/s\nFastThreadLocalBenchmark.jdkThreadLocalGet thrpt 20 64619,961 ± 4825,228 ops/s\n\n^ original benchmark, modified netty\n```\n\nBack to Recycler now I am considering removing per-thread stacks now that that my version is more like a linked list...wdyt @normanmaurer ? Would save most of these ThreadLocal.get(). The cost of volatile should still be cheaper than threadlocal.get()\nLet's see what happens if I remove thread local recyclers, make a shared one and then implement an optimistic linked queue...playing with it right now.\n", "@ninja- please hold of for now... I want to investigate a bit more \n", "@ninja- \ncould you run it on Java 7 as well please? Maybe it's just a change in the runtime that makes `FastThreadLocalBenchmark` regress. Also looking at the commit 7d37af5dfb33c9ca67c596, it seems like this change was justified due to improved performance in `ByteBufAllocatorBenchmark` and `RecyclableArrayListBenchmark`. So it would be interesting to see if that's still true for those two benchmarks. \n\nAlso, your benchmark numbers seem to be very different with each run. Do you know why?\n", "@buchgr Regarding byte buf allocator I have some other (quite simple) ideas so I'll test it as a whole\n(recycler/threadlocal.get())\n@buchgr are you referring to that millions-something score benchmark? I left a notice under it that it's a version that uses less iterations to test. Below is the non-modified benchmark\n", "@ninja- please if you do any benchmarks do it with \"isolated\" changes and not change multiple things at the same time. \n", "@normanmaurer @buchgr's question is easy to test: either these two benchmarks with jdk's threadlocal will regress or not. So that's not a problem to test too\n", "> @buchgr are you referring to that millions-something score benchmark? I left a notice under it that it's a version that uses less iterations to test. Below is the non-modified benchmark\n\nlooking at your first post, numbers seem significantly different between the runs.\n", "@buchgr pc load I think. It would be 2x faster in every attempt to test \n", "@ninja- can reproduce it... looking :)\n", "@normanmaurer I have a patch that removes thread-local recycler stack lookups(lookup once) in places where the object is either thread local already (like the PoolThreadCache) or called only inside the event loop so: PoolThreadCache(memory regions),PooledByteBufAllocator(pooled buffer objects),ChannelOutboundBuffer(outbound buffer entries). No chance to break anything and benchmarks should like this change(I'll test) and doesn't depend on other changes. Thread locals would never be free as long as Thread.currentThread() is there.\n", "Like I said please hold off\n\n> Am 29.10.2015 um 12:35 schrieb ninja [email protected]:\n> \n> @normanmaurer I have a patch that removes thread-local recycler stack lookups(lookup once) in places where the object is either thread local already (like the PoolThreadCache) or called only inside the event loop so: PoolThreadCache(memory regions),PooledByteBufAllocator(pooled buffer objects),ChannelOutboundBuffer(outbound buffer entries). No chance to break anything and benchmarks should like this change(I'll test) and doesn't depend on other changes. Thread locals would never be free as long as Thread.currentThread() is there.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@normanmaurer ...it's barely related\n" ]

[]

[]

[ "microbench/src/main/java/io/netty/microbench/util/AbstractMicrobenchmark.java" ]

[ "microbench/src/main/java/io/netty/microbench/util/AbstractMicrobenchmark.java" ]

[]

[ "timestamp(timedelta=95230.0, similarity=0.8515996131608482)" ]

[ "@louiscryan - I think we can close this now?\n" ]

[]

[]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpHeaders.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpHeaders.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpHeaders.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpHeaders.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/HttpHeadersTest.java" ]

[ "timestamp(timedelta=20815.0, similarity=0.9127999973381334)" ]

[ "@louiscryan can be closed ?\n", "@louiscryan - Thanks for the fix!\n" ]

[]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Headers.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Headers.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2HeadersTest.java" ]

[ "timestamp(timedelta=20987.0, similarity=0.8665203016177547)" ]

[ "@louiscryan I think this can be closed ?\n", "@louiscryan - thanks!\n" ]

[]

[]

[ "codec/src/main/java/io/netty/handler/codec/DefaultHeaders.java" ]

[ "codec/src/main/java/io/netty/handler/codec/DefaultHeaders.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/HttpHeadersTest.java", "codec/src/test/java/io/netty/handler/codec/DefaultHeadersTest.java" ]

[ "timestamp(timedelta=22.0, similarity=0.873253803792159)" ]

[ "I didn't expect people to produce such amount of load obviously. Let me get the query ID allocation done on per-server basis rather than on per-resolver basis.\n", "Thanks. If you want to keep an array structure, perhaps size could be a configurable setting? In either case it should be possible to get the originally submitted DnsQuestion in the listener.\n", "I think this was fixed by @trustin ... @trustin WDYT ?\n", "Not fixed yet\n\n## \n\nhttps://twitter.com/trustin\nhttps://twitter.com/trustin_ko\nhttps://twitter.com/netty_project\n\nOn Wed, Sep 9, 2015, at 09:42 PM, Norman Maurer wrote:\n\n> I think this was fixed by @trustin[1] ... @trustin[2] WDYT ?\n> \n> —\n> Reply to this email directly or view it on GitHub[3]. \n\nLinks:\n1. https://github.com/trustin\n2. https://github.com/trustin\n3. https://github.com/netty/netty/issues/3972#issuecomment-138896842\n", "DNS txid space is limited to 0-65535, like @trustin suggested we can use a txid pool per server, but maybe we could do a txid pool per source port ?\n\nI think a dns query identifier is `source port` + `transaction id` we can send the same transaction id to a same server, but using a different source port. This way we have a far bigger query id space.\n", "I'd support any solution that would extend the query ID space...\n", "Busy at work - will fix when I have time. PR welcome.\n", "Fixed by #4455\n" ]

[ "We should not do this as this allows this to escape from the constructor and so we may see not fully initialise instances .\n", "can't we use a ConcurrentMap here ?\n", "We can't because we insert multiple entries.\n", "I guess that's not gonna happen because `DnsQueryContext` is only accessed later when a matching DNS response is received.\n" ]

[ "defect" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsQueryContext.java" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsQueryContext.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsQueryContextManager.java" ]

[]

[ "timestamp(timedelta=30.0, similarity=0.9127479533480994)" ]

[ "No this is a bug\n\n> Am 08.11.2015 um 22:03 schrieb Brendt [email protected]:\n> \n> Netty Version: master (latest snapshot a6816bd)\n> \n> I've notice that if a DefaultFullHttpRequest or DefaultFullHttpResponse have been released, then calls to hashCode() will throw an IllegalRefCountException. I was wondering if this is considered the correct behaviour?\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Is it just a matter of returning 0 if the content has been freed? If so I could provide a PR for it. I wonder what would happen to lookups for entries in a hash-based container (such as a HashMap<DefaultFullHttpRequest, SomeObject>)?\n", "@blucas - Semantics would change (the lookup may fail the second time) if we use an arbitrary constant if the content has been released. Perhaps we should cache some (or all) of the hash code that is considered \"imitable\". I can submit a PR to cleanup.\n", "@Scottmitch @blucas not sure what is the best way to fix this but I think we should not throw for equals(....), toString() and hashCode().\n", "@Scottmitch - To give you some background, I discovered this issue when attempting to clear out data stored in a `DefaultHttpDataFactory` for a given request which had already been freed. It threw the `IllegalRefCountException` when attempting to lookup a value which was keyed by `DefaultFullHttpRequest` [here](https://github.com/netty/netty/blob/master/codec-http/src/main/java/io/netty/handler/codec/http/multipart/DefaultHttpDataFactory.java#L250). Maybe it isn't a good idea to use `FullHttpRequest` or `FullHttpResponse` objects as keys in maps?\n\n@normanmaurer - I agree, we shouldn't throw, regardless of what the fix is.\n", "@blucas - In general the hashCode and equals were added for testing purposes. However it is not unreasonable to use these items as keys in a map. For example maybe a cache may use a request as a key into a map. You have to be aware of with the current implementation is that the content is used in the hash code generation, and it may be an expensive operation.\n\nWe can discuss the `DefaultHttpDataFactory` details in another issue if that is OK?\n", "See https://github.com/netty/netty/pull/4463\n", "@Scottmitch - thanks for the quick PR. With regards to `DefaultHttpDataFactory` I was just curious. No need to raise another issue unless you think it warrants one.\n", "@blucas - If you are happy then I am happy :). Honestly I would need to look into `DefaultHttpDataFactory` in more detail to see if it is necessary to use the entire `Full` message or not.\n", "Fixed by https://github.com/netty/netty/pull/4463. Thanks for reporting @blucas !\n" ]

[ "nit: `e` -> `ignore`\n", "nit: `e` -> `ignore`\n" ]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpResponse.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpResponse.java" ]

[]

[ "timestamp(timedelta=175.0, similarity=0.9108850013441069)" ]

[ "Will check soon\n\n> Am 18.11.2015 um 14:01 schrieb Stephane Landelle [email protected]:\n> \n> Netty 4.0.33 (client)\n> Jetty 9.3 (server)\n> OSX 10.11.1\n> Hotspot 1.8.0_66\n> AUTO_READ is set to false, read is manually triggered from channelActive and readComplete \n> Hi,\n> \n> I ran into an issue while trying to upgrade AsyncHttpClient test suite to Jetty 9.3, see AsyncHttpClient/async-http-client#1035.\n> \n> The test is about sending a small ChunkedStream to a Jetty server expecting Basic auth, without providing the Authorization header. In this case, Jetty will respond with 401 and then close the socket as soon as it receives the headers and realize Authorization is missing.\n> \n> It seems that Netty (or JDK NIO?) randomly fails to read the HTTP response and directly jumps to closing the socket.\n> It checked with Wireshark and Jetty indeed sends the expected 401 response.\n> \n> Please have a look at this reproducer (client is pure Netty, not AHC): https://github.com/slandelle/netty-missing-events-issue\n> \n> Note that running from mvn test from the terminal seems to increase the failure probability.\n> \n> Regards\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@normanmaurer Thanks!\n\nNote: I removed the auto_read disabling comment and updated the code as I can reproduce without.\n", "@slandelle - I took a look at the Netty unit test you linked above. Let me describe the behavior I am seeing and you can confirm if this is what you are seeing:\n\n`testUnauthenticatedGetThenChunkedPostWtf` test behavior:\n1. Client sends GET request\n2. Server responds with 401 (with some body)\n3. Client reads 401.\n 3.1 MyHandler.read() called with `DefaultHttpResponse`\n 3.2 Myhandler.read() called with `DefaultLastHttpContent` (319 bytes)\n4. Client sends POST request (chunked encoding) `4\\r\\ntest\\r\\r\\n`)\n5. Client detects that channel is inactive, `MyHandler.channelInactive` prints out `Channel became inactive before reading anything!!!`.\n\nAt some point after 4 the server closes the connection. The code that initiates step 4 above is in a listener when the connection is established, and a flush operation (`channel.writeAndFlush(LastHttpContent.EMPTY_LAST_CONTENT)`) fails with a `java.io.IOException: Broken pipe` in [AbstractChannel](https://github.com/netty/netty/blob/4.0/transport/src/main/java/io/netty/channel/AbstractChannel.java#L776).\n\nI'll need to think about what can be done at this point. Can we close just the outbound portion of the channel? Should we be completing the close promise before we attempt a read? @normanmaurer - Does this make sense?\n", "@Scottmitch just so I understand it correctly... Jetty closes the channel before we are done writing ?\n", "@Scottmitch @normanmaurer Thanks for taking a look.\n\nI've just pushed a new commit on the issue demo, to make sure I don't try to recreate a Jetty server.\n\nBeware that every request creates a new client instance with its own event loop and bootstrap.\n\nI should have been more explicit:\n- `testUnauthenticatedGet` and `testUnauthenticatedChunkedPost` are only there to show that, taken separately, those are more likely to succeed.\n- `testUnauthenticatedGetThenChunkedPostWtf` basically runs the previous two, but for some reason, is more likely to fail.\n\nSo, regarding your sequence, please note that [1, 3] happen in a first (event loop, bootstrap) while [4, 5] happen in a second one.\n\nTo be more specific:\n1. First client instance opens new Channel\n2. First client sends GET request\n3. Server responds with 401 (with some body) and closes the socket\n4. First client reads 401.\n4.1 First client: MyHandler.read() called with DefaultHttpResponse\n4.2 First client: Myhandler.read() called with DefaultLastHttpContent (319 bytes)\n5. First client gets closed\n6. Second client instance opens new Channel\n7. Second client sends POST request (chunked encoding) 4\\r\\ntest\\r\\r\\n)\n8. Server responds with 401 (with some body) and closes the socket\n9. Client detects that channel is inactive, MyHandler.channelInactive prints out Channel became inactive before reading anything!!!\n10. Test fails with:\n\n```\ntestUnauthenticatedGetThenChunkedPostWtf(io.gatling.issues.MissingEventsTest) Time elapsed: 0.015 sec <<< FAILURE!\njava.lang.AssertionError: Should have been able to send request and read response\n at org.junit.Assert.fail(Assert.java:88)\n at org.junit.Assert.assertTrue(Assert.java:41)\n at io.gatling.issues.MissingEventsTest.doTest(MissingEventsTest.java:37)\n at io.gatling.issues.MissingEventsTest.testUnauthenticatedGetThenChunkedPostWtf(MissingEventsTest.java:54)\n at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)\n at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)\n at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)\n at java.lang.reflect.Method.invoke(Method.java:497)\n at org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:50)\n at org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12)\n at org.junit.runners.model.FrameworkMethod.invokeExplosively(FrameworkMethod.java:47)\n at org.junit.internal.runners.statements.InvokeMethod.evaluate(InvokeMethod.java:17)\n at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:325)\n at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:78)\n at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:57)\n at org.junit.runners.ParentRunner$3.run(ParentRunner.java:290)\n at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:71)\n at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:288)\n at org.junit.runners.ParentRunner.access$000(ParentRunner.java:58)\n at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:268)\n at org.junit.internal.runners.statements.RunBefores.evaluate(RunBefores.java:26)\n at org.junit.internal.runners.statements.RunAfters.evaluate(RunAfters.java:27)\n at org.junit.runners.ParentRunner.run(ParentRunner.java:363)\n at org.apache.maven.surefire.junit4.JUnit4Provider.execute(JUnit4Provider.java:252)\n at org.apache.maven.surefire.junit4.JUnit4Provider.executeTestSet(JUnit4Provider.java:141)\n at org.apache.maven.surefire.junit4.JUnit4Provider.invoke(JUnit4Provider.java:112)\n at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)\n at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)\n at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)\n at java.lang.reflect.Method.invoke(Method.java:497)\n at org.apache.maven.surefire.util.ReflectionUtils.invokeMethodWithArray(ReflectionUtils.java:189)\n at org.apache.maven.surefire.booter.ProviderFactory$ProviderProxy.invoke(ProviderFactory.java:165)\n at org.apache.maven.surefire.booter.ProviderFactory.invokeProvider(ProviderFactory.java:85)\n at org.apache.maven.surefire.booter.ForkedBooter.runSuitesInProcess(ForkedBooter.java:115)\n at org.apache.maven.surefire.booter.ForkedBooter.main(ForkedBooter.java:75)\n```\n\n> Note: Failure is very random. Test can pass 10 times, then fail 3 times in a row.\n", "@slandelle - Yip. The details you added are the same I'm seeing on my end.\n\nCan you add `.option(ChannelOption.AUTO_CLOSE, false);` to your bootstrap and let me know if this gives the expected behavior? \n", "@slandelle - Also I know this is just an example but I also changed your `channelRead` to make sure there was no premature close.\n\n``` java\n@Override\npublic void channelRead(ChannelHandlerContext ctx, Object msg) throws Exception {\n read = true;\n if (msg instanceof LastHttpContent) {\n ctx.close();\n }\n}\n```\n", "@Scottmitch Bull's eye, explicitly setting `AUTO_CLOSE` to false did the trick! I run 30 times without being able to reproduce.\n\nThen, I don't get it. `AUTO_CLOSE`'s javadoc says `false` is the default value.\n", "@slandelle - Thanks for pointing out the javadoc error. The \"real\" default is [true](https://github.com/netty/netty/blob/4.0/transport/src/main/java/io/netty/channel/DefaultChannelConfig.java#L62) :wink:. Let me update the docs so they are not misleading. We can debate what the default should be as part of another issue if necessary.\n", "See https://github.com/netty/netty/pull/4510\n", "@slandelle - I'll close this for now. Please re-open (or open another issue) if the issue pops back up.\n", "@Scottmitch Isn't there anything to do to respect temporality?\nI mean that the response was received BEFORE the socket got closed and the next chunk write failed.\n\nThen, what's the behaviour when AUTO_READ is disabled? I mean that it doesn't seem that `exceptionCaught` is notified.\n\nAnd, I'd really like to hear your opinion on what the default should be. What's the impact if I make AUTO_READ the default in AsyncHttpClient?\n", "> I mean that the response was received BEFORE the socket got closed and the next chunk write failed.....I mean that it doesn't seem that exceptionCaught is notified.\n\nFor \"outbound\" operations generally the promise associated with that operation is failed and `fireExceptionCaught` is not called. This behavior was recently \"clarified\" in PR https://github.com/netty/netty/pull/4498.\n\n> Then, what's the behaviour when AUTO_READ is disabled?\n\nI wouldn't expect the behavior to change, other than you must explicitly call `read()`. When you call `read()` you should get the data, and if the socket is \"not suitable for reading\" anymore when you call `read()` you should get a `channelInactive()` (and maybe an `exceptionCaught`).\n\n> And, I'd really like to hear your opinion on what the default should be.\n\nI think the default from Netty's perspective (in 4.0 and 4.1) for `AUTO_CLOSE` has to remain `true` to avoid changing the existing behavior (established [a while ago](https://github.com/netty/netty/blame/4.0/transport/src/main/java/io/netty/channel/DefaultChannelConfig.java#L62)).\n\n> What's the impact if I make AUTO_READ the default in AsyncHttpClient?\n\n(caveat - I'm not familiar with the details of AHC.) Generally turning AUTO_READ off just allows the user of Netty to determine when `read()` is called. Can explain what your are trying to do by turning auto_read off and what your trigger points will be for reading?\n", "Sorry, late here: I meant AUTO_CLOSE, not AUTO_READ...\n", "@slandelle - Thanks for clarifying. Changing `AUTO_CLOSE` to false will just mean that failures to write will not automatically close the transport (for situations like these). This can be useful for \"half closed\" situations like the one uncovered in your unit test. I would imagine AHC would want `AUTO_CLOSE` set to `false` to handle this type of scenario and avoid potentially losing data.\n", "@Scottmitch But does this mean that I then am in charge of closing the channel?\n", "@slandelle - As a result of write failures yes. However you will still get notified if problems are detected on the read side.\n", "@Scottmitch Thanks a lot for your explanations\n" ]

[]

[ "cleanup" ]

[ "transport/src/main/java/io/netty/channel/ChannelOption.java" ]

[ "transport/src/main/java/io/netty/channel/ChannelOption.java" ]

[]

[ "timestamp(timedelta=4.0, similarity=0.8464062515270071)" ]

[ "I guess the root cause of this issue is at `HttpClientCodec.upgradeFrom()`, which removes the `Decoder` immediately. Perhaps we should remove it using `ctx.executor().execute()`?\n", "To add a bit more, I'm using 101 to upgrade from HTTP/1 to HTTP/2. I sometimes get an `IllegalReferenceCountException`, a `First received frame was not SETTINGS.` error, or both. After removing the decoder using `ctx.executor().execute()`, the problem(s) went away.\n\nThe 'first received frame was not SETTINGS' error seems to be because the `channelRead` event that contains the first `SETTINGS` frame is produced _before_ `Http2ConnectionHandler` is added and thus lost at the end of the pipeline.\n", "@trustin sounds like a valid fix. \n", "sounds reasonable to me too\n", "@trustin sgtm\n", "@trustin will you submit a PR for this ?\n" ]

[]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpClientCodec.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpClientCodec.java" ]

[]

[ "timestamp(timedelta=14.0, similarity=0.8718083460186403)" ]

[ "Seems like we should use `HttpHeaderValues` here. The specification calls out the lowercase variant [1]. Should we change `HttpHeaderValues.UPGRADE` to lowercase, and should/are we using case insensitive compare when evaluating this header value?\n\n[1] https://tools.ietf.org/html/rfc7230#section-6.7\n\n> When Upgrade is sent, the sender MUST also send a Connection header\n> field (Section 6.1) that contains an \"upgrade\" connection option\n", "We are using case-insensitive comparison, which is fine in my opinion. (more tolerating)\n\nJetty uses case-sensitive comparison and expects `Upgrade`, which seems incorrect. (They just do `.contains()` on a `List<String>` IIRC.)\n\nSo, we could:\n- Use `HttpHeaderValues.UPGRADE` instead,\n- Change it from `Upgrade` to `upgrade`, and\n- Ping the Jetty folks about this so they fix this or they have a chance to disagree.\n", "Here's the code that uses `.contains(\"Upgrade\")`: \n\nhttps://github.com/eclipse/jetty.project/blob/65d33f8b49bd51140d9fe9b024c5dd42ad7571d8/jetty-server/src/main/java/org/eclipse/jetty/server/HttpChannelOverHttp.java#L373\n\n`git blame` tells @gregw could help us here.\n", "@trustin Ooops that looks wrong. Fixing now..... standby.....\n", "@trustin Thanks for confirming we use case-insensitive compare. I agree we should change `HttpHeaderValues.UPGRADE` to lowercase, and use it here.\n\n@gregw - Thanks for the fix!\n", "@trustin @Scottmitch Fix is committed to jetty 9.3 and master branches: https://bugs.eclipse.org/bugs/show_bug.cgi?id=483039\n", "@gregw thanks!\n\n@Scottmitch should I take care ?\n", "@normanmaurer - Sure. Thanks!\n", "Ok working on a fix\n" ]

[]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpClientUpgradeHandler.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpHeaderValues.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpClientUpgradeHandler.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpHeaderValues.java" ]

[]

[ "timestamp(timedelta=110249.0, similarity=0.8467921632658658)" ]

[ "@louiscryan - Let me summarize the face value tradeoffs and follow up with some questions. Feel free to elaborate. It seems like goal is to potentially reduce the amount of data frames written (less data frame header overhead on the wire) at the cost of an extra iteration step before writing out data frames (until a different frame type, or a frame for a different stream is encountered).\n\nQuestions:\n- What does \"small\" mean? How many \"small\" frames should we combine? Is this based upon network conditions (i.e. is \"small\" for a cell network the same as for corp network), and/or defined by the application?\n- Seems like a goal is the peer will potentially generate less WINDOW_UPDATE frames to replenish bytes to flow control (as stated in #3737)? I guess the \"small\" qualification is to avoid the risk of sacrificing responsiveness for some use-cases where a WINDOW_UPDATE could be sent more immediately when we don't combine?\n- Would this require some sort of \"look ahead\" step in the flow controller to determine \"if we combine this next frame, would that exceed my allotted allocation\"? \n\n> In this case we can address by simply draining the flow-controller for the relevant stream and then writing the RST.\n\nDoes this mean we need to potentially be able to queue every frame type in the flow control queues? For example what if we can't \"drain\" the queues because we are blocked due to flow control constraints?\n", "Thats a fair summary of the goal with the associated benefit of larger DATA frame reads on the receiving side. \n\nI don't know that we necessarily need to 'flush' DATA for a stream if a different frame type or stream output is encountered. Enqueueing data should consume flow-control which would prevent ordering problems in the consumption of flow-control on eventual write. In the case of RST_STREAM we already discard any pending DATA waiting for flow-control so discarding any pending DATA waiting for flush would not be a significant semantic shift. The processing of a WINDOW_UPDATE after a write should similarly be consumed by the already enqueued but not written data.\n- I don't think we need to define 'small'. We don't actually commit anything to the wire without flushing the channel anyway.\n- Responsiveness is really dictated by how often the application calls 'flush' which I believe is the desired behavior.\n- I think this is moot given that we don't need to define small.\n\nSee note above about ordering semantics of RST, I'm not sure we actually have an issue. A RST would cause the stream to become inactive and any pending-consumed window could be returned without actually writing the bytes.\n", "I should point out that a perfectly reasonable option is that we don't want to do any of this and applications are responsible for making their writes chunky.\n\nSee parallel discussion in GRPC about doing just that\nhttps://github.com/grpc/grpc-java/issues/502\n", "I think it may make sense. Just a bunch of things to think about :) For example does this negatively impact responsiveness (see below), is it possible users would want the different data frames (different application layer semantics), etc... I can't think of anything but seems like you have thought about this a bit more than I have. \n\nDoes this negatively impact a lossy channel (i.e. cellular)? For lossy channels it may be more likely to get the \"smaller\" chunks up to the application in a timely fashion. However I guess having less overall data (less frames, less window updates) may be better in a lossy environment? Maybe these endpoints are already limiting their flow control window sizes as well?\n", "@louiscryan looking at the [current API](https://github.com/netty/netty/blob/master/codec-http2/src/main/java/io/netty/handler/codec/http2/Http2RemoteFlowController.java#L44), this would be a fairly significant redesign, since `FlowControllable`s would need to support merging. This gets a bit hairier when considering that `HEADERS` are `FlowControllable` too, and we wouldn't support merging them with other headers or with data frames. It certainly possible to do this, but I wonder what the resulting API would look like. Any thoughts?\n", "Something like\n\nboolean merge(FlowControllable next)\n\nif return is true then they were merged successfully, if false they werent and 'next' should be added to the queue\n", "@louiscryan right, that's pretty much what my initial thought was as well. Were you thinking of cooking up a PR to see what this would look like?\n", "I started hacking on something, primarily so I can compare the performance of implicit merging vs explicit merging at the application layer. If implicit is close then it's probably a good idea, if it's not then perhaps explicit application merging is the recommendation and Netty remains unchanged other than adding doc.\n", "SGTM\n", "Just an update on this. I have most of this working but at the moment it performs worse than before in GRPC. This is largely due to the fact that for each DATA write we attach a listener to the future and use that to return flow-control to the application. This causes flow-control to be returned to a stream for each small message rather than as a block for the coalesced write.\n\nTo fix this we need to be notified of how much data the flow-controller wrote when it was flushed. Ill look into that tomorrow.\n", "OK. Fixed performance issues by allowing for more granular notifications out of the flow-controller.\n\nIn GRPC I see about a 2x throughput improvement for streams of small messages.\n", "@Scottmitch @ejona86 @nmittler \n\nAn update on whether to support window update coalescing.... \nIn short I'm inclined not to make the change as it only seems to help when the flow-control window is improperly sized to the network.\n\nI've started using netem with our benchmarks so we can better simulate real network behaviors over loopback. For 1gb links we can saturate them pretty easily, for 10gb links we currently get to about 60% which isn't too shabby but we'd like to do better.\n\n1gb link, 200us latency\nsudo tc qdisc change dev lo root netem delay 0.2ms rate 1gbit\n\nWith WU coalescing 4.1B6\n\n```\nBenchmark (clientInboundFlowWindow) (maxConcurrentStreams) (responseSize) Mode Cnt Score Error Units\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond MEDIUM 1 LARGE thrpt 100 404.407 ± 3.885 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond LARGE 1 LARGE thrpt 100 889.155 ± 2.494 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond JUMBO 1 LARGE thrpt 100 949.817 ± 0.400 ops/s\n```\n\nHead\n\n```\nBenchmark (clientInboundFlowWindow) (maxConcurrentStreams) (responseSize) Mode Cnt Score Error Units\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond MEDIUM 1 LARGE thrpt 100 398.030 ± 3.349 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond LARGE 1 LARGE thrpt 100 869.634 ± 2.837 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond JUMBO 1 LARGE thrpt 100 950.606 ± 0.270 ops/s\n```\n\n10gb link\nsudo tc qdisc change dev lo root netem delay 0.2ms rate 10gbit\n\nWith WU coalescing 4.1B6\n\n```\nBenchmark (clientInboundFlowWindow) (maxConcurrentStreams) (responseSize) Mode Cnt Score Error Units\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond MEDIUM 1 LARGE thrpt 100 653.040 ± 4.803 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond LARGE 1 LARGE thrpt 100 5337.066 ± 49.104 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond JUMBO 1 LARGE thrpt 100 4277.931 ± 711.438 ops/s\n```\n\nMaster\n\n```\nBenchmark (clientInboundFlowWindow) (maxConcurrentStreams) (responseSize) Mode Cnt Score Error Units\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond MEDIUM 1 LARGE thrpt 30 489.693 ± 6.204 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond LARGE 1 LARGE thrpt 30 5916.676 ± 173.040 ops/s\nStreamingResponseBandwidthBenchmark.stream:megabitsPerSecond JUMBO 1 LARGE thrpt 30 5048.846 ± 1229.370 ops/s\n```\n\n@normanmaurer - Random aside, have you seen much use of netem with Netty before? I wonder if it would be useful to specify a standard 'network' configuration that folks can use to run benchmarks with to improve comparison across environments. E..g setup a tuntap network with a well known name, apply netem to it and then detect its presence in the benchmark using java.net.NetworkInterface\n", "@louiscryan - Thanks for the update. It seemed the PR which coalesced data was assuming this change would rectify some of the degradation it introduced for larger payload sizes. Do we need to take any follow-up action to correct this?\n", "No, the degradation for larger payload was caused by using a payload\nByteBuf that was too large to be allocated out of the pool. I don't think\nthere is any followup specific to HTTP2 here. The appropraite fix has been\nmade in the GRPC benchmarks https://github.com/grpc/grpc-java/pull/563\n\nOn Thu, Jun 25, 2015 at 12:08 PM, Scott Mitchell [email protected]\nwrote:\n\n> @louiscryan https://github.com/louiscryan - Thanks for the update. It\n> seemed the PR which coalesced data was assuming this change would rectify\n> some of the degradation it introduced for larger payload sizes. Do we need\n> to take any follow-up action to correct this?\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/3852#issuecomment-115365415.\n", "@louiscryan - I'm not sure how reducing the payload size by itself would mitigate an issue, just seems like we are taking a different data point, maybe you should also add another size for what was `JUMBO`? I recall you saying there was some extra copy operations going on in the benchmarks? I could see may if the benchmarks were somehow copying the now bigger payloads that may cause some overhead in this situation.\n", "@louiscryan - I think this is already done. Re-open if there is more work needed here.\n" ]

[]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameWriter.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2FrameRoundtripTest.java" ]

[ "timestamp(timedelta=477.0, similarity=0.8917241508522582)" ]

[ "@Scottmitch @nmittler please have a look guys :)\n", "I wonder if `HttpConversionUtil` should just group all multi-valued headers into single entries (i.e. not treat `Cookie` as special here). It wouldn't be invalid and I suspect would perform better end-to-end. \n\n@Scottmitch WDYT?\n", "@nmittler +1\n", "@nmittler - As far as I can tell, that won't work. What is true for one header is not necessarily true for others. Specifically I can think of the `Set-Cookie` header. In this case (depending which version/browser you use) sending each cookie in a separate `Set-Cookie` header is actually the correct thing to do (log into or out of facebook to see an example of this).\n", "Yup, looking at https://tools.ietf.org/html/rfc7230#section-3.2.2 ...\n\n`Set-Cooke` is the exception. It also sounds for Netty to function as a proxy, it should not merge headers.\n\nAnd as described in http://tools.ietf.org/html/rfc6265#section-5.4: `the user agent MUST NOT attach more than one Cookie header field`.\n\n@blucas Are you saying that Chrome is sending multiple `Cookie` headers, or that it's sending a multi-valued `Cookie` header that Netty is then splitting into separate headers? The former would appear to be a chrome bug.\n", "+1 for not merging headers :smiley: \n\n@nmittler I haven't checked that yet. My assumption is that it is a Netty related issue, as Firefox is also affected by this. I will check now and get back to you\n", "Ok, so I still don't know if it is the browser's or netty's (possibly twitter hpack's) fault. All I can tell you is\n- Both Firefox and Chrome exhibit the issue\n- Single call to [addFragment](https://github.com/netty/netty/blob/master/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java#L669) which has `endOfHeaders` set to true.\n\nMaybe @Scottmitch could shed some light on this one?\n\nBTW, looking at chrome://net-internals for the request generated this stack:\n\n```\nt= 38394 [st= 35693] HTTP2_SESSION_SEND_HEADERS\n --> fin = true\n --> :authority: www.somesite.net\n :method: GET\n :path: /managemain/menu.cfm\n :scheme: https\n accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8\n accept-encoding: gzip, deflate, sdch\n accept-language: en-US,en;q=0.8\n cache-control: max-age=0\n cookie: [118 bytes were stripped]\n upgrade-insecure-requests: 1\n user-agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_7_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/46.0.2490.80 Safari/537.36\n --> priority = 0\n --> stream_id = 3\n --> unidirectional = false\n```\n\nChrome did the `[118 bytes were stripped]` not me.\n", "@blucas - Can you post your server code somewhere (or some reduced set of it)? I just modified the example http2 helloworld server and client to both use the HTTP/1.x translation layer and added a cookie from the client (`request.headers().add(HttpHeaderNames.COOKIE, \"foo1, foo2\");`) and this is what I see on the server side:\n\n```\nDefaultFullHttpRequest(decodeResult: success, version: HTTP/1.1, content: UnpooledUnsafeHeapByteBuf(ridx: 0, widx: 0, cap: 0))\nGET /whatever HTTP/1.1\nx-http2-scheme: http\nhost: 127.0.0.1:8080\naccept-encoding: gzip\naccept-encoding: deflate\ncookie: foo1, foo2\nx-http2-stream-id: 3\ncontent-length: 0\n```\n\nIt looks like CSV headers are preserved (`cookie`), and multiple headers as different entries (`accept-encoding`) are preserved. Are you thinking we should always force everything to be a single CSV separated value unless otherwise prohibited from doing this by the spec?\n", "@Scottmitch - I don't have a reproducer handy for you to test with, but all you need to do is make your HTTP/2 client send multiple `cookie` headers to your netty HTTP/2 -> HTTP/1.x translation server.\n\n> Are you thinking we should always force everything to be a single CSV separated value unless otherwise prohibited from doing this by the spec?\n\nNo. We shouldn't force everything to use single CSV.\n\nI took a look at the HTTP/2 spec, and finally think I've found what I'm trying to explain. \n\n> To allow for better compression efficiency, the Cookie header field MAY be split into separate header fields, each with one or more cookie-pairs.\n\nSo, according to the [spec](https://http2.github.io/http2-spec/#CompressCookie), it is perfectly valid for a HTTP/2 client to send multiple `cookie` headers (thus Firefox and Chrome are complying with the spec). //cc @nmittler - I believe this answers your question.\n\n> If there are multiple Cookie header fields after decompression, these MUST be concatenated into a single octet string using the two-octet delimiter of 0x3B, 0x20 (the ASCII string \"; \") before being passed into a non-HTTP/2 context, such as an HTTP/1.1 connection, or a generic HTTP server application.\n\nSo I interpret that as Netty's `HttpConversionUtil.Http2ToHttpHeaderTranslator` should be responsible for concatenating one or more HTTP/2 `cookie` headers into a single HTTP/1.x `Cookie` header.\n\n```\n// Example 1 - Client sends multiple single value HTTP/2 cookie headers\ncookie: a=b\ncookie: c=d\ncookie: e=f\n\n// Netty should translate that to a single HTTP/1.x Cookie header\ncookie: a=b; c=d; e=f\n\n// Example 2 - Client sends single and multi-value HTTP/2 cookie headers\ncookie: a=b; c=d\ncookie: e=f\n\n// Netty should translate that to a single HTTP/1.x Cookie header\ncookie: a=b; c=d; e=f\n```\n\nI hope that makes more sense :smiley: \n", "@Scottmitch - gentle ping :)\n", "@Scottmitch ^^ \n", "@blucas - Sorry for the delay and thank you for the reference to the specification. I will submit a PR soon.\n" ]

[ "yikes ... was this missing? :)\n", "Not new code, but why does the `if` statement above check for the value of `trailers`? Seems like checking the name equals `te` is sufficient, no?\n", "final?\n", "yikes :)\n", "Are we guaranteed that we need to skip exactly 2 characters?\n", "Maybe check for the exceptional case first? Then you don't need the else.\n", "Yes sir it was :(\n\nWe were previously using all `AsciiString` in tests apparently so the value comparisons using POJO `equals` was sufficient.\n", "It is defined/declared in an interface, so it is final by definition.\n\n> /Users/scott_mitchell/git/netty/common/src/main/java/io/netty/util/ByteProcessor.java:126:5: Redundant 'final' modifier.\n", "https://tools.ietf.org/html/rfc7540#section-8.1.2.2\n\n> The only exception to this is the TE header field, which MAY be\n> present in an HTTP/2 request; when it is, it MUST NOT contain any\n> value other than \"trailers\".\n", "Yip :frowning: A few of the new unit tests now test this.\n", "Right, I'm just not sure why this logic cares about the value. It would we would only care if we were going to enforce that requirement (i.e. throw if the value is NOT `trailers`).\n", "I believe so. Let me add a reference to the cookie RFC to clarify.\n\nhttps://tools.ietf.org/html/rfc6265#section-4.2.1\n\n> cookie-header = \"Cookie:\" OWS cookie-string OWS\n> cookie-string = cookie-pair *( \";\" SP cookie-pair )\n\nhttps://tools.ietf.org/html/rfc6265#section-4.1.1\n\n> cookie-pair = cookie-name \"=\" cookie-value\n", "ah right ... didn't notice that this was an interface. Sorry for the noise :)\n", "done\n", "I guess we are enforcing it by not including it in the resulting headers, but we can throw to be more explicit if you prefer.\n", "I think throwing is probably the right thing to do ... otherwise, AFAICT we're silently dropping this header on the floor. WDYT?\n", "done.\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Headers.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java", "common/src/main/java/io/netty/util/AsciiString.java", "common/src/main/java/io/netty/util/ByteProcessor.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Headers.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java", "common/src/main/java/io/netty/util/AsciiString.java", "common/src/main/java/io/netty/util/ByteProcessor.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandlerTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/InboundHttp2ToHttpAdapterTest.java", "common/src/test/java/io/netty/util/AsciiStringMemoryTest.java" ]

[ "timestamp(timedelta=39.0, similarity=0.8664136669792186)" ]

[ "[This is the Workaround](https://github.com/saxsys/SynchronizeFX/blob/9789b889a160051c4f74097446a4bc563f2beabb/transmitter/netty-transmitter/src/main/java/de/saxsys/synchronizefx/netty/websockets/WhiteSpaceInPathWebSocketClientHandshaker13.java) i use. I basically use [getRawPath()](http://docs.oracle.com/javase/8/docs/api/java/net/URI.html#getRawPath--) instead of `getPath()`.\n", "@rbi - Thanks for reporting. Would you mind submitting a PR?\n", "Working on a fix...\n", "Sorry for not answering but thank you for fixing it. I'm looking forward to Netty 4.0.34. 👍\n", "Fixed via https://github.com/netty/netty/pull/4552\n" ]

[ "Extract the return value of `getQuery()`\n", "Fixed...\n" ]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker00.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker07.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker08.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker13.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker00.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker07.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker08.java", "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker13.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker00Test.java", "codec-http/src/test/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker07Test.java", "codec-http/src/test/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker08Test.java", "codec-http/src/test/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshaker13Test.java", "codec-http/src/test/java/io/netty/handler/codec/http/websocketx/WebSocketClientHandshakerTest.java" ]

[ "timestamp(timedelta=14.0, similarity=0.8830104167249984)" ]

[ "Fixed by https://github.com/netty/netty/pull/4565\n" ]

[]

[ "defect" ]

[ "common/src/main/java/io/netty/util/AsciiString.java" ]

[ "common/src/main/java/io/netty/util/AsciiString.java" ]

[ "common/src/test/java/io/netty/util/AsciiStringCharacterTest.java" ]

[ "timestamp(timedelta=15.0, similarity=0.933704597804567)" ]

[ "@nmittler - FYI\n", "@Scottmitch good find! Any thoughts on a fix?\n", "@nmittler - Yes. See https://github.com/netty/netty/pull/4599\n" ]

[ "Should `streamableBytes` just be `pendingBytes` or something like that. `streamableBytes` is a derived value based on the stream window.\n", "spelling: `writing`\n", "what if `windowSize == 0 && pendingBytes == 0`? Seems like we would want to write an empty frame.\n", "done\n", "`pendingBytes == 0` should be accounted for by `hasFrame`. `windowSize == 0` is permitted if we are not writing. The rational being we allow it to be added to the queue to write 1 time, but if it is 0 during/after we write then we don't add it again (until updateStreamableBytes is called again).\n\nAm I missing something?\n", "default methods would be nice for this ... let me see what this looks like because the distributors will need to calculate this value anyways.\n", "Ok, I think this will work.\n", "+1\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/PriorityStreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/StreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/UniformStreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/WeightedFairQueueByteDistributor.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/PriorityStreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/StreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/UniformStreamByteDistributor.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/WeightedFairQueueByteDistributor.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/PriorityStreamByteDistributorTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/UniformStreamByteDistributorFlowControllerTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/UniformStreamByteDistributorTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/WeightedFairQueueByteDistributorTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/WeightedFairQueueRemoteFlowControllerTest.java" ]

[ "timestamp(timedelta=15.0, similarity=0.9236626301501217)" ]

[ "@nmittler - FYI\n", "@Scottmitch good catch! ... do you want to throw together a PR with the updated test?\n", "@nmittler - Yes. I can handle this. However the new interface required to fix this is introduced and used in PR #4538. Would you mind if I fixed after this issue after PR #4538 is resolved?\n", "@Scottmitch sgtm, thanks!\n", "pr pending ...\n" ]

[ "Maybe instead of using another variable `writeAllowed`, it might be more clear if we just use `windowSize` directly:\n\n``` java\nif (state.windowSize <= 0) {\n continue;\n}\n```\n", "@nmittler - We could do this, but boolean should be less memory, and this would be the only reason why we need `windowSize`. Would changing the variable name or adding javadocs on the variable be sufficient?\n", "Also I think the condition would be `< 0` (instead of `<= 0`).\n", "@Scottmitch ok, but can we name it to something more clear? Maybe something like `windowCollapsed` or `hasWindow`?\n", "@nmittler - `windowNegative` work for you?\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/UniformStreamByteDistributor.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/UniformStreamByteDistributor.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/UniformStreamByteDistributorTest.java" ]

[ "timestamp(timedelta=4.0, similarity=0.9201898721503884)" ]

[ "@louiscryan @nmittler @ejona86 - I checked in gRPC and it looks like this is still not being used? Is the plan still to use this or eventually consume the stream writability changes?\n", "So far it has not proven critical to performance and there are other perf improvements I would prioritize in GRPC above the use-case for this, particularly with the improvements in flow-control write allocations and frame merging.\n\nIm OK with getting rid of this\n", "after we get PR https://github.com/netty/netty/pull/4599 pulled in I'll submit a PR for this (I have one pending)\n", "Thanks Scott\n\nOn Fri, Dec 18, 2015 at 1:27 PM, Scott Mitchell [email protected]\nwrote:\n\n> after we get PR #4599 https://github.com/netty/netty/pull/4599 pulled\n> in I'll submit a PR for this.\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/4600#issuecomment-165901318.\n" ]

[]

[ "cleanup" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2RemoteFlowController.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2RemoteFlowController.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java" ]

[ "timestamp(timedelta=27.0, similarity=0.8533197798933183)" ]

[ "Sounds like an issue with loading the jni stuff. Can you show me how you specify the dependencies ?\n", "```\n <dependency>\n <groupId>io.netty</groupId>\n <artifactId>netty-all</artifactId>\n <version>5.0.0.Alpha2</version>\n <scope>test</scope>\n </dependency>\n```\n\nit works if I comment out isKeepAlive() call. I think the problem is that isKeepAlive is missing in io_netty_channel_epoll_Native.h \n", "@vrozov please use 4.0.33.Final or 4.1.0.Beta8. It should work there. If not re-open\n", "Actually It seems like there is still the bug. Reopen now\n", "@vrozov fixed via https://github.com/netty/netty/pull/4619\n", "Fixed\n" ]

[ "consider making this a member variable and using a `@Before` to initialize and `@After` to tear it down.\n" ]

[ "defect" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_unix_Socket.c", "transport-native-epoll/src/main/c/io_netty_channel_unix_Socket.h" ]

[ "transport-native-epoll/src/main/c/io_netty_channel_unix_Socket.c", "transport-native-epoll/src/main/c/io_netty_channel_unix_Socket.h" ]

[ "transport-native-epoll/src/test/java/io/netty/channel/unix/SocketTest.java" ]

[ "timestamp(timedelta=32.0, similarity=0.8424088020104541)" ]

[ "@Scottmitch is this still valid ?\n", "@normanmaurer - I think so. I have not done anything to fix it.\n", "Fixed\n" ]

[ "can you explain why we need an array?\n", "check `if (!timeout[0].isDone()) {..}` before adding the object?\n", "I wonder if we need to keep a collection of timeouts. Because the time is fixed, each new event will alway fire later (in time) than the previous event. Can we just keep a reference to the \"last\" event, and cancel it when necessary (scheduling new event, cleanup, etc..)?\n", "sounds good.\n", "I guess the array is used @windie tries to access the future that is returned by `schedule(...)` in the `OneTimeTask` itself. Consider just using create a `OneTimeTask` subclass which will contain the `Future` in a field. Something like this:\n\n``` java\nprivate class TimeoutTask extends OneTimeTask {\n Future<?> future;\n\n @Override\n public void run() {\n ...\n if (future != null) {\n timeouts.remove(future); \n }\n }\n}\n\n....\nTimeoutTask task = new TimeoutTask();\n\ntask.future = ctx.executor().schedule(task, timeoutNanos, TimeUnit.NANOSECONDS);\n...\n```\n", "+1\n", "Addressed and rebased\n", "Addressed and rebased\n", "need to check if timeout != null a well\n", "So it's guaranteed that ChannelPromise's listener will be called before the next `write` event?\n", "task.timeout is assigned before this line\n", "consider renaming it to scheduledFuture.\n", "rename to promise\n", "static\n", "doh! I need more coffee... you are right.\n", "it needs to access the outer `timeouts`\n", "no it's not... but it is guaranteed it is called once it is written or the write failed. I just wonder if we can somehow remove the overhead of the Set in terms of memory + remove/add.\n", "> So it's guaranteed that ChannelPromise's listener will be called before the next write event?\n\nJust want to confirm this one before removing the collection of timeouts\n", "what we also could do is make the WriteTimeoutTask basically a doubly-linked list and so have a ref to `prev` and `next` task. This way adding and removing would be O(1) for our use-case and we not need any extra data structure. Not sure if this is better or not though\n", "> no it's not... but it is guaranteed it is called once it is written or the write failed. I just wonder if we can somehow remove the overhead of the Set in terms of memory + remove/add.\n\nSo if just storing a reference to the last `timeout`, it may override the previous one. Then the previous one won't be cleaned in `handlerRemoved`. Seems we need a collection to store active timeouts at least.\n", "@windie yeah it seems so :( What you think about my \"doubly-linked-list\" idea ?\n", "> @windie yeah it seems so :( What you think about my \"doubly-linked-list\" idea ?\n\nSounds good. Let me try it.\n", "nit: remove final\n", "makes sense\n", "Can we also have WriteTimeoutTask implement ChannelFutureListener and so safe one object allocation ?\n" ]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/timeout/WriteTimeoutHandler.java" ]

[ "handler/src/main/java/io/netty/handler/timeout/WriteTimeoutHandler.java" ]

[]

[ "timestamp(timedelta=48.0, similarity=0.8704097078767499)" ]

[ "If you can make it with breakage use 4.0\n\n> Am 05.10.2015 um 17:36 schrieb Stephane Landelle [email protected]:\n> \n> Netty version: 4.0.32.Final\n> \n> Context:\n> In AHC, we have our own builder for requests, where we internally store headers into a HttpHeaders instance.\n> \n> Sadly, there's no way to directly pass a HttpHeaders instance to DefaultHttpMessage hierarchy and we have to set our instance into the DefaultHttpMessage's internal one that's been eagerly created.\n> \n> This results is an excessive allocation that could be avoided.\n> \n> Please let me know if you're interested in a PR, and which branch to target.\n> \n> Regards\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Will do :)\n", "> If you can make it with breakage use 4.0\n\nJust to clarify ... If you can make it **without** breaking use 4.0. @normanmaurer ?\n", "lol.. yes. Fat fingers :D\n", "Yeah, that made me smile, but I got it right, no worries ;-)\n", "@slandelle ping :) \n", "Fixed by https://github.com/netty/netty/pull/4639\n" ]

[ "ObjectUtils.checkNonNull(...)\n", "Replace with ObjectUtils.checkNonNull(...)\n", "the getStatus -> the status\n", "getMethod -> method\n" ]

[ "improvement" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpResponse.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpMessage.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpResponse.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultFullHttpResponse.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpMessage.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpRequest.java", "codec-http/src/main/java/io/netty/handler/codec/http/DefaultHttpResponse.java" ]

[]

[ "timestamp(timedelta=38.0, similarity=0.9066869632289366)" ]

[ "will have a look\n", "Fixed in #4681\n", "@valodzka could you test with latest 4.0 branch: \n\nThis test works for me:\n\n```\n\n @Test\n public void testToStringComposite() {\n CompositeByteBuf cBuff = compositeBuffer();\n cBuff.addComponent(buffer(8).writeLong(0));\n cBuff.addComponent(buffer(8).writeLong(0));\n\n // work as expected\n System.err.println(\"str=\"+ cBuff.toString(0, 16, CharsetUtil.UTF_8));\n // fails\n System.err.println(\"str=\"+ cBuff.toString(3, 13, CharsetUtil.UTF_8));\n }\n```\n", "@valodzka doh! I did not notice @windie fixed it and @Scottmitch pulled it in. @windie @Scottmitch you rock!\n" ]

[ "Can you use a more descriptive test name? something like `testToStringDoesNotThrowIndexOutOfBounds`\n" ]

[ "defect" ]

[ "buffer/src/main/java/io/netty/buffer/ByteBufUtil.java" ]

[ "buffer/src/main/java/io/netty/buffer/ByteBufUtil.java" ]

[ "buffer/src/test/java/io/netty/buffer/ByteBufUtilTest.java" ]

[ "timestamp(timedelta=96943.0, similarity=0.8577363022834855)" ]

[ "> the DefaultFullHttpResponse implements LastHttpContent, after encode any of DefaultFullHttpResponse msg, HttpObjectEncoder#encodeChunkedContent will reset state to ST_INIT, then rest msg can not be encoded.\n\nI think this is the expected behaviour since it's a completed http message.\n", "in `HttpStaticFileServer` only the response header was encoded, but the `chunked` file content can not be encoded.\nAnd I think this will also cause `Keep-Alive` not work.\n", "> the sample code not set transfer-encoding header to chunked. in HttpObjectEncoder will process as chunked response only the response msg has chunked header.\n\nJust sent #4683 to fix it.\n", "> in HttpStaticFileServer only the response header was encoded, but the chunked file content can not be encoded.\n> And I think this will also cause Keep-Alive not work.\n\nHttpStaticFileServerHandler uses DefaultHttpResponse to send the file:\nhttps://github.com/netty/netty/blob/b6a4f5de9d213cdbe32ff7772be4bcf6a1ba9094/example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java#L177\n\n`DefaultFullHttpResponse` is used to send the file list and doesn't use HttpChunkedInput.\n", "HttpStaticFileServerHandler uses `DefaultHttpResponse` to send the `header`:\nhttps://github.com/netty/netty/blob/b6a4f5de9d213cdbe32ff7772be4bcf6a1ba9094/example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java#L186\n\nand uses `HttpChunkedInput` to send the `content`:\nhttps://github.com/netty/netty/blob/b6a4f5de9d213cdbe32ff7772be4bcf6a1ba9094/example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java#L198\n\nbut only the `header` was encoded by `HttpObjectEncoder`, the `content` can not be encoded.\n", "When I ran `HttpStaticFileServer` with ssl and downloaded a file, I did see `HttpObjectEncoder` processes `DefaultHttpResponse` and the following `DefaultHttpContent`s. Did you see any unexpected behaviour?\n", "`IllegalStateException` will be throwed in `HttpObjectEncoder` both http and https:\nhttps://github.com/netty/netty/blob/netty-5.0.0.Alpha2/codec-http/src/main/java/io/netty/handler/codec/http/HttpObjectEncoder.java#L92\n\nbecause after the `DefaultHttpResponse header` was encoded, the `state` has been reset to `ST_INIT`:\nhttps://github.com/netty/netty/blob/netty-5.0.0.Alpha2/codec-http/src/main/java/io/netty/handler/codec/http/HttpObjectEncoder.java#L171\n", "Sorry, I was testing the branch 4.1. Yes, `HttpStaticFileServer` is broken in `5.0.0.Alpha2`. However, since 5.0 has been dropped (#4466), could you test branch 4.1 and see if it still doesn't work?\n", "I tried `4.1` and got the same error.\n", "@cpf624 can you please tell us the exact steps you did. Tried with 4.1 branch with and without ssl and it \"just works\".\n", "Sorry, it's my fault, I write wrong `HttpResponse` as `FullHttpResponse`:\n\n``` java\nFullHttpResponse response = new DefaultFullHttpResponse(HTTP_1_1, OK);\n```\n\nit should be:\nhttps://github.com/netty/netty/blob/4.1/example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java#L177\n", "Ok closing...\n" ]

[]

[]

[ "example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java" ]

[ "example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java" ]

[]

[ "timestamp(timedelta=38.0, similarity=0.8717868222015914)" ]

[ "@normanmaurer @trustin - Any thoughts?\n", "+1 thanks for creating @Scottmitch \n", "@Scottmitch +1\n", "@Scottmitch +1\n", "@nmittler - Looks like we have our answer :) I'll take the assignment for now so it doesn't get dropped...feel free to take it though.\n", "@Scottmitch FYI this isn't a super high priority item for us ... I'm probably not going to pick it up any time soon.\n", "Taking so we get this fixed before the next release (which will be this week!) :)\n" ]

[ "Can the level change? If not, would it be make sense to save off some sort of allocator for the leak aware buffers?\n", "Yes it can change\n", "Since you're sharing this now, would it make sense to move this to some sort of common util class? It seems to be the only thing that uses `ACQUIRE_AND_RELEASE_ONLY`, so that could move too.\n", "I thought this is overkill as it is only one method\n\n> Am 15.01.2016 um 15:50 schrieb Nathan Mittler [email protected]:\n> \n> In buffer/src/main/java/io/netty/buffer/AdvancedLeakAwareByteBuf.java:\n> \n> > @@ -53,7 +53,7 @@\n> > this.leak = leak;\n> > }\n> > - private void recordLeakNonRefCountingOperation() {\n> > - static void recordLeakNonRefCountingOperation(ResourceLeak leak) {\n> > Since you're sharing this now, would it make sense to move this to some sort of common util class? It seems to be the only thing that uses ACQUIRE_AND_RELEASE_ONLY, so that could move too.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Not really related, but could you add documentation to this class about how `maxNumComponents` works. It's not really a hard max, correct?\n", "will do in a follow up pr\n" ]

[ "improvement", "feature" ]

[ "buffer/src/main/java/io/netty/buffer/AbstractByteBufAllocator.java", "buffer/src/main/java/io/netty/buffer/AbstractReferenceCountedByteBuf.java", "buffer/src/main/java/io/netty/buffer/AdvancedLeakAwareByteBuf.java", "buffer/src/main/java/io/netty/buffer/CompositeByteBuf.java", "buffer/src/main/java/io/netty/buffer/Unpooled.java" ]

[ "buffer/src/main/java/io/netty/buffer/AbstractByteBufAllocator.java", "buffer/src/main/java/io/netty/buffer/AbstractReferenceCountedByteBuf.java", "buffer/src/main/java/io/netty/buffer/AdvancedLeakAwareByteBuf.java", "buffer/src/main/java/io/netty/buffer/AdvancedLeakAwareCompositeByteBuf.java", "buffer/src/main/java/io/netty/buffer/CompositeByteBuf.java", "buffer/src/main/java/io/netty/buffer/SimpleLeakAwareCompositeByteBuf.java", "buffer/src/main/java/io/netty/buffer/Unpooled.java", "buffer/src/main/java/io/netty/buffer/WrappedCompositeByteBuf.java" ]

[]

[ "timestamp(timedelta=66518.0, similarity=0.8475097645824541)" ]

[ "@lw346 could you please comment ? \n", "I guess he meant any number `&` with `0x3F` (63) must be less than 64.\n", "Fixed in #4731\n" ]

[ "Removed `private` to call it in test\n", "Removed `private` to call it in test\n", "`decoded` and `Unpooled.wrappedBuffer(...)` need to be released.\n", "+1\n", "Just call ByteBuf.release() ?\n", "just call `in.release()`?\n", "@normanmaurer - +1 :wink: \n" ]

[]

[ "codec/src/main/java/io/netty/handler/codec/compression/Snappy.java" ]

[ "codec/src/main/java/io/netty/handler/codec/compression/Snappy.java" ]

[ "codec/src/test/java/io/netty/handler/codec/compression/SnappyTest.java" ]

[ "timestamp(timedelta=12.0, similarity=0.9466880115333267)" ]

[ "@normanmaurer @trustin - Any objections?\n", "@ejona86 works for me. Please open a pr\n" ]

[]

[ "feature" ]

[ "microbench/src/main/java/io/netty/microbench/channel/EmbeddedChannelWriteReleaseHandlerContext.java", "transport/src/main/java/io/netty/channel/AbstractChannelHandlerContext.java", "transport/src/main/java/io/netty/channel/ChannelHandlerContext.java", "transport/src/main/java/io/netty/channel/CombinedChannelDuplexHandler.java" ]

[ "microbench/src/main/java/io/netty/microbench/channel/EmbeddedChannelWriteReleaseHandlerContext.java", "transport/src/main/java/io/netty/channel/AbstractChannelHandlerContext.java", "transport/src/main/java/io/netty/channel/ChannelHandlerContext.java", "transport/src/main/java/io/netty/channel/CombinedChannelDuplexHandler.java" ]

[]

[ "timestamp(timedelta=46.0, similarity=0.8595048383618223)" ]

[]

[]

[]

[ "transport/src/main/java/io/netty/channel/ChannelOutboundHandlerAdapter.java" ]

[ "transport/src/main/java/io/netty/channel/ChannelOutboundHandlerAdapter.java" ]

[]

[ "timestamp(timedelta=87.0, similarity=0.9265861370024806)" ]

[ "/cc @normanmaurer @nmittler @trustin \n", "@trustin ping again\n" ]

[ "unused import ?\n", "ReferenceCountUtil.safeRelease(...) ?\n", "see above\n", "I was just anticipating your next comments :wink: \n", "I originally had this, but figured we could avoid the instanceof check and casting by the simple `try/catch` block.\n" ]

[ "defect" ]

[ "buffer/src/main/java/io/netty/buffer/CompositeByteBuf.java" ]

[ "buffer/src/main/java/io/netty/buffer/CompositeByteBuf.java" ]

[]

[ "timestamp(timedelta=7.0, similarity=0.8969352858274092)" ]

[ "@louiscryan - makes sense.\n", "@louiscryan - Any updates on this? It would be nice to get this in before the next release. I can take over if you don't have cycles.\n", "@louiscryan - I'll take this over for now. We can improve tests later.\n", "Sorry Scott. Let me dig up my PR tomorrow and at least send that over so\nyou can take it over. It's pretty small\n\nOn Wed, Jan 20, 2016 at 5:12 PM, Scott Mitchell [email protected]\nwrote:\n\n> @louiscryan https://github.com/louiscryan - I'll take this over for\n> now. We can improve tests later.\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/4573#issuecomment-173418841.\n", "@louiscryan - No problem. I have a PR pending, but if you push yours so I can make sure I covered everything. I'll also ping you for a review of my PR.\n" ]

[ "Why wouldn't we just increment by frame.size()? Is the padding not merged?\n", "The padding is currently not necessarily additive when a merge occurs. It is currently taken as `max`.\n\nSee https://github.com/netty/netty/issues/4573 for details too.\n", "@nmittler - PTAL. This is to address https://github.com/netty/netty/issues/4773.\n\nI could add a unit test, but we don't have an environment setup to mock buffers that go through this code so it would require a huge allocation amount or take a bit more time to mock...\n", "@Scottmitch perfect :)\n", "sgtm\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2ConnectionEncoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2ConnectionEncoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowController.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2RemoteFlowControllerTest.java" ]

[ "timestamp(timedelta=23.0, similarity=0.8896217346671011)" ]

[ "@mrniko you are right... I will fix it. In the meantime just create a new instance everytime.\n", "@mrniko PTAL https://github.com/netty/netty/pull/4769\n", "@normanmaurer PTAL ? Sorry, what? :)\n", "@mrniko Please take a look ;) \n", "@mrniko actually only WebSocketClientCompressionHandler can be `@Sharable` as the other is keeping state. Fixed the PR.\n", "@normanmaurer OK, I got you!\n", "Fixed by https://github.com/netty/netty/pull/4769\n", "@normanmaurer @mrniko why WebSocketServerCompressionHandler cannot be be @Sharable . i can't understand. Can you explain it to me？", "It's a subclass of `WebSocketServerExtensionHandler` which keeps state per `connection` so it cant be shared. " ]

[ "What is the need for the `INSTANCE` member if it can't be reused and the constructor is public?\n", "ups sorry missed to remove it ... \n" ]

[ "improvement" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/extensions/compression/WebSocketClientCompressionHandler.java", "example/src/main/java/io/netty/example/http/websocketx/client/WebSocketClient.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/websocketx/extensions/compression/WebSocketClientCompressionHandler.java", "example/src/main/java/io/netty/example/http/websocketx/client/WebSocketClient.java" ]

[]

[ "timestamp(timedelta=65338.0, similarity=0.8604314867155635)" ]

[ "Fixed in #4706\n", "Corresponding Jython bug: http://bugs.jython.org/issue2401 - we should be able to build shortly, then Nick can see if he reproduce in his environment\n", "Can we also merge this into 4.0?\n", "This will need a more general fix... Looking into it now.\n\nAs workaround you can add a task o the EventLoop that will actually add the handler to the pipeline\n", "This is a big tricky... so may take some time.\n", "So I applied the fix from #4706 by subclassing SslHandler and setting ctx in channelActive. That fixed the race condition for me and I don't see any other negative effects, at least so far. Would you recommend against actually utilizing that as a fix until it can be fixed more generally?\n", "@nickmbailey could you try to use netty 4.1 and use this patch ? :\n\nhttps://github.com/netty/netty/pull/4716\n\nI hope this is a fundamental fix.\n", "I'll have to set up another environment for reproducing the issue, but I'll try to get to that sometime today and verify the issue.\n\nSince we need a fix in Jython and won't be able to upgrade netty to 4.1 in the immediate future, is the approach I described above relatively safe in your opinion?\n", "It is a workaround and should work…\n\nThat said once I know it works I will backport the fix to 4.0 as well.\n\n> On 15 Jan 2016, at 18:53, Nick Bailey [email protected] wrote:\n> \n> I'll have to set up another environment for reproducing the issue, but I'll try to get to that sometime today and verify the issue.\n> \n> Since we need a fix in Jython and won't be able to upgrade netty to 4.1 in the immediate future, is the approach I described above relatively safe in your opinion?\n> \n> —\n> Reply to this email directly or view it on GitHub https://github.com/netty/netty/issues/4705#issuecomment-172033188.\n", "So since I'm hitting the error running through jython on 4.0.33, I can't easily test out this fix.. I tried switching out 4.0.33 with 4.1 but apparently there are big enough changes that it isn't a drop in replacement. \n", "Can you share the project?\n\n> Am 17.01.2016 um 01:35 schrieb Nick Bailey [email protected]:\n> \n> So since I'm hitting the error running through jython on 4.0.33, I can't easily test out this fix.. I tried switching out 4.0.33 with 4.1 but apparently there are big enough changes that it isn't a drop in replacement.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "https://bitbucket.org/jython/jython/src\n", "Jython seems to be using Netty: https://bitbucket.org/jython/jython/src/6855fa289d48cf741a3daed2a0bb9c9c87cdf07f/build.xml?at=default&fileviewer=file-view-default#build.xml-181:185\n", "Yeah jython uses netty under the hood. I'm hitting the issue using jython, I have a script that reproduces the issue and I can provide that. I only see the issue on ubuntu 14.04 and not on OSX for whatever reason.\n", "Here's the script I can reproduce with and an example run:\n\nhttps://gist.github.com/nickmbailey/f80ad965a048c1d4df22\n", "@nickmbailey thanks.. what are the setup steps for jython ?\n", "There isn't any. You can download the jar from here: http://mvnrepository.com/artifact/org.python/jython-standalone/2.7.1b2\n\nI commented on the gist with how to run that script, so once you download that jar and script, that should be all you need.\n", "@nickmbailey got it... will check\n", "@nickmbailey can you give me any pointers how I can use a custom netty jar with jython ?\n", "@nickmbailey @normanmaurer I assume adding it to the classpath first should be ok? Unfortunately I am still seeing issues with 27df0cc6945d43e20d6469ce2bf2fa3a9b8f61dd\n\n```\n[johno@habbo-dev-1 netty]$ git log --name-status HEAD^..HEAD\ncommit 27df0cc6945d43e20d6469ce2bf2fa3a9b8f61dd\nAuthor: Norman Maurer <[email protected]>\nDate: Fri Jan 15 14:23:41 2016 +0100\n\n [#4705] Ensure ChannelHandler.handlerAdded(...) is always called as first method of the handler\n\n Motivation:\n\n If a user adds a ChannelHandler from outside the EventLoop it is possible to get into the situation that handlerAdded(...) is scheduled on the EventLoop and so called after another methods of the ChannelHandler as the EventLoop may a\n\n Modification:\n\n - Ensure we always check if the handlerAdded(...) method was called already and if not add the currently needed call to the EventLoop so it will be picked up after handlerAdded(...) was called. This works as if the handler is added t\n - Some cleanup in the DefaultChannelPipeline\n\n Result:\n\n Correctly order of method executions of ChannelHandler.\n\nM transport/src/main/java/io/netty/channel/AbstractChannelHandlerContext.java\nM transport/src/main/java/io/netty/channel/DefaultChannelPipeline.java\n```\n\n```\n[johno@habbo-dev-1 netty]$ java -cp all/target/netty-all-4.1.0.CR1-SNAPSHOT.jar:jython-standalone-2.7.1b2.jar org.python.util.jython netty-ssl-bug.py\n200 OK\n0\n200 OK\n1\n200 OK\n2\n200 OK\n3\nTraceback (most recent call last):\n File \"netty-ssl-bug.py\", line 4, in <module>\n conn.request(\"GET\", \"/\")\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1001, in request\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1035, in _send_request\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 997, in endheaders\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 850, in _send_output\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 812, in send\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1204, in connect\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/_socket.py\", line 357, in handle_exception\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/_socket.py\", line 357, in handle_exception\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 320, in wrap_socket\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 147, in __init__\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 196, in do_handshake\n_socket.SSLError: [Errno 1] Unmapped exception: java.lang.NullPointerException\n```\n\n```\n[johno@habbo-dev-1 netty]$ java -cp all/target/netty-all-4.1.0.CR1-SNAPSHOT.jar:jython-standalone-2.7.1b2.jar org.python.util.jython netty-ssl-bug.py\n200 OK\n0\n200 OK\n1\n200 OK\n2\n200 OK\n3\nTraceback (most recent call last):\n File \"netty-ssl-bug.py\", line 4, in <module>\n conn.request(\"GET\", \"/\")\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1001, in request\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1035, in _send_request\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 997, in endheaders\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 850, in _send_output\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 812, in send\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/httplib.py\", line 1204, in connect\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/_socket.py\", line 357, in handle_exception\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/_socket.py\", line 357, in handle_exception\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 320, in wrap_socket\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 147, in __init__\n File \"/home/johno/netty/jython-standalone-2.7.1b2.jar/Lib/ssl.py\", line 196, in do_handshake\n_socket.SSLError: [Errno 1] Illegal state exception\n```\n", "@johnou I don't think so as it seems like they shadow the netty version /cc @nickmbailey \n", "@normanmaurer ah I see what you mean, it's in the jar as io/netty and org/python/netty.\n", "@normanmaurer looking at their build.xml it's just an uber jar created with jarjar.\n\nI took the pipeline_race branch, built the artifacts, unzipped the all artifact and replaced the netty classes in jython-standalone-2.7.1b2.jar, repeating the same class patching process with a shaded jar created with jarjar (see ant script below).\n\nI can no longer reproduce the errors I pasted above https://github.com/netty/netty/issues/4705#issuecomment-174468261 (ran the script 10 times+).\n\nI uploaded an unofficial patched version of jython to my github branch https://github.com/johnou/netty/commit/b08bcdfef205df4eb43bc8228a6e2e3a91036775\n\n```\n[johno@habbo-dev-1 netty]$ unzip all/target/netty-all-4.1.0.CR1-SNAPSHOT.jar\n[johno@habbo-dev-1 netty]$ zip -r jython-standalone-2.7.1b2.jar io\n.. same with the shaded jar (nb. cd and directory structure must match for patching to work)\n```\n\n```\n<project name=\"MyProject\" default=\"dist\" basedir=\".\">\n<target name=\"dist\">\n <taskdef name=\"jarjar\" classname=\"com.tonicsystems.jarjar.JarJarTask\"\n classpath=\"jarjar-1.4.jar\"/>\n <jarjar jarfile=\"netty-all-4.1.0.CR1-shaded.jar\">\n <zipfileset src=\"netty-all-4.1.0.CR1-SNAPSHOT.jar\" excludes=\"META-INF/**\"/>\n <rule pattern=\"io.netty.**\" result=\"org.python.netty.@1\"/>\n </jarjar>\n</target>\n</project>\n```\n\n```\n[johno@habbo-dev-1 netty]$ java -jar jython-standalone-2.7.1b2-patched.jar netty-ssl-bug.py\n200 OK\n0\n200 OK\n1\n200 OK\n2\n200 OK\n3\n200 OK\n4\n200 OK\n5\n200 OK\n......\n200 OK\n94\n200 OK\n95\n200 OK\n96\n200 OK\n97\n200 OK\n98\n200 OK\n99\n```\n", "Yeah, when I tried to run the pipeline branch previously I copied the updated netty jars into the jython checkout and updated build.xml to use those instead. I think the approach @johnou took should work as well.\n", "@nickmbailey @johnou ok cool. Let me backport the change to 4.0 as well.\n", "@normanmaurer are you actively working on this? If not I have some time to backport your commit to 4.0.\n", "Yep should have it done later today.\n\n> Am 26.01.2016 um 16:11 schrieb Johno Crawford [email protected]:\n> \n> @normanmaurer are you actively working on this? If not I have some time to backport your commit to 4.0.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "\n", "Fixed\n" ]

[]

[ "defect" ]

[ "src/main/java/org/jboss/netty/handler/ssl/SslHandler.java" ]

[ "src/main/java/org/jboss/netty/handler/ssl/SslHandler.java" ]

[]

[ "timestamp(timedelta=21.0, similarity=0.8585134029208575)" ]

[ "@jknair I think this is the same issue as :\n\nhttps://github.com/netty/netty/issues/4722\n\nLooking into it atm.\n", "@jknair can you share a \"full\" reproducer ?\n", "sorry for the delay, I will try to make one. Problem is I cant find any SNI enabled public host that will fail the handshake \n", "@jknair maybe there is no need at all. Let me check if I can write one myself\n", "@jknair can you show me a basic client setup that works with JDK version ?\n\nDo you do the same as here:\nhttps://www-01.ibm.com/support/knowledgecenter/#!/SSYKE2_8.0.0/com.ibm.java.security.component.80.doc/security-component/jsse2Docs/sni_extension.html\n ?\n", "the following code does **NOT** work with Openssl client \n\n``` java\nSSLEngine sslEngine = sslCtx.newEngine(ctx.channel().alloc(), peerHost, peerPort);\n SSLParameters sslParameters = sslEngine.getSSLParameters();\nsslParameters.setServerNames(new ArrayList<SNIServerName>(1) {{\n add(new SNIHostName(peerHost));\n}});\nsslEngine.setSSLParameters(sslParameters);\nfinal SslHandler ssllHandler = new SslHandler(sslEngine);\n```\n\nand for the JDK version this works perfectly dont need to add any `SNIServerName` : \n\n``` java\nSSLEngine sslEngine = sslCtx.newEngine(ctx.channel().alloc(), peerHost, peerPort);\nfinal SslHandler ssllHandler = new SslHandler(sslEngine);\n```\n", "@jknair and peerHost and peerPort are anything public that I can hit ?\n", "@jknair ok I think I know whats wrong.. working on a fix.\n", "Fixed by https://github.com/netty/netty/pull/4791\n", "sorry for the delay but this worked perfectly. Thanks @normanmaurer \n" ]

[]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslEngine.java", "pom.xml" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslEngine.java", "pom.xml" ]

[ "handler/src/test/java/io/netty/handler/ssl/SniClientTest.java" ]

[ "timestamp(timedelta=36.0, similarity=0.8991627586810774)" ]

[ "@floragunncom which java version and java brand ?\n", "Oracle 1.8.0_45 on Mac OS Yosemite\n\njava version \"1.8.0_45\"\nJava(TM) SE Runtime Environment (build 1.8.0_45-b14)\nJava HotSpot(TM) 64-Bit Server VM (build 25.45-b02, mixed mode)\n\nI need to create a new Open SSL Engine and i do it with:\n\n<pre>\nsslContext.newEngine(PooledByteBufAllocator.DEFAULT);\n</pre>\n\nMaybe this is the flaw?\n", "@floragunncom that said I think we should use JDK provider if Unsafe is not present. I will come up with a patch\n", "@floragunncom actually I think we not need unsafe for OpenSsl. Working on a fix\n", "Fixed...\n" ]

[]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSsl.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslContext.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslEngine.java" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSsl.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslContext.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslEngine.java" ]

[]

[ "timestamp(timedelta=19.0, similarity=0.8923300350807587)" ]

[ "cc: @nmittler & @normanmaurer\n", "Why not use a CHMv8 here? It seems the original concern was `AddressResolver` instantiation cost, but that's what `computeIfAbsent` is for. This would let us remove all the synchronized blocks.\n", "Because it's java8 only... Will fix the race though\n\n> Am 03.02.2016 um 23:09 schrieb Stephane Landelle [email protected]:\n> \n> Why not use a CHMv8 here? It seems the original concern was AddressResolver instantiation cost, but that's what computeIfAbsent is for. This would let us remove all the synchronized blocks.\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@normanmaurer I'm referring to your internal fork: https://github.com/netty/netty/blob/4.1/common/src/main/java/io/netty/util/internal/chmv8/ConcurrentHashMapV8.java\n", "@trustin WDYT?\n\n> Am 04.02.2016 um 08:08 schrieb Stephane Landelle [email protected]:\n> \n> @normanmaurer I'm referring to your internal fork: https://github.com/netty/netty/blob/4.1/common/src/main/java/io/netty/util/internal/chmv8/ConcurrentHashMapV8.java\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "Fixed by https://github.com/netty/netty/pull/4844\n" ]

[]

[ "defect" ]

[ "resolver/src/main/java/io/netty/resolver/AddressResolverGroup.java" ]

[ "resolver/src/main/java/io/netty/resolver/AddressResolverGroup.java" ]

[]

[ "timestamp(timedelta=33.0, similarity=0.917887492488813)" ]

[ "sgtm\n", "Fixed in #4852\n", "Thanks for the patch @windie !\n" ]

[]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/cors/CorsConfigBuilder.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/cors/CorsConfigBuilder.java" ]

[]

[ "timestamp(timedelta=247.0, similarity=0.9668094919673911)" ]

[ "Thanks for capturing this issue @danbev!\n\nJust as #3271 added the ability to automatically combine and CSV escape values associated with the same header name; we should also add the ability to automatically unescape values returned by the accessor methods. For example `CharSequence get(Charsequence name)` in `DefaultTextHeaders` should be unescaping the values that are returned (under some condition...which may be the same condition that activates escaping).\n", "@danbev - Any updates on this issue?\n", "Sorry this parenting thing was more time consuming than I expected. I don't\nthink I'll be of much use before I return to work in July I'm afraid.\n\nonsdag 8 april 2015 skrev Scott Mitchell [email protected]:\n\n> @danbev https://github.com/danbev - Any updates on this issue?\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/3435#issuecomment-90806591.\n", "@danbev - O wow I had no idea...congratulations! Take care of your family and forget about this ;)\n", "@danbev Congratulations! :-D\n", "Thanks guys!\n\nfredag 10 april 2015 skrev Trustin Lee [email protected]:\n\n> @danbev https://github.com/danbev Congratulations! :-D\n> \n> —\n> Reply to this email directly or view it on GitHub\n> https://github.com/netty/netty/issues/3435#issuecomment-91419961.\n", "Thanks @windie for taking care of this! The second part of this issue was moved to https://github.com/netty/netty/issues/4855.\n" ]

[ "++i would be better.\n", "It doesn't look like the `or the value unchanged` is honored like in `escapeCsv`\n", "Doesn't really matter. Javac generates the same bytes for both `i++` and `++i` if the return value is not used.\n", "Fixed. Now just return the value if it's not escaped.\n", "Remove the else\n", "nit: import static for `unescapeCsv` and `escapeCsv`\n", "if `quoted == FALSE` we are just validating proper CSV format and not doing any unescaping, right? Does it make sense to split these two functions up?\n\n``` java\npublic static void unescapeCsv(CharSequence seq) {\n boolean quoted = ...;\n if (quoted) {\n validateCsvFormat(seq);\n return seq;\n }\n StringBuilder builder = ...;\n for(/**/) {\n // unescape seq into builder\n }\n return builder.toString();\n}\n\npublic static void validateCsvFormat(CharSequence seq) {\n..\n// if invalid throw exception\n}\n```\n", "Done\n", "There is an `escapeCsv` method in `StringUtilTest`. So I just used the static import for unescapeCsv.\n", "- Could you extract `new IllegalArgumentException(...)` into a new private static factory method to remove the duplication?\n- Exception message: `invalid escaped CSV field:`\n", "Done\n", "How about `newInvalidEscapedCsvFieldException`? `new` and `CSV` -> `Csv`\n", "Done\n", "Consider adding an argument for an index to give more context as to where the problem is.\n", "Consider removing temp variable. It is only used 1 time.\n", "Consider removing explicit null check ... You \"dereference\" on this line anyways and NPE will already be thrown. There is also only 1 argument so I don't think much context will be lost by not having \"value\" in the exception message.\n", "I think it's still useful to have this check for any user-calling public methods.\n", "done\n", "done\n", "okey, I will keep it.\n" ]

[ "improvement" ]

[ "common/src/main/java/io/netty/util/internal/StringUtil.java" ]

[ "common/src/main/java/io/netty/util/internal/StringUtil.java" ]

[ "common/src/test/java/io/netty/util/internal/StringUtilTest.java" ]

[ "timestamp(timedelta=10.0, similarity=0.9219346625170636)" ]

[ "@ejona86 - Thanks for reporting! See https://github.com/netty/netty/pull/4857\n" ]

[ "Wait ... is this right? I think we still need something here to make sure that the other endpoint doesn't know about the stream (see https://github.com/netty/netty/pull/4502). Or can we not get into `IDLE` via a remote-initiated stream?\n", "#4502 was necessary because we used to queue the initial headers frames and it was possible for the user to do a `writeHeaders()` (headers would be queued) and then `writeReset()` (reset would be sent, but remote doesn't know about the stream and responds with a protocol error and the connection is closed). Now that we don't queue initial headers this scenario should not be possible (stream is opened and the remote peer doesn't know it) and we are now just trying to enforce the following:\n\nhttps://tools.ietf.org/html/rfc7540#section-6.4\n\n> RST_STREAM frames MUST NOT be sent for a stream in the \"idle\" state.\n> If a RST_STREAM frame identifying an idle stream is received, the\n> recipient MUST treat this as a connection error (Section 5.4.1) of\n> type PROTOCOL_ERROR.\n\nHowever this restriction seems strange if the peer has sent a PRIORITY frame which creates a stream in the idle state ... we are not allowed to reset it ... and the peer may think it can later send frames on this stream ... but I think we now send a stream error in this case in [DefaultHttp2ConnectionDecoder](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2ConnectionDecoder.java#L562).\n", "sgtm ... I knew you were working on not-queuing initial headers, but I wasn't sure it had gone in yet. Thanks!\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2ConnectionHandlerTest.java" ]

[ "timestamp(timedelta=4.0, similarity=0.9871739659685546)" ]

[ "@nmittler - FYI\n\nrelated to https://github.com/netty/netty/issues/4780#issuecomment-179849253\n", "@Scottmitch yikes! Are you working on a patch?\n", "@nmittler - I was planning on it, but haven't got around to it yet....does this one peek your interest?\n", "pr pending\n" ]

[ "Why not just take a `ChannelPromise`?\n", "Just curious ... not sure how Netty handles this case in general: what if `promise` is a void promise? If the caller tries to call `addListener` it will fail ... maybe we don't care, since they're the one that passed in the void promise?\n", "Don't we have a light-weight promise aggregator that basically does this?\n", "Would it make sense to just loop through and close them all and then optionally clear the map after the loop?\n", "Much better :)\n", "If the user passed in a void promise and tries to add a listener to it ... returning the same promise is also done in other places in Netty (e.g. `channel.write()`) so I don't think this should be unexpected.\n", "It makes unit testing easier (given all the Mock objects that we have, creating a `ChannelPromise` would require more setup) and there is no reason why the code has to have a `ChannelPromise` if it can work with a `Promise<Void>`.\n", "[PromiseNotifier](https://github.com/netty/netty/blob/4.1/common/src/main/java/io/netty/util/concurrent/PromiseNotifier.java) may work ... not sure if this is \"light weight\" as it creates an array.\n", "Possibly but I would rather not invalidate any logic (which may result in a connection error) that may be going on during active stream iteration. Granted how we are currently using it is when the connection is being shut (or has already been shut down) down anyways, but I think it is easy enough to respect the active stream iteration so I just did this ... WDYT?\n", "I will move this code to a utility class and submit a followup PR (or just use PromiseNotifier after I fix a bug in it :smile: )\n", "I'm just concerned about the complexity introduced by not using `ChannelPromise`. It a real setup, it will always be a `ChannelPromise` so it would seem that some of this code could go away. How much harder does this make your tests? I seem to recall mocking `ChannelPromise` without too much of an issue, no?\n", "> I'm just concerned about the complexity introduced by not using ChannelPromise\n\nWhat complexity are you referring to? The instanceof to support the void promise check?\n", "Yeah, the if gets a bit simpler that's all. I don't feel that strongly.\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Connection.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/DefaultHttp2ConnectionTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2ConnectionHandlerTest.java" ]

[ "timestamp(timedelta=44.0, similarity=0.8687496447467746)" ]

[]

[ "maybe use some lower capacity as the default one ?\n", "consider optimistically incrementing `i` here (`++i`) and decrement if found to be invalid `throw newInvalidEscapedCsvFieldException(value, i - 1);`\n" ]

[ "feature" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/CombinedHttpHeaders.java", "common/src/main/java/io/netty/util/internal/StringUtil.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/CombinedHttpHeaders.java", "common/src/main/java/io/netty/util/internal/StringUtil.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/CombinedHttpHeadersTest.java", "common/src/test/java/io/netty/util/internal/StringUtilTest.java" ]

[ "timestamp(timedelta=5583.0, similarity=0.8777541916621916)" ]

[ "@marcuslinke because we not implemented it yet ;) I will take care... Thanks for reporting\n", "Fixed by #4883\n" ]

[ "do we need this change ?\n", "future -> promise\n", "do we need this ?\n", "do we need this ?\n", "found another way, so don't need `public` anymore, fixed\n", "done\n", "no, fixed\n", "no, fixed\n" ]

[ "feature" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollSocketChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/unix/DomainSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/SocketChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollSocketChannel.java", "transport-native-epoll/src/main/java/io/netty/channel/unix/DomainSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/DuplexChannel.java", "transport/src/main/java/io/netty/channel/socket/SocketChannel.java" ]

[]

[ "timestamp(timedelta=7329.0, similarity=0.9337282066886308)" ]

[ "@rtimush yes!\n", "Duplicates https://github.com/netty/netty/issues/4881\n", "Lets close this in favor of #4881 (reported first).\n" ]

[ "do we need this change ?\n", "future -> promise\n", "do we need this ?\n", "do we need this ?\n", "found another way, so don't need `public` anymore, fixed\n", "done\n", "no, fixed\n", "no, fixed\n" ]

[ "duplicate" ]