Datasets:

---

feedback-to-code

/

mozzarella

like 1

Follow

Feedback-2-Code 7

[ "timestamp(timedelta=22.0, similarity=0.876440567581754)" ]

[ "At the moment not through this interface. What type of socket channel are you dealing with as there is likely a way to do it using an implementation of this interface.\n", "NioSocketChannel, EpollSocketChannel or OioSocketChannel\n", "I see you want it all ... stand by for a PR.\n", "Fixed by https://github.com/netty/netty/pull/5269\n" ]

[ "@normanmaurer - making `isInputShutdown()` abstract and removing this method is an API change ... i can be more strict with the API but this will change will only be for 4.1, and its OIO... let me know if you think I should spend more effort here to preserve the API\n", "this is also an API change, but IMO was \"broken\" before (see the commit message)\n", "@trustin WDYT ?\n", "@trustin WDYT ?\n", "@Scottmitch does this somehow change the behaviour of our current impl and so may break stuff ?\n", "javadocs for protected methods please\n", "same question as before \n", "setSuccess() ?\n", "should we log t if cause is not null ?\n", "should we log t if cause is not null ?\n", "setSuccess()\n", "I think this is not correct as voidPromise() will fail when you try to add listeners to it. So better just return a `newFailedFuture()` ?\n", "same as below\n", "done\n", "done\n", "done\n", "done\n", "done\n", "yes I debated this but no-one is currently attaching to the future ... it only returns a ChannelFuture so it doesn't conflict with the existing `ChannelFuture shutdownInput()` from the base interface.\n\nHowever I can change this in case future code changes i guess\n", "@normanmaurer - previously `shutdownInput` set state internally and did not attempt to set the channel state or change the state of the underlying transport. Now that we do interact with the underlying transport we either close the underlying channel or shutdown channel (if the user enabled this behavior). NIO already [checks the input is open](http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8u40-b25/sun/nio/ch/SocketChannelImpl.java#300) before reading anything. This is also now more consistent with EPOLL.\n", "done\n", "This should be `UnsupportedOperationException` \n", "This should be `UnsupportedOperationException` \n", "debug ?\n", "debug ?\n", "factory out the above code into a new method `shutdownOutput0(...)` and delegate to it.\n", "See above and delegate to `shutdownOutput0(...)`\n", "same as above\n", "same as above\n", "same as above\n", "+1\n", "+1\n", "sure\n", "sure\n", "I'm not sure it is necessary here. It is makes sense where this is currently done in `NioSocketChannel` because it can be called in a few spots ... but here this method just calls itself (so no duplication). The `shutdown` method is different because it must treat the exceptions differently so we can't really share must there.\n", "see https://github.com/netty/netty/pull/5192#discussion_r62754131\n", "see https://github.com/netty/netty/pull/5192#discussion_r62754131\n", "see https://github.com/netty/netty/pull/5192#discussion_r62754131\n", "This code is different than above. We treat exceptions differently here.\n", "see https://github.com/netty/netty/pull/5192#discussion_r62543513\n", "I think it's fine because it was broken before and it affects the transport implementations only.\n", "notify -> shutdown the input and notify\n", "Isn't `inEventLoop()` relatively an expensive operation? I'd prefer not letting it called twice when we can avoid.\n", "Sgtm I will factor it out\n" ]

[ "feature" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-rxtx/src/main/java/io/netty/channel/rxtx/RxtxChannel.java", "transport-udt/src/main/java/io/netty/channel/udt/nio/NioUdtByteConnectorChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioByteChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioMessageChannel.java", "transport/src/main/java/io/netty/channel/oio/AbstractOioByteChannel.java", "transport/src/main/java/io/netty/channel/socket/DuplexChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/oio/OioSocketChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-rxtx/src/main/java/io/netty/channel/rxtx/RxtxChannel.java", "transport-udt/src/main/java/io/netty/channel/udt/nio/NioUdtByteConnectorChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioByteChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioMessageChannel.java", "transport/src/main/java/io/netty/channel/oio/AbstractOioByteChannel.java", "transport/src/main/java/io/netty/channel/socket/DuplexChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/oio/OioSocketChannel.java" ]

[ "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollSocketShutdownOutputByPeerTest.java", "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollSocketShutdownOutputBySelfTest.java" ]

[ "timestamp(timedelta=18.0, similarity=0.8524211396772581)" ]

[ "It only checks the fingerprint. It does not traverse the certificate chain at all. Theoretically, there's a chance where an attacker can forge a certificate with identical fingerprint, although it's not going to be very easy.\n\nI agree that it needs better documentation though. Would you be interested in sending us a PR?\n", "Ah ok, it makes sense to point that out. Will make a PR to improve docs.\n", "Thanks a lot, @CodingFabian !\n", "@CodingFabian any progress here ?\n", "woops, forgot. coming\n", "@CodingFabian 😍 \n", "Fixed by https://github.com/netty/netty/pull/5195\n" ]

[ "@CodingFabian maybe add a `<a href=\"\"...>` for `Man-in-the-middle` \n", "be consistent ... either use Man or man .\n", "`This {@link TrustManagerFactory} will` -> `This {@link TrustManagerFactory} will <strong>only</strong>`\n" ]

[ "cleanup", "documentation" ]

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

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

[]

[ "timestamp(timedelta=4.0, similarity=0.8883271627404382)" ]

[ "@Zpetkov the 5.x branch has been deprecated and is no longer supported. However, the issue also exists in 4.1. I ll take care of this.\n", "Fixed by https://github.com/netty/netty/pull/5203\n" ]

[]

[ "defect" ]

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

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

[ "codec/src/test/java/io/netty/handler/codec/json/JsonObjectDecoderTest.java" ]

[ "timestamp(timedelta=3.0, similarity=0.9025616792679517)" ]

[ "I've added this to the checklist in #3667. What you've proposed sounds good.\n\nGOAWAY and RST_STREAM can simply be ignored since the channel will closed automatically (they are propagated for additional information to those that want it; HTTP/1 doesn't want it). Stream ID should always be null on a child channel (at least for now), so there should be nothing to handle there either.\n", "Not sure if this is covered somewhere or is understood but just to clarify on the following....\n\n> HeadersFrame is endStream => LastHttpContent with Trailers\n\nA HEADERS frame with EOS doesn't necessarily mean the headers are trailers. This frame could just be the first HEADERS frame and there are no DATA frames to follow.\n", "OK, that's good to know. So that would be converted to a FullHttpRequest with an empty content body?\n" ]

[ "+1\n", "fix formating \n", "fix formating\n", "I think we should retain as late as possible to mimize the risk of leak data in case of exceptions.\n", "fix formating. \n", "see above\n", "Add `.` on EOL.\n", "also assert the return value of all write calls (also in other tests).\n", "also call channel.finish() and assert the return value\n", "`FullHttpResponse` can also have `trailingHeaders()`\n", "looks like this is covered in `encodeLastContent` ... just started looking but do you see an issue with this?\n", "LastHttpContent is not mutually exclusive with HttpResponse (as is seen with FullHttpResponse). Instead of if-elses, I'd suggest having separate ifs for each part of the message, and don't treat FullHttpResponse specially.\n\n```\nif (obj instanceof HttpMessage) {\n out.add(headers);\n}\nif (obj instanceof HttpContent) {\n if (hasContent) {\n out.add(content /* no eos */);\n }\n}\nif (obj instanceof LastHttpContent) {\n if (lastContent.trailingHeaders().isEmpty()) {\n out.add(emptyDataFrameWithEOS);\n } else {\n out.add(trailers);\n }\n}\n```\n\nIt would be possible to optimize it to avoid `emptyDataFrameWithEOS`, but I question how much that matters.\n", "I'm not sure if we should attempt to insert \"filler\". Seems more expected to handle the objects as they are...\n\n``` java\nLastHttpContent last = ...;\nif (last.trailingHeaders().isEmpty()) {\n out.add(new dataframe(last.content(), true));\n} else {\n out.add(new dataframe(last.content(), false));\n out.add(new headersframe(convert(last.trailingHeaders(), ...), true));\n}\n```\n", "Ah, you're right. It's fine.\n", "consider renaming `toHttpRequest` -> `toFullHttpRequest` and renaming this `toHttpChunkedRequest` -> `toHttpRequest`. The \"chunked\" terminology may be confused with \"chunked\" encoding as defined in https://tools.ietf.org/html/rfc7230#section-4.1\n", "the suggestion in https://github.com/netty/netty/pull/5204#discussion_r62067945 by @ejona86 would also help mitigate this issue.\n", "this will create problems if an exception occurs (e.g. during headers conversion). sending a stream error with stream id 0 is not allowed by the spec https://tools.ietf.org/html/rfc7540#section-6.4. We should also get the \"real\" stream ID to ease understanding.\n\n@ejona86 - what is the best way to do this?\n", "`0` -> `id` (assuming id eventually represents the \"real\" stream id)\n", "@Scottmitch, I've got to go to a meeting, so I can't dig too deeply. But there shouldn't be much reason to need the id with the childchan API. If we really have to, we could grab it from an attribute (which currently doesn't exist) on the channel.\n", "Is the intention of this conditional to distinguish between the \"headers\" and the \"trailers\"? Perhaps a better way to do this is to check the stream state. @ejona86 - is there a way to get the stream state in the new API?\n\nAlso as a side note @nmittler I don't think we are currently enforcing this (are we?):\n\nhttps://tools.ietf.org/html/rfc7540#section-8.1.2.3\n\n> All HTTP/2 requests MUST include exactly one valid value for the\n> \":method\", \":scheme\", and \":path\" pseudo-header fields, unless it is\n> a CONNECT request (Section 8.3). An HTTP request that omits\n> mandatory pseudo-header fields is malformed (Section 8.1.2.6).\n", "https://tools.ietf.org/html/rfc7540#section-8.1.2.6\n\n> Malformed requests or responses that are\n> detected MUST be treated as a stream error (Section 5.4.2) of type\n> PROTOCOL_ERROR.\n> \n> For malformed requests, a server MAY send an HTTP response prior to\n> closing or resetting the stream. Clients MUST NOT accept a malformed\n> response. Note that these requirements are intended to protect\n> against several types of common attacks against HTTP; they are\n> deliberately strict because being permissive can expose\n> implementations to these vulnerabilities.\n\nSo maybe its good we don't enforce this on the server ... either it seems best to avoid using the presence/absence of these headers fields to indicate state of the stream.\n", "typically the `validateHeaders` is exposed and configurable by the user. validating headers can add non-trivial overhead that some users decide to forgo.\n", "also expose the `validateHeaders` so it can be configured by users (same as https://github.com/netty/netty/pull/5204#discussion_r62077393)\n", "expose the `validateHeaders` so it can be configured by users (same as https://github.com/netty/netty/pull/5204#discussion_r62077393)\n", "@ejona86 - is the idea that the exception should propagate through the child channel's pipeline and translated to the appropriate type of HTTP/2 frame by the http/2 codec under the hood? In the long term this seems like a good thing (assuming we can distinguish between connection/stream errors if necessary)... in the near term our Http2Exception requires the StreamId ... and if the connection stream id is used a stream error will be translated into a connection error (see [Http2Exception](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Exception.java#L124)) ... we should have some story to handle this in the near term ... wdyt?\n", "> Perhaps a better way to do this is to check the stream state. @ejona86 - is there a way to get the stream state in the new API?\n\nNo, there isn't, and currently we don't track this sort of state anywhere (even in the old API). It could be something we add, but I'd hope to not need it, especially given how hard it will be to resolve races. If the handler wasn't shareable it'd be trivial to deal with, but shareable is a laudable goal.\n\nI don't have too much concern about relying on :method for requests and :status for responses, since they are so core to HTTP and _must_ be present in HTTP/1 during parsing. That would even behave correctly in the face of 1xx status codes where you end up having multiple sets of \"headers\". (For those unaware of this particular dark corner of HTTP, multiple header sets is a real thing in both [HTTP/1.1](https://tools.ietf.org/html/rfc7231#section-6.2) and [HTTP/2](https://tools.ietf.org/html/rfc7540#section-8.1). Yay!)\n\nAnother option is to add a bit to Http2HeadersFrame indicating whether it is a \"header\" or \"trailer,\" but that sounds a bit icky because it has to agree with endStream.\n", "So I looked into this some. The value has two usages: stream exceptions and converting to non-childchan HTTP/2. Converting to non-childchan HTTP/2 doesn't seem important.\n\nExceptions on a stream's channel have to be handled like normal exceptions; in this case with exceptionCaught. If uncaught it just logs. Applications will commonly write exceptionCaught to close the channel, which then becomes [Http2ResetFrame(CANCEL)](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/Http2MultiplexCodec.java#L413)). An application could write its own Http2ResetFrame with special error code if it wanted.\n\nSo the Http2Exceptions aren't special with h2childchan. That's true even if you're on the parent channel. And in fact it's true in the non-childchan API as well when you are later in the pipeline than the Http2ConnectionHandler; it's just that that almost never happens.\n", "> No, there isn't, and currently we don't track this sort of state anywhere (even in the old API)\n\nIn the existing API we generally pass around the stream ID or the Http2Stream object. If you have the stream ID then you can use `Http2Connection.stream(id)` to get the Http2Stream. The Http2Stream allows access to the [State](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Stream.java#L68).\n\n> I don't have too much concern about relying on :method for requests and :status for responses, since they are so core to HTTP and must be present in HTTP/1 during parsing\n\nConsider the following sequence of messages:\n\nHeaders (:method=get) StreamId=3 EOS=0\nData StreamId=3 EOS=0\nHeaders (:method=get) StreamId=3 EOS=1\n\nThis is likely a separate issue but where do we enforce that these \"special\" headers can only exist \"exactly once\" before the \"data\" (unless its a connect request)?\n", "Yes this seems reasonable ... only concern is if a childchan uses an API which intends to throw a \"Connection Error\" (which would typically be converted into a GO_AWAY) ... although it seems like most of these scenarios should be on the parent channel decoding stuff before it gets to the childchan\n", "nit: `this(true)`\n", "done, thanks!\n", "> The Http2Stream allows access to the State.\n\nBut that state is 1) not directly usable because it has already been modified in response to the frames, 2) this logic using the state may be delayed, even on the same thread and 3) this logic using the state may be run from a different thread from where the state is maintained/changed.\n\nFor point #1, consider these new streams:\n\nIDLE: (isn't actually tracked yet; it will [never](https://github.com/netty/netty/blob/bd6040a/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java#L311) show up as IDLE)\nOPEN: Headers (:method=get) StreamId=3 EOS=0 # These are headers\nOPEN: Headers () StreamId=3 EOS=1 # These are trailers\nHALF_CLOSED_REMOTE: (yes, it happens [afterward](https://github.com/netty/netty/blob/83c349f/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2ConnectionDecoder.java#L324))\n\nIDLE: (isn't actually tracked yet; it won't show up as IDLE)\nHALF_CLOSED_REMOTE: Headers (:method=get) StreamId=5 EOS=1 # These are headers\n(yes, this state happens [immediately](https://github.com/netty/netty/blob/bd6040a/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2Connection.java#L315))\n\nSo the State is _never_ useful for this distinction today. That could possibly be fixed, but the child channel has its own queue. The application could simply not read from the child channel immediately and the state will be HALF_CLOSED_REMOTE in both cases. When the child channel is using a different event loop, we'd expect HALF_CLOSED_REMOTE in both cases.\n\n> This is likely a separate issue but where do we enforce that these \"special\" headers can only exist \"exactly once\" before the \"data\" (unless its a connect request)?\n\nIt looks like we should be prohibiting that (still taking into account 1xx):\n\n> Pseudo-header fields are only valid in the context in which they are defined. Pseudo-header fields defined for requests MUST NOT appear in responses; pseudo-header fields defined for responses MUST NOT appear in requests. Pseudo-header fields MUST NOT appear in trailers. Endpoints MUST treat a request or response that contains undefined or invalid pseudo-header fields as malformed (Section 8.1.2.6).\n", "I would hope the Http2Exceptions would go away (or be an implementation detail) in the child chan API (for both child channel and parent channel). They stop being helpful for control flow when you have multiple handlers, since the exceptions are propagated toward the tail of the pipeline, not the head. They could still be used for notifications of failures, but it seems they couldn't trigger RST_STREAM and GOAWAY like they do today.\n", "Agreed the state as it exists today won't be very useful for this purpose\n", "`full.content().readableBytes() == 0` -> `!full.content().isReadable()`\n", "why not make non-static and just use `validateHeaders` as a member variable instead of as a parameter?\n", "its not clear why we should try to simulate the user sending a data frame to make the encoding work. Seems more intuitive to just encode what the user provides and if their logic is flawed they should fix it. Can we just mimic the logic in [HttpToHttp2ConnectionHandler.write](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/HttpToHttp2ConnectionHandler.java) or improve the logic their if it is not good enough? It would be nice to share code or at least use consistent logic if possible.\n", "We need to handle the case where a user sends a `LastHttpContent.EMPTY_LAST_CONTENT` in order to close their streamed message, but there's no need to send a message otherwise. I don't think this is incorrect behavior.\n\nIt looks like in HttpToHttp2ConnectionHandler, it sends a data frame regardless of whether there's content or not, which seems unnecessary. I can clean it up, do you mind if we push to a subsequent review though?\n", "> I don't think this is incorrect behavior.\n\nAgreed. Thanks for clarifying.\n\n> I can clean it up, do you mind if we push to a subsequent review though?\n\nFollowup PR sgtm thanks!\n", "@mosesn - I just merged this an noticed a leak in the build log. can you add another commit (or pr) where these `request` objects are released in a finally block for all these tests?\n", "yep, how can I check if I'm leak-free?\n", "The best thing we have ATM is [-Dio.netty.leakDetectionLevel=paranoid](http://netty.io/wiki/reference-counted-objects.html#leak-detection-levels).\n", "thanks :)\n", "@mosesn run a few times with `-Pleak` and see if you see any `LEAK:` logs in the console\n", "note that you can also use [releaseLater](http://netty.io/wiki/reference-counted-objects.html#fixing-leaks-in-unit-tests) instead of using a `try`/`finally` block.\n", "I've run it a few times, but can't repro locally. Here's what I'm doing:\n\n``` bash\nmvn test -pl codec-http2 -Dtest=Http2ServerDowngraderTest -Pleak\n```\n\nIs there anything I'm missing? Is there a way I can see the failed build, so I can fix based on that?\n", "@mosesn - the build won't fail locally, but you should see log statements that look like this https://garage.netty.io/teamcity/viewLog.html?tab=buildLog&logTab=tree&filter=debug&expand=all&buildId=8253#_focus=18593\n", "Weird, I don't see that failure locally. Do you see it consistently? Do I need to change a log level somewhere?\n", "Not sure why, but it just started working locally. Working on a fix now–looks like it's just for one test.\n", "I made a PR, but I'm not sure if this is the right fix: #5230.\n" ]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/InboundHttp2ToHttpAdapter.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ServerDowngrader.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/HttpConversionUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/InboundHttp2ToHttpAdapter.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2ServerDowngraderTest.java" ]

[ "timestamp(timedelta=22.0, similarity=0.8452019802756294)" ]

[ "@wanganran - Thanks for reporting.\n\n@trustin - Looks like this was introduced in https://github.com/netty/netty/commit/ada61d4985571f5406ea6158b02755d51392ebff . Can you take care?\n", "will take care\n", "I think NioDatagramChannel should report an error: https://github.com/netty/netty/blob/4.1/transport/src/main/java/io/netty/channel/socket/nio/NioDatagramChannel.java#L104\n", "Fixed by https://github.com/netty/netty/pull/5210\n" ]

[]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/socket/nio/NioDatagramChannelConfig.java" ]

[ "transport/src/main/java/io/netty/channel/socket/nio/NioDatagramChannelConfig.java" ]

[]

[ "timestamp(timedelta=61.0, similarity=0.9070950966957434)" ]

[ "@nayato you are right... let me fix it.\n", "Fixed https://github.com/netty/netty/pull/5219\n" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=11.0, similarity=0.8916239172691428)" ]

[ "That's a good point. If a user specifies a promise when registering, a user doesn't really need to specify a channel because an event loop can get the channel from the specified promise. Let me fix this in 4.1.\n", "@colalife @trustin makes sense\n", "really encourage me，btw，i see it implements from EventLoopGroup，some classes have it\n", "Are you suggesting that `EventLoop` extending `EventLoopGroup` is a problem, or..?\n", "I think the reasonable fix for this issue is to remove `register(Channel, ChannelPromise)` (with proper deprecation,) and let people always use `register(Channel)`. What do you think?\n", "@trustin +1\n", "yeah, i agree, and of course other classes et. MultithreadEventLoopGroup... who implement it\n", "@trustin I guess we can only do this for 5.0.0 as it will break the api\n", "@trustin do we want to \"fix this\" for 4.1.0.Final or not ?\n", "Fixed by https://github.com/netty/netty/pull/5231\n" ]

[]

[ "defect" ]

[ "transport/src/main/java/io/netty/channel/EventLoopGroup.java", "transport/src/main/java/io/netty/channel/MultithreadEventLoopGroup.java", "transport/src/main/java/io/netty/channel/SingleThreadEventLoop.java", "transport/src/main/java/io/netty/channel/ThreadPerChannelEventLoop.java", "transport/src/main/java/io/netty/channel/ThreadPerChannelEventLoopGroup.java", "transport/src/main/java/io/netty/channel/embedded/EmbeddedEventLoop.java", "transport/src/main/java/io/netty/channel/oio/OioEventLoopGroup.java" ]

[ "transport/src/main/java/io/netty/channel/EventLoopGroup.java", "transport/src/main/java/io/netty/channel/MultithreadEventLoopGroup.java", "transport/src/main/java/io/netty/channel/SingleThreadEventLoop.java", "transport/src/main/java/io/netty/channel/ThreadPerChannelEventLoop.java", "transport/src/main/java/io/netty/channel/ThreadPerChannelEventLoopGroup.java", "transport/src/main/java/io/netty/channel/embedded/EmbeddedEventLoop.java", "transport/src/main/java/io/netty/channel/oio/OioEventLoopGroup.java" ]

[ "transport/src/test/java/io/netty/bootstrap/BootstrapTest.java", "transport/src/test/java/io/netty/channel/SingleThreadEventLoopTest.java", "transport/src/test/java/io/netty/channel/ThreadPerChannelEventLoopGroupTest.java" ]

[ "timestamp(timedelta=23.0, similarity=0.890638655932035)" ]

[ "the underlying storage is currently exposed by `options` and `attr` and there is synchronization that must be applied when dealing with these collections. I'm guessing they are not exposed so the storage can be directly exposed (no copy etc...) and its assumed users of the interface \"play nice\" with synchronization and modifiability. We should consider these factors before making changes to these interfaces.\n", "Yeah I think the synchronisation is an issue. That said maybe we could just expose a copy of the attributes / options. Let me cook up something...\n", "Fixed by https://github.com/netty/netty/pull/5243\n" ]

[ " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", "`toString()` please\n", "`AbstractBootstrapConfig` and make it top-level?\n", "Could we move this class top-level?\n", "Could we move this to top level?\n", "`toString()`\n", "- u -> U\n- `{@link #config()}.` -> `{@link #config()} instead.`\n", "- u -> U\n- . -> instead.\n", "Global comment: Return's'\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n", " Remove usage of generic wildcard type. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1452)\n" ]

[ "improvement", "feature" ]

[ "transport/src/main/java/io/netty/bootstrap/AbstractBootstrap.java", "transport/src/main/java/io/netty/bootstrap/Bootstrap.java", "transport/src/main/java/io/netty/bootstrap/ServerBootstrap.java" ]

[ "transport/src/main/java/io/netty/bootstrap/AbstractBootstrap.java", "transport/src/main/java/io/netty/bootstrap/AbstractBootstrapConfig.java", "transport/src/main/java/io/netty/bootstrap/Bootstrap.java", "transport/src/main/java/io/netty/bootstrap/BootstrapConfig.java", "transport/src/main/java/io/netty/bootstrap/ServerBootstrap.java", "transport/src/main/java/io/netty/bootstrap/ServerBootstrapConfig.java" ]

[]

[ "timestamp(timedelta=16.0, similarity=0.8965582301594414)" ]

[ "@nayato there is no good reason... let me fix it.\n", "Fixed by https://github.com/netty/netty/pull/5251\n" ]

[]

[ "cleanup" ]

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

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

[]

[ "timestamp(timedelta=24.0, similarity=0.8564786573284091)" ]

[ "@Underbalanced we love contributions ;)\n", "I will read up on the how to contribute as well as your code, I am pretty new to using proper design patterns. Started back in April with java. Though I have learned enough to write a Craftbukkit like server tool for another game, I skipped some important fundamentals.\n", "Fixed by https://github.com/netty/netty/pull/5255\n" ]

[ "that match -> that are matched by\n\n.. because it's not a channel that matches a matcher?\n" ]

[ "feature" ]

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

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

[]

[ "timestamp(timedelta=18.0, similarity=0.9098434823812074)" ]

[ "Fixed by https://github.com/netty/netty/pull/5265\n" ]

[ "use `new LinkedHashSet<...>(domanNamePatterns.length)`\n", "Also ensure this is read-only ?\n", "Added the length hint. Is there any `Map.Entry` implementation in netty?\n", "> Also ensure this is read-only ?\n\nDone.\n" ]

[ "improvement", "feature" ]

[ "common/src/main/java/io/netty/util/DomainMappingBuilder.java", "common/src/main/java/io/netty/util/DomainNameMapping.java" ]

[ "common/src/main/java/io/netty/util/DomainMappingBuilder.java", "common/src/main/java/io/netty/util/DomainNameMapping.java" ]

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

[ "timestamp(timedelta=28.0, similarity=0.9338759135638064)" ]

[ "This is the patch I wrote for [Armeria](https://github.com/line/armeria):\n\nhttps://github.com/line/armeria/pull/155/files\n\nAs shown in the PR above, we could write a decorating `NameResolver` that works for all `NameResolver` implementations. However, I'm not sure what the name of that class would be. Any suggestions?\n", "/cc @slandelle \n", "InflightDnsNameResolver ? Not sure its the best name though\n\n> Am 27.04.2016 um 12:18 schrieb Trustin Lee [email protected]:\n> \n> This is the patch I wrote for Armeria:\n> \n> https://github.com/line/armeria/pull/155/files\n> \n> As shown in the PR above, we could write a decorating NameResolver that works for all NameResolver implementations. However, I'm not sure what the name of that class would be. Any suggestions?\n> \n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly or view it on GitHub\n", "@trustin WDYT ?\n", "FlyingQueriesAwareDnsNameResolver ? Pro is that such name will make Spring people jealous ;-) \n", "InProgressAware?\n", "@trustin also we can just make it package private for now and find a good name later if we feel we not have a good one atm.\n", "Alright. Let me cook some tomorrow.\n", "any news ?\n", "Fixed by https://github.com/netty/netty/pull/5267\n" ]

[ " Take the required action to fix the issue indicated by this comment. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1134)\n" ]

[ "defect" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsAddressResolverGroup.java" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsAddressResolverGroup.java", "resolver-dns/src/main/java/io/netty/resolver/dns/InflightNameResolver.java" ]

[]

[ "timestamp(timedelta=45.0, similarity=0.876440567581754)" ]

[ "At the moment not through this interface. What type of socket channel are you dealing with as there is likely a way to do it using an implementation of this interface.\n", "NioSocketChannel, EpollSocketChannel or OioSocketChannel\n", "I see you want it all ... stand by for a PR.\n", "Fixed by https://github.com/netty/netty/pull/5269\n" ]

[]

[ "feature" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-rxtx/src/main/java/io/netty/channel/rxtx/RxtxChannel.java", "transport-udt/src/main/java/io/netty/channel/udt/nio/NioUdtByteConnectorChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioByteChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioMessageChannel.java", "transport/src/main/java/io/netty/channel/oio/AbstractOioByteChannel.java", "transport/src/main/java/io/netty/channel/socket/DuplexChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/oio/OioSocketChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java", "transport-rxtx/src/main/java/io/netty/channel/rxtx/RxtxChannel.java", "transport-udt/src/main/java/io/netty/channel/udt/nio/NioUdtByteConnectorChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioByteChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioChannel.java", "transport/src/main/java/io/netty/channel/nio/AbstractNioMessageChannel.java", "transport/src/main/java/io/netty/channel/oio/AbstractOioByteChannel.java", "transport/src/main/java/io/netty/channel/socket/DuplexChannel.java", "transport/src/main/java/io/netty/channel/socket/nio/NioSocketChannel.java", "transport/src/main/java/io/netty/channel/socket/oio/OioSocketChannel.java" ]

[ "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollSocketShutdownOutputByPeerTest.java", "transport-native-epoll/src/test/java/io/netty/channel/epoll/EpollSocketShutdownOutputBySelfTest.java" ]

[ "timestamp(timedelta=28.0, similarity=0.8719758372520758)" ]

[ "Note that those `Submitter`s are just `int` wrappers, so I think the leak is more of an annoying warning than a severe issue.\n", "Also the Netty `ForkJoinPool` is just a copy of the OpenJDK FJP. So maybe a simple up date to the latest OpenJDK version would resolve this.\n", "@buchgr I wonder if it's actually a fork of Doug Lea's initial work for JSR166-e, before it was merged into OpenJDK and modified again.\n\nNetty's version has this [ThreadLocal<Subscriber>](https://github.com/netty/netty/blob/netty-4.0.31.Final/common/src/main/java/io/netty/util/internal/chmv8/ForkJoinPool.java#L1082) than [OpenJDK hasn't](http://hg.openjdk.java.net/jdk8/jdk8/jdk/file/687fd7c7986d/src/share/classes/java/util/concurrent/ForkJoinPool.java#l1098).\n", "@slandelle good point. I am not exactly sure. Just saying that updating it might resolve such things.\n", "@buchgr I'm not sure you guys will want to fork OpenJDK as it's GPLv2. You might want to use [Doug Lea's JSR166](http://g.oswego.edu/dl/concurrency-interest/) work that's Creative Commons - public domain.\n", "But I agree, upgrading should fix the issue as Tomcat doesn't report it against regular CHMv8.\n", "@slandelle - Thanks for the heads up!\n\n@normanmaurer - I'll assign to you for now...pending discussion about 5.0.\n", "Fixed by https://github.com/netty/netty/pull/5270\n" ]

[]

[ "defect" ]

[ "common/src/main/java/io/netty/util/internal/chmv8/ForkJoinPool.java" ]

[ "common/src/main/java/io/netty/util/internal/chmv8/ForkJoinPool.java" ]

[]

[ "timestamp(timedelta=22.0, similarity=0.9284216160032346)" ]

[ "PR https://github.com/netty/netty/pull/5321\n", "Fixed by https://github.com/netty/netty/pull/5321\n" ]

[ "Please define a static byte[] and reuse everytime. Something like:\n\nprivate static final byte[] ID = new byte[] {'n', 'e', 't', 't', 'y'};\n" ]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=18.0, similarity=0.9191116549719136)" ]

[ "@MilosFabian we love PRs.. so maybe you can provide one as I don't have any good OSGI knowledge :(\n", "Alright, I will work on it :) \n", "Thanks ❤️!\n\n> Am 12.04.2016 um 09:33 schrieb MilosFabian [email protected]:\n> \n> Alright, I will work on it :)\n> \n> —\n> You are receiving this because you commented.\n> Reply to this email directly or view it on GitHub\n", "I am sorry for a delay...finally I found some time take a look at it.\nI managed produce bundle (include manifest in jar) for netty-transport-native-epoll-*-linux-x86_64\nHowever, when I was trying to use the bundle in OSGi (Apache Karaf) I hit a problem when loading native library.\nThe Epoll.isAvailable() returns false with following unavailability cause.\n\njava.lang.ClassNotFoundException: io.netty.channel.epoll.NativeStaticallyReferencedJniMethods\n at org.eclipse.osgi.internal.loader.BundleLoader.findClassInternal(BundleLoader.java:501)\n at org.eclipse.osgi.internal.loader.BundleLoader.findClass(BundleLoader.java:421)\n at org.eclipse.osgi.internal.loader.BundleLoader.findClass(BundleLoader.java:412)\n at org.eclipse.osgi.internal.baseadaptor.DefaultClassLoader.loadClass(DefaultClassLoader.java:107)\n at java.lang.ClassLoader.loadClass(ClassLoader.java:357)[:1.8.0_45]\n at java.lang.ClassLoader$NativeLibrary.load(Native Method)[:1.8.0_45]\n at java.lang.ClassLoader.loadLibrary0(ClassLoader.java:1937)[:1.8.0_45]\n at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1822)[:1.8.0_45]\n at java.lang.Runtime.load0(Runtime.java:809)[:1.8.0_45]\n at java.lang.System.load(System.java:1086)[:1.8.0_45]\n at io.netty.util.internal.NativeLibraryLoader.load(NativeLibraryLoader.java:214)\n at io.netty.channel.epoll.Native.loadNativeLibrary(Native.java:261)\n at io.netty.channel.epoll.Native.<clinit>(Native.java:62)\n at io.netty.channel.epoll.Epoll.<clinit>(Epoll.java:33)\n\nMore investigation needed.\n", "Fixed by https://github.com/netty/netty/pull/5326\n" ]

[]

[ "defect" ]

[ "common/pom.xml", "transport-native-epoll/pom.xml" ]

[ "common/pom.xml", "transport-native-epoll/pom.xml" ]

[]

[ "timestamp(timedelta=2632.0, similarity=0.9519604256868078)" ]

[ "@maseev I think you should try latest 4.1 branch and see how it behaves as we now also pool Slices if you use the PooledByteBufAllocator here.\n", "@maseev maybe you want to send a PR which implements an \"optimised\" `LengthFieldBasedFrameDecoder` and `FixedLengthFrameDecoder` ?\n", "I don't think your handler can be Sharable with that buffer bytebuf. \n" ]

[ "Maybe the check can be moved to some constructor?\n", "No need to assign writeIndex here I guess?\n", "Maybe I am missing something, but I assume a decoder like this one may throw every frame decoded to out via buffer.retainSlice(), etc. I could not quite follow the point here, even after reading the tests. Correct me if I am wrong. :-)\n", "nit: mismatch of the if condition and the exception message?\n", "I don't think so. Either way, you have to call an appropriate method which depends on an offset value. The same approach is used in LengthFieldBasedFrameDecoder. Seems legit for me.\n", "Thanks! Will fix that very soon.\n", "Do you mean `buffer.retain().slice()`? Well, it's not how it supposed to be working. The whole point of this decoder is to batch all incoming ByteBuf messages into one preallocated ByteBuf object. When the `channelReadComplete` method gets called we simply copy data from the buffer to another chunk buffer which would contain precisely N messages (where N >= 1) and pass it to next pipeline element. By doing so we're able to reuse this buffer, again and again, no matter how many messages we get. If I chose retain-slice approach I would have to keep an eye on the buffer reference counter because slice creates only a view of buffer sub-region, so you wouldn't be able to reuse it, if for example, the channel wasn't ready for write (channel.isWritable() returns false).\n", "See my comment below.\n" ]

[]

[]

[ "codec/src/main/java/io/netty/handler/codec/BulkFixedLengthFrameDecoder.java", "codec/src/main/java/io/netty/handler/codec/BulkFrameDecoder.java", "codec/src/main/java/io/netty/handler/codec/BulkLengthFieldBasedFrameDecoder.java" ]

[ "codec/src/test/java/io/netty/handler/codec/BulkFixedLengthFrameDecoderTest.java", "codec/src/test/java/io/netty/handler/codec/BulkLengthFieldBasedFrameDecoderTest.java" ]

[ "timestamp(timedelta=14.0, similarity=0.8851513607779111)" ]

[ "@nmittler @ejona86 - WDYT?\n", "@Scottmitch Yeah I think that seems reasonable. I suspect we'll want to set some sort of timeout on how long we wait for all of the streams to close before sending a `hard` `GOAWAY` and just closing the connection. Having all of this \"pluggable\" in some way might be handy.\n", "@Scottmitch SGTM\n", "> I suspect we'll want to set some sort of timeout on how long we wait for all of the streams to close before sending a hard GOAWAY\n\nWhen close is called we already have a timeout mechanism [gracefulShutdownTimeoutMillis](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/Http2ConnectionHandler.java#L73) which attempts to wait for all streams to be closed before the channel is actually closed. Can you clarify if you think we need more than this, and what you mean by `hard GOAWAY`?\n", "@Scottmitch ah right, I had forgotten about that :). SGTM\n" ]

[ "Should the comment above be moved this this `else`?\n", "Maybe add a comment here to describe what you're doing?\n", "I will clarify the comments.\n", "I will clarify the comments.\n" ]

[ "improvement" ]

[ "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=13.0, similarity=0.9126192891211647)" ]

[ "@rkapsi - Would you be interested in submitting a PR?\n", "@Scottmitch On it...\n\n... but `io.netty.util.collection.*` seem to be missing on the 4.1 branch?!\n", "@rkapsi its generated ... run mvn clean compile in the common module\n", "@normanmaurer thanks!\n", "Fixed by https://github.com/netty/netty/pull/5377\n" ]

[ "Pass a `ByteBufAllocator` as a parameter and use it to allocate this buffer. We can use `UnpooledByteBufAllocator.DEFAULT` by default but at least this gives control over allocations done in this method.\n", "Pass a `ByteBufAllocator` as a parameter and use it to allocate this buffer. We can use `UnpooledByteBufAllocator.DEFAULT` by default but at least this gives control over allocations done in this method.\n", "if `pem.content().isDirect()` is true we should be able to just use `pem.content().retainedSlice()` instead of allocating a new buffer and copying.\n", "we may also want have the new classes that construct the PEM cert/key allocate direct by default.\n", "not introduced by this PR but if we adjust the ordering of events such that the encoding is done first ... we should know the exact size to allocate `pem` (`BEGIN_CERT.length + base64.readableBytes() + END_CERT.length`)\n", "discussion: is this necessary ATM? I wonder if we can do with out this for now?\n", "nit: add some descriptive text? `content is not accessible after this object has been destroyed`\n", "Please elaborate. You mean classes such as `SelfSignedCertificate` create and return PemX509Certificates?\n", "any reason why we can't implement this?\n", "`PKCS#8` \nhttps://docs.oracle.com/javase/7/docs/api/java/security/Key.html#getFormat()\n", "can/should we support this based upon https://docs.oracle.com/javase/7/docs/api/java/security/Key.html#getAlgorithm() ?\n", "`PemX509Certificate` and `PemPrivateKey` currently use `Unpooled.buffer()` to allocate the buffer they use for internal storage. However we could change this to `ByteBufAllocator.directBuffer(...)` to make it more likely that `pem.content().isDirect()` is true and thus avoid the copy operation.\n", "is it correct to return 0 if the length is 0 ? If so we need to update the javadocs.\n", "move this in another finally block as just to be safe ?\n", "to a null check via ObjectUtil.checkNonNull(...) ?\n", "maybe you could just extend `DefaultByteBufHolder` and so reuse some code ?\n", "directly call content.touch(...)\n", "+1\n", "+1\n", "Use javadoc notion and use {@link...} \n\nSame above\n", "Again why not extend `DefaultByteBufHolder` ?\n", "null checks\n", "null check\n", "@normanmaurer `DefaultByteBufHolder` doesn't have a `#destroyed()` method which it used to fill the ByteBuf with zero bytes if the value is sensitive (PrivateKey).\n", "and ? What is the problem to just add it to PemValue ? \n", "How about throwing an IllegalArgumentException if length==0?\n", "yep sounds good\n", "So, `PrivateKey` implements that `Destroyable` interface but it doesn't say when it was introduced. Netty's build agent didn't like my `@Overrides`.\n", "Something like... extend and then...?\n\n``` java\n@Override\npublic boolean release() {\n if (super.release()) {\n // zero out\n return true;\n }\n return false;\n}\n```\n", "nit: just pass `ByteBufAllocator.DEFAULT` directly ... no need for temp variable (even if we think JDK may optimize it away)\n", "Actually that will not work, `DefaultByteBufHolder#release()` returns true/false after it has released the underlying ByteBuf at which point it's no longer possible for me to wipe the bytes.\n", "can this actually happen?\n", "The alloc is passed ~2 lines further down into the `toBIO()` call. Feels more appropriate to have it as a temp variable and ideally passed in as method argument (but that would break existing API). Happy to change it to `ByteBufAllocator.DEFAULT` though.\n", "if the constructor of PemValue were to throw for some reason, it seems like we would leak the buffer. Consider adding comments on the constructor that it must take ownership of the buffer, or handle the release here:\n\n``` java\nPemEncoded result = new PemValue(pem, false);\nsuccess = true;\nreturn result;\n```\n", "nit: consider using a tertiary statement.\n", "only if `chain.length==0` is passed in. Happy to turn that into a precondition that throws an IllegalArgumentException or so.\n", "Ah ok... Sorry for the noise\n\n> Am 09.06.2016 um 19:28 schrieb Roger [email protected]:\n> \n> In handler/src/main/java/io/netty/handler/ssl/PemValue.java:\n> \n> > - _/\n> > +package io.netty.handler.ssl;\n> > +\n> > +import io.netty.buffer.ByteBuf;\n> > +import io.netty.buffer.ByteBufAllocator;\n> > +import io.netty.util.AbstractReferenceCounted;\n> > +import io.netty.util.IllegalReferenceCountException;\n> > +\n> > +/_*\n> > - \\* A PEM encoded value.\n> > - *\n> > - \\* @see PemEncoded\n> > - \\* @see PemPrivateKey#toPEM(ByteBufAllocator, boolean, java.security.PrivateKey)\n> > - \\* @see PemX509Certificate#toPEM(ByteBufAllocator, boolean, java.security.cert.X509Certificate[])\n> > - */\n> > +class PemValue extends AbstractReferenceCounted implements PemEncoded {\n> > Actually that will not work, DefaultByteBufHolder#release() returns true/false after it has released the underlying ByteBuf at which point it's no longer possible for me to wipe the bytes.\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "can this be package private?\n", "just for content :)\n", "I missed that it was used in 2 spots originally ... thanks for clarifying ... I'm fine leaving it as is.\n", "is this necessary any more now that we don't implement `Destroyable`?\n", "nvm ... `PrivateKey` brings `Destroyable` in to the heirarchy ... add `@Override`?\n", "add `@Override`?\n", "yah lets handle that case explicitly as a precondition then\n", "I'm a bit on the fence. Java's PrivateKey interface extends Destroyable. No idea when that was introduced. I assume JDK8 because its got `default` impls. straight on the interface. Yesterday's build errors were because of that.\n", "consider simplifying this `try` block to:\n\n``` java\n boolean success = false;\n final ByteBuf pem = useDirect ? allocator.directBuffer(size) : allocator.buffer(size);\n try {\n pem.writeBytes(BEGIN_PRIVATE_KEY);\n pem.writeBytes(base64);\n pem.writeBytes(END_PRIVATE_KEY);\n\n PemValue value = new PemValue(pem, true);\n success = true;\n return value;\n } finally {\n // Make sure we never leak that PEM ByteBuf if there's an Exception.\n if (!success) {\n SslUtils.zerooutAndRelease(pem);\n }\n }\n```\n", "The build agent didn't like that yesterday which lead to that error\n\n```\n[Compiler] Compilation failure\n where T is a type-variable:\n T extends Object declared in method <T>attr(AttributeKey<T>)\n/var/lib/teamcity-agent/work/a00bdf77e621d005/handler/src/main/java/io/netty/handler/ssl/PemPrivateKey.java:[178,4] error: method does not override or implement a method from a supertype\n```\n", "@rkapsi - Good point. `PrivateKey` started extending `Destroyable` in [JDK8](http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8-b132/java/security/PrivateKey.java/). I recall you had the `//@Override // JDK8` and I now understand why :) ... Maybe add something like this back `//@Override will only in work with JDK8 but Netty must be JDK6 compatible so we can't have nice things`\n", "+1 that makes sense ... we can discuss above in https://github.com/netty/netty/pull/5377#discussion_r66490755\n", "+1\n", "sgtm\n", "@rkapsi just address this comment and I will pull in. Great work!\n" ]

[ "improvement", "feature" ]

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

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslContext.java", "handler/src/main/java/io/netty/handler/ssl/PemEncoded.java", "handler/src/main/java/io/netty/handler/ssl/PemPrivateKey.java", "handler/src/main/java/io/netty/handler/ssl/PemValue.java", "handler/src/main/java/io/netty/handler/ssl/PemX509Certificate.java", "handler/src/main/java/io/netty/handler/ssl/SslUtils.java" ]

[]

[ "timestamp(timedelta=22.0, similarity=0.8790954761111075)" ]

[ "Splitted the proposal to new issue: https://github.com/netty/netty/issues/5352\n", "@mickare - Thanks for bringing this to our attention. The javadocs should be updated. Let me take care.\n" ]

[]

[ "documentation" ]

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

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

[]

[ "timestamp(timedelta=299.0, similarity=0.868294864826654)" ]

[ "@slandelle you are right... let me fix.\n", "Awesome, thanks!\nI can provide a PR if you want, but I first wanted to be sure it was a bug and not intended behavior.\n", "@slandelle a pr would be awesome! Please also add a test that we catch when `ChannelFactory.newChannel(...)` throws and notify the promise.\n", "PR on its way\n", "Actually, having a look at `BootstrapTest#testLateRegistrationConnect` makes me think that `connect` is expected to possibly throw exceptions.\n\nI can't say I like this behavior, but it is what it is, and IMHO can't be changed/fixed outside a new version.\n\nWDYT?\n", "Should Exception be thrown AND listener notified?\n", "@slandelle I think you misread the test case. It only throws because we do `syncUninterruptibly()`.\n", "My bad...\n", "@slandelle no worries!\n", "Done, see #5388\n", "Fixed by https://github.com/netty/netty/pull/5388\n" ]

[ "Please create the exception before the bootstrap and just throw it. This way you can use `assertSame(...)` below to check if the same exception. This is more robust compared to check for the message.\n", "remove empty line\n", "See below.\n", "See above.\n" ]

[ "defect" ]

[ "transport/src/main/java/io/netty/bootstrap/AbstractBootstrap.java" ]

[ "transport/src/main/java/io/netty/bootstrap/AbstractBootstrap.java" ]

[ "transport/src/test/java/io/netty/bootstrap/BootstrapTest.java" ]

[ "timestamp(timedelta=73.0, similarity=0.882209135474392)" ]

[ "@mrokitka wow there are still people using HTTP/1.0 ? You are right we are currently only support it with HTTP/1.1, not sure if we will ever support HTTP/1.0. But at least we should \"skip\" HTTP/1.0 messages. Let me come up with a PR for this\n", "@normanmaurer - I've run into this since we're using Squid as a caching proxy to handle caching of content for outbound proxy requests from a Netty based proxy server I'm implementing. Squid doesn't fully support HTTP 1.1, so although it handles HTTP compression properly responses are always HTTP 1.0.\n\nKind of a drag, but if the compressor could handle the compression for HTTP 1.0 without the need for chunked encoding that would be great.\n", "@mrokitka the problem with this is that we would need buffer the full content in memory as we will need to set the \"Content-Length\" header. We could do this but not sure if this is really good :)\n", "@normanmaurer perhaps a flag that could be programmatically set on the compressor to indicate if we want to support HTTP 1.0? A true value would compress and buffer to set the content-length header for 1.0 responses, false would just skip 1.0 messages?\n", "@mrokitka maybe... let me see what I can come up with.\n", "Fixed by https://github.com/netty/netty/pull/5405 . \n\n@mrokitka I may come up with a follow up to support compression for http 1.0 as well but currently very busy. If you like you can also contribute a PR.\n", "https://github.com/netty/netty/pull/5405\n" ]

[]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=12.0, similarity=0.9073717107325305)" ]

[ "@vietj so what ? I think there is not much we can do without doing some crazy jni stuff :(\n", "now in Vert.x we simply resolve localhost to `127.0.0.1` on Windows. Perhaps doing the same would make sense in Netty too.\n", "Yeah maybe… and logging a warning. Let me do this.\n\n> On 13 Jun 2016, at 11:49, Julien Viet [email protected] wrote:\n> \n> now in Vert.x we simply resolve localhost to 127.0.0.1 on Windows. Perhaps doing the same would make sense in Netty too.\n> \n> —\n> You are receiving this because you commented.\n> Reply to this email directly, view it on GitHub https://github.com/netty/netty/issues/5386#issuecomment-225536427, or mute the thread https://github.com/notifications/unsubscribe/AAa0QrENFN-8OEAFmlqfsVSJxiXLgnmVks5qLSevgaJpZM4I0JNJ.\n", "a workaround is to add this to the list of parsed hosts when it does not exists using the blocking API (which is not a problem usually as this is done in the resolver itself)\n", "i.e you can still get the localhost value once and cache it in a static.\n", "```\n try {\n if (!entries.containsKey(\"localhost\")) {\n entries.put(\"localhost\", InetAddress.getLocalHost());\n }\n } catch (UnknownHostException ignore) {\n }\n builder.hostsFileEntriesResolver(entries::get);\n```\n", "Fixed by https://github.com/netty/netty/pull/5410\n" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=11.0, similarity=0.9507042869265819)" ]

[ "will check...\n\n@vietj thanks for reporting!\n", "it made my day epic :-)\n", "DNS epic day\n", "lol\n", "Fixed by https://github.com/netty/netty/pull/5413\n" ]

[ "Why need to create this `decodeName0` method? Is it for providing a hook method to allow people to change the behavior?\n", "Yes\n\n> Am 16.06.2016 um 22:38 schrieb Xiaoyan Lin [email protected]:\n> \n> In codec-dns/src/main/java/io/netty/handler/codec/dns/DefaultDnsRecordDecoder.java:\n> \n> > @@ -106,7 +106,19 @@ protected DnsRecord decodeRecord(\n> > \\* @param in the byte buffer containing the DNS packet\n> > \\* @return the domain name for an entry\n> > */\n> > - protected String decodeName(ByteBuf in) {\n> > - protected String decodeName0(ByteBuf in) {\n> > Why need to create this decodeName0 method? Is it for providing a hook method to allow people to change the behavior?\n> \n> —\n> You are receiving this because you were assigned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n" ]

[ "defect" ]

[ "codec-dns/src/main/java/io/netty/handler/codec/dns/DefaultDnsRecordDecoder.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolverContext.java" ]

[ "codec-dns/src/main/java/io/netty/handler/codec/dns/DefaultDnsRecordDecoder.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolverContext.java" ]

[]

[ "timestamp(timedelta=3.0, similarity=0.8565223779274694)" ]

[ "Thanks for reporting @npordash! Are you in a situation where `listener.onDataRead` is called multiple times in `DelegatingDecompressorFrameListener`, and you are trying to call `consumeBytes(..)` for data which was not returned by previous calls to `Http2FrameListener.onDataRead`?\n", "> I tested this with both 4.1.0.Final and 4.1.1.Final.\n\nDo you have a unit test which you would be willing to contribute? I have a PR but it would be nice to verify against a unit test if you already have one.\n", "Thanks for the quick response!\n\nI'm not in a situation where `onDataRead` is being called multiple times, at least I have yet to observe that happening. The exception gets thrown on the first data frame received.\n\nMy use-case is that I'm exposing inbound data frames as an `rx.Observable<ByteBuf>` and I only want to report the processing of the bytes when a `ByteBuf` has been sent to a subscriber of the `Observable`. It's possible that the `ByteBuf` is sent to a subscriber inside the call to `onDataRead` if the processing is synchronous which means that the call to `consumeBytes` might happen before `onDataRead` returns as well.\n\nUnfortunately, I do not have an existing unit test for this yet. I stumbled across this while I was prototyping which involved manual testing with `nghttp` to send data to a netty http2 server 😦 \n", "> which means that the call to consumeBytes might happen before onDataRead returns as well.\n\nAs long as you are not double releasing the bytes you should be OK. I think I found the issue and I'll ask you to verify with the PR when its ready.\n", "Awesome, thanks again!\n" ]

[ "nit: spelling on the method name ... should be `...Bytes`\n", "Is this condition ever possible? I think it only would be if padding were `< 0`, which will never happen. Suggest either getting rid of this condition or moving it somewhere more clear.\n", "Same comment as below.\n", "Don't we also have to remove the property from the stream?\n", "Don't we have to mark these bytes as consumed?\n", "+1\n", "I guess this is no longer needed now that it is being used strictly to increment ... overflow should be guarded by flow control.\n", "removed\n", "There was an ordering issue related to the stream being removed if an error occurred and attempting to return bytes to flow control. So if we removed the property we would not be able to return bytes to flow control in this condition. Ideally we would remove it, but I don't think it hurts anything to leave it in the property map.\n", "Good point ... we should consume immediately instead of returning the bytes. I'll add a comment on the return statement too.\n", "Shouldn't this be a call to `decompressor.consumeBytes()`?\n", "The `decompressor` just keeps state to be able to convert bytes the users sees (decompressed) to bytes the flow controller knows about (compressed). `ConsumedBytesConverter` wraps the flow controller so that when ever bytes are returned it uses the `decompressor` to translate from what the user knows (decompressed) to what the flow controller knows (compressed). So I think we actually want to return bytes to the flow controller here ... this is effectively what is done when bytes are returned by this method (done by `DefaultHttp2ConnectionDecoder`) however bcz there may be multiple frames decompressed it makes sense to return the bytes immediately instead of waiting for the method to return.\n", "@nmittler - WDYT?\n", "@Scottmitch Ah I had forgotten that the flow controller is actually the `ConsumedBytesConverter`. Maybe add a comment?\n", "done\n" ]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=22.0, similarity=0.9334949369019993)" ]

[ "@caillette yet that sounds not right... let me come up with a fix quickly.\n", "Fixed https://github.com/netty/netty/pull/5419\n" ]

[ "`HttpScheme.HTTPS.port()`\n", "consider using `HttpScheme` here\n", "nit: you could just do `return originValue + ':' + wsPort;`\n", "nit: HttpScheme also provides the `https` portion with `HttpScheme.name()`.\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/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" ]

[ "timestamp(timedelta=14.0, similarity=0.908510300263825)" ]

[ "@trustin PTAL\n\n> Am 20.06.2016 um 16:09 schrieb Franck Chevassu [email protected]:\n> \n> [ netty-4.1.1.Final ]\n> \n> There seem to be a couple of things in HttpProxyHandler that do not comply with RFC 2817 (Upgrading to TLS Within HTTP/1.1)\n> \n> The HTTP version required to support the CONNECT method is 1.1,\n> whereas Netty's HttpProxyHandler class specifies a 1.0 version (line 123)\n> \n> The \"Host\" header should apparently hold the name of the target host,\n> whereas Netty specifies the proxy address instead, (lines 127-131).\n> \n> Changing the original code in lines 122-131 :\n> \n> FullHttpRequest req = new DefaultFullHttpRequest(\n> HttpVersion.HTTP_1_0, HttpMethod.CONNECT,\n> rhost + ':' + raddr.getPort(),\n> Unpooled.EMPTY_BUFFER, false);\n> \n> SocketAddress proxyAddress = proxyAddress();\n> if (proxyAddress instanceof InetSocketAddress) {\n> InetSocketAddress hostAddr = (InetSocketAddress) proxyAddress;\n> req.headers().set(HttpHeaderNames.HOST, hostAddr.getHostString() + ':' + hostAddr.getPort());\n> }\n> To something like this would solve both issues :\n> \n> final String host = rhost + ':' + raddr.getPort();\n> \n> FullHttpRequest req = new DefaultFullHttpRequest(\n> HttpVersion.HTTP_1_1, HttpMethod.CONNECT,\n> host,\n> Unpooled.EMPTY_BUFFER, false);\n> \n> req.headers().set(HttpHeaderNames.HOST, host);\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "@fchevassu Care to send a PR? I don't have enough bandwidth these days.\n", "Fixed by https://github.com/netty/netty/pull/5437\n" ]

[]

[ "defect" ]

[ "handler-proxy/src/main/java/io/netty/handler/proxy/HttpProxyHandler.java" ]

[ "handler-proxy/src/main/java/io/netty/handler/proxy/HttpProxyHandler.java" ]

[]

[ "timestamp(timedelta=41.0, similarity=1.0000000000000002)" ]

[ "`HpackUtil.equals` currently implements an equality comparison which \"doesn't leak timing information\" [1]. @jpinner - can you provide some motivation for why this type of comparison is necessary in this specific case?\n\n[1] https://github.com/twitter/hpack/commit/84e3438ae60efad587c7058c87b04064d7d6a536#diff-d23f44552b6381004428efb52e1667baR26\n", "@Scottmitch I think it has something to do with security http://httpwg.org/specs/rfc7541.html#Security\n", "@buchgr that's correct\n@Scottmitch it's to help prevent timing-based side-channel attacks on header field values\n", "Understood it is to prevent timing side-channel attacks. I guess the compromise was made to leak when the length is correct, but not how many characters are matching.\n", "Thanks for looking into this Scott! I believe this is quite an urgent issue for gRPC that would be great to have in the next Netty 4.1.2. So please do let us know if you don't have enough cycles, we are happy to help out! cc: @carl-mastrangelo \n", "@buchgr - Np. PR is pending.\n", "This might be a dumb question, but does the constant time compare really need to be used for non-security sensitive headers? HPACK provides a means to not use indexing and never indexing, which seems like it implies they sensitive. Most of the time we do want the fast, time-leaking compare.\n", "@carl-mastrangelo - +1\n\n@jpinner - WDYT?\n", "I'd have to take a pass through the code and see exactly how it's used. The compare does predate the inclusion of the \"never index\" flag. That being said I can certainly think of \"sensitive\" headers that you might want to index for performance in the table (e.g. domain cookies sent on every request).\n\nI personally would be cautious here.\n\nSent from my iPhone\n\n> On Jun 22, 2016, at 2:21 PM, Scott Mitchell [email protected] wrote:\n> \n> @carl-mastrangelo - +1\n> \n> @jpinner - WDYT?\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n" ]

[ "@Scottmitch could you not just use the methods in PlatformDependent0 ?\n", "can you clarify which method you are suggesting I use? `PlatformDependent0` doesn't have a equals which accepts a `CharSequence`.\n", "I ran benchmarks comparing some equals functions, and found that it is significantly faster to use an int rather than a byte:\n\n```\nHeadersBenchmark.arrays 4096 sample 129777 1227.096 ± 4.670 ns/op\nHeadersBenchmark.digest 4096 sample 127528 2488.555 ± 6.749 ns/op\nHeadersBenchmark.theBytes 4096 sample 99710 3170.983 ± 9.587 ns/op\nHeadersBenchmark.theChars 4096 sample 105714 2992.226 ± 8.468 ns/op\nHeadersBenchmark.theInts 4096 sample 126811 2497.360 ± 7.106 ns/op\nHeadersBenchmark.theLongs 4096 sample 99672 3181.179 ± 8.380 ns/op\nHeadersBenchmark.theShorts 4096 sample 100008 3158.683 ± 8.164 ns/op\n```\n\n`arrays` is Arrays.equals (for comparison)\n`digest` is MessageDigest.isEqual (broken upto in partly java6, fixed in 2009)\nAll the others are copy paste versions of MessageDigest.isEqual using different accumulators.\n", "The [Go](https://golang.org/src/crypto/subtle/constant_time.go?s=490:531#L2) implementation uses a constant time comparison for the final check here. You might consider using it too.\n", "thanks @carl-mastrangelo !\n", "never mind ... I thought this two methods are a copy of what you have implemented...\n", "@carl-mastrangelo - I repeated your benchmark when charsequence is used and not surprisingly I see the same results ... i'll make the same change here too\n\n```\nBenchmark Mode Cnt Score Error Units\nTestBenchmark.theChars sample 152044 2105.563 ± 7.328 ns/op\nTestBenchmark.theInts sample 141490 1155.404 ± 6.013 ns/op\nTestBenchmark.theLongs sample 108386 1498.885 ± 8.700 ns/op\nTestBenchmark.theShorts sample 151317 2119.505 ± 8.116 ns/op\n```\n", "+1. Done\n", "@carl-mastrangelo - note this change ... before it was skipping out early if the name didn't match ... which may yield some timing info.\n\nI propose we do a follow up investigation if you think we can reduce the amount of \"constant time comparisons\" that we must do ... or just generally if we are using the \"constant time comparisons\" where/how we should be.\n", "I asked internally how \"constant\" a constant function needs to be, and so far the answer has been \"nothing is constant and will basically be impossible with an optimizing compiler / processor\".\n\nThis is best effort, and copying behavior of other security reviewed code feels like said effort.\n", "Are you using & to avoid short circuit behavior? If so, that's subtle and deserves a comment. If not, then && seems appropriate.\n", "Yes. I can add a comment (this preserves existing behavior).\n", "That should be placed into `equalsConstantTime` in `ConstantTimeUtils` and `PlatformDependent` as that's generally useful for have in an equals method. Shorter strings are smaller or so.\n", "I think the name is misleading, as it's not constant time equals?\n", "We should probably come up with a better name for this class ... `EqualsUtils`?\n", "Can you clarify your suggestion? `ConstantTimeUtils` for `CharSequence` does have this check. However the variant which operates on `byte[]` accepts a length parameter ... so it is implied the length which will be iterated is the same (and bounds checks are omitted are assumed to be outside the scope of the method per the docs).\n", "Previously I was using `Consistent` instead of `Constant`. However [Go](https://golang.org/src/crypto/subtle/constant_time.go?s=490:531#L2) and various internet searches seem to associate `constant time` with this concept so I figured I would use the prevailing nomenclature ... do you have another suggestion?\n", "This comment seems to be related to https://github.com/netty/netty/pull/5446#discussion_r68375984. lets discuss their.\n", "The name constant implies that the contents of the array does not affect the time it takes to process. There are always going to be compiler tricks or processor tricks or OS tricks that will mess with the timing. The important thing is to not have different behavior based on the contents. That's where the \"constant\" comes from.\n\nThat said, to @buchgr's comment, I think you may still need to do a bounds check. Something like bytes1.length - startPos1 == bytes2.length - start2Pos.\n", "> I think you may still need to do a bounds check. Something like bytes1.length - startPos1 == bytes2.length - start2Pos.\n\nThe bounds check is done through the length check. The \"no bounds checking behavior\" is also specified in the [javadocs](https://github.com/netty/netty/pull/5446/files#diff-ea8860513cb8af4c4392d1e295326fc6R639), PlatformDependent/ConstantTimeUtils are part of the internal API, and this behavior is consistent with [PlatformDependent.equals](https://github.com/netty/netty/blame/4.1/common/src/main/java/io/netty/util/internal/PlatformDependent.java#L625).\n", "Sorry to be nitpicky, but my intuition says that this will be slower than the forward path just because it would be easier for the cache to take advantage of this. \n\nAlso, how are you sure that byte arrays are word aligned? Doing an unaligned access is pretty slow. Arrays that are stack allocated may not have the same alignment guarantees as heap allocated. \n\nIs there benchmarking for this method? \n\nAnd fyi, I am proposing a similar change for Go: https://go-review.googlesource.com/#/c/24422/1/src/crypto/subtle/constant_time.go\n", "> Sorry to be nitpicky, but my intuition says that this will be slower than the forward path just because it would be easier for the cache to take advantage of this.\n> \n> Is there benchmarking for this method?\n\n`PlatformDependentBenchmark` exists which I will run for a before/after. This was shown to provide benefits on smaller inputs when running benchmarks on `equalsConstantTime`. The overhead of allocating `i`, `j`, and `end` can be avoided for smaller inputs that don't need these variables.\n\n> Also, how are you sure that byte arrays are word aligned?\n\n`PlatformDependent.equalsConstantTime` checks if aligned access is allowed in general, but no check is done on the individual arrays. Do you know how I would check, and an example allocation which I can use to verify this method?\n", "@carl-mastrangelo - Here are the results of `PlatformDependentBenchmark`. I switched the measurement to average time and added `1` for a size.\n\nWe gain on the smaller inputs, but start to perform worse around input size of `100`. However the gain may be negligible (or likely a loss) if there are larger headers involved ... let me revert this.\n\n```\nWith this PR\nBenchmark (size) Mode Cnt Score Error Units\nPlatformDependentBenchmark.unsafeBytesEqual 1 avgt 5 4.453 ± 0.081 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 7 avgt 5 4.756 ± 0.139 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 10 avgt 5 5.493 ± 0.223 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 50 avgt 5 9.630 ± 0.262 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 100 avgt 5 13.559 ± 0.306 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 1000 avgt 5 107.538 ± 1.410 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 10000 avgt 5 995.796 ± 16.488 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 100000 avgt 5 9942.613 ± 350.996 ns/op\n\nBefore This PR\nBenchmark (size) Mode Cnt Score Error Units\nPlatformDependentBenchmark.unsafeBytesEqual 1 avgt 5 4.906 ± 0.404 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 7 avgt 5 5.469 ± 0.099 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 10 avgt 5 6.051 ± 0.098 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 50 avgt 5 9.634 ± 0.235 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 100 avgt 5 12.056 ± 0.455 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 1000 avgt 5 73.085 ± 1.784 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 10000 avgt 5 649.883 ± 23.863 ns/op\nPlatformDependentBenchmark.unsafeBytesEqual 100000 avgt 5 7254.138 ± 475.926 ns/op\n```\n", "@carl-mastrangelo - The following was collapsed but still open ...\n\n> Also, how are you sure that byte arrays are word aligned?\n\n`PlatformDependent.equalsConstantTime` checks if unaligned access is allowed in general, but no check is done on the individual arrays. Do you know how I would check, and an example allocation which I can use to verify this method?\n\n> And fyi, I am proposing a similar change for Go: https://go-review.googlesource.com/#/c/24422/1/src/crypto/subtle/constant_time.go\n\nI guess this would depend upon the answer to the first question ... maybe investigate this as a followup?\n", "Not sure how you would check, but being able to do unaligned access has a noticeble performance degradation. Intel lets you do it, but just does it slower. (i.e. has to load two cache lines instead of one, and takes longer to go through the pipeline). \n\nThe benchmark numbers on the Go version saw a boost (~460ns -> 320ns comparing two 4k buffers) by having correct alignment. I have no idea how Java tells you if something is aligned, since I don't know how the memory allocator works. Since java has a compacting GC, any pointers to data are an extra level of indirection away right?\n", "I think array contents on hotspot 64-bit intel should by default be 8 byte aligned. Due to compressed oops the array object is 8 byte aligned, followed by a 12 byte object header and the 4 byte array length, followed by array contents 😄 \n\n> Since java has a compacting GC, any pointers to data are an extra level of indirection away right?\n\nafaik the references to an object are updated when moved. so an array variable holds a pointer to the array contents (minus some offset for the object header + size).\n\nThat being said, apparently you can get the address of an object/array [somehow](http://hg.openjdk.java.net/code-tools/jol/file/30c3fbadf049/jol-samples/src/main/java/org/openjdk/jol/samples/JOLSample_17_Allocation.java#l64) (best guess via Unsafe)\n\nHowever, since java GC can relocate an object between getting the address and doing the comparison, you might never know whether it's the right address at the time of comparison :-).\n\nI think the whole complexity might not be worth it, especially since for headers we are likely not comparing 4K buffers but a few dozen bytes at a time.\n", "@buchgr - Thanks for the details. I agree it doesn't seem practical to do this in java and may be more trouble than it is worth for this use case anyways.\n", "@buchgr are you sure it has the same alignment, even when the array is stack allocated due to escape analysis? Also, there were still about ~12% perf gain by using aligned access. Whether or not this matters is debatable.\n" ]

[ "improvement" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HeaderField.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HpackUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/StaticTable.java", "common/src/main/java/io/netty/util/internal/PlatformDependent.java", "common/src/main/java/io/netty/util/internal/PlatformDependent0.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HeaderField.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HpackUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/StaticTable.java", "common/src/main/java/io/netty/util/internal/ConstantTimeUtils.java", "common/src/main/java/io/netty/util/internal/PlatformDependent.java", "common/src/main/java/io/netty/util/internal/PlatformDependent0.java", "microbench/src/main/java/io/netty/microbench/http2/internal/hpack/HpackUtilBenchmark.java" ]

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

[ "timestamp(timedelta=5.0, similarity=0.8870567172644754)" ]

[ "We also need to audit outbound path.\n", "cc: @Scottmitch, @normanmaurer, @nmittler \n", "@ejona86 should I take this?\n", "@buchgr, please do.\n\nThis will also need to be updated:\nhttps://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2DataFrame.java#L72\n", "If we add 1 to readPadding() we will have be sure to account that the byte is already read and the [int dataLength = payload.readableBytes() - padding;](https://github.com/netty/netty/blob/netty-4.1.1.Final/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java#L403) must account for this (and also for the other usages of `readPadding()`). An alternative would be to just adding `1` when we call the FrameListener [1] to ensure it is accounted for in flow control ... then we don't have to update usages of `readPadding()`. We should also update the comments on the `FrameListener` if we are going to include the \"padding length byte\" in the `padding` value.\n\n[1] https://github.com/netty/netty/blob/netty-4.1.1.Final/codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2FrameReader.java#L410\n", "ok i ll take this. will see what turns out best in code. can discuss on the pr.\n", "thanks @buchgr !\n", "A quick look at the outbound path seems to indicate, that the pad length field is also not counted towards flow control there.\n\nInterestingly, the spec seems to allow zero length padding, that is the pad field is present with a value of zero.\n\n> Note: A frame can be increased in size by one octet by including a Pad Length field with a value of zero.\n\nI have to take a more detailed look tomorrow morning (11pm here), but this might indicate that we are doing padding wrong as when padding is 1, we really add two bytes (pad length and 1 byte padding). So probably we should just have padding to be a value in [0, 256], with a padding of 1 being special cased on the frame writer as pad length field present with value zero.\n" ]

[ "Either keep the return value int or don't cast to short?\n", "You might comment that the +1 is due to the pad length, which comes before most of the payload whereas the padding itself comes after.\n", "Probably should add a bit of javadoc to make it clear how it is different from `padding`. The implementation is \"obvious\" in what it does, but not quite _why_ it does.\n", "can we share this method? It seems to be used in a few spots. Maybe put it in `Http2CodecUtil`?\n", "explain how this padding value will be translated to bytes on the wire. it may be unexpected that 256 is allowed when the frame definition only allows for 8 byte length.\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2DataFrame.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/DefaultHttp2HeadersFrame.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2DataWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameListener.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2LocalFlowController.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2DataFrame.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/DefaultHttp2HeadersFrame.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2DataWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameListener.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameWriter.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2LocalFlowController.java" ]

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

[ "timestamp(timedelta=23.0, similarity=0.9354724155765581)" ]

[ "checking...\n", "Verified that the issue not exists in 4.x. Looking into 3.10.x now.\n", "Hi @normanmaurer I would like to know if there are any plans to fix this soon as I would like to see 3.10.6.Final released as soon as it is humanly possible.\n", "I had no time yet... If you want help submit a PR 👍\n\n> Am 11.04.2016 um 15:30 schrieb Guido Medina [email protected]:\n> \n> Hi @normanmaurer I would like to know if there are any plans to fix this soon as I would like to see 3.10.6.Final released as soon as it is humanly possible.\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly or view it on GitHub\n", "Fixed by https://github.com/netty/netty/pull/5451\n" ]

[ "This is a copy & paste from `4.0` branch, I noticed it preferred direct access vs `for-each byte` which creates an iterator.\n", "This is the actual bug fix, to not return true when `mask == -1` which is also a copy & paste from `4.0` branch.\n" ]

[ "defect" ]

[ "src/main/java/org/jboss/netty/handler/ipfilter/IpV4Subnet.java" ]

[ "src/main/java/org/jboss/netty/handler/ipfilter/IpV4Subnet.java" ]

[]

[ "timestamp(timedelta=13.0, similarity=0.948771872386056)" ]

[ "@vietj +1 ... Want to provide a pr?\n", "will do soon @normanmaurer \n", "@vietj thanks a lot!\n", "I'm seing an issue in tests because the channel is not yet initialized on registration, i.e the code:\n\n```\n b.handler(new ChannelInitializer<DatagramChannel>() {\n @Override\n protected void initChannel(DatagramChannel ch) throws Exception {\n ch.pipeline().addLast(DECODER, ENCODER, responseHandler);\n }\n });\nb.option(ChannelOption.DATAGRAM_CHANNEL_ACTIVE_ON_REGISTRATION, true);\n\nChannelFuture channelFuture = b.register();\n```\n\ndoes not guarantee that the `initChannel` is invoked when the `channelFuture` is succeeded. As consequence sometimes during tests the `DatagramDnsQuery` are not encoded as the `ENCODER` is not yet added to the pipeline.\n\nOne solution is to use a promise that is resolved after the pipeline has been modified instead of using the channelFuture.\n\nThoughts ?\n", "@vietj sounds good.\n", "@vietj I think even better would be to have the promise be notified once channelActive is called.\n", "Fixed by https://github.com/netty/netty/pull/5461\n" ]

[ "woups\n", "@vietj make the constructor package private\n" ]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=19.0, similarity=0.9095846852576694)" ]

[ "For `hash(CharSequence)`, it's a bit dangerous to use `hashCode()`. In `CharSequence`'s Javadoc:\n\n> This interface does not refine the general contracts of the equals and hashCode methods. The result of comparing two objects that implement CharSequence is therefore, in general, undefined.\n", "@ejona86 - Great point. `name.hashCode()` would not work. I've updated the issue description to reflect this.\n" ]

[]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec/src/main/java/io/netty/handler/codec/DefaultHeaders.java", "common/src/main/java/io/netty/util/AsciiString.java", "microbench/src/main/java/io/netty/microbench/http2/internal/hpack/EncoderBenchmark.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec/src/main/java/io/netty/handler/codec/DefaultHeaders.java", "common/src/main/java/io/netty/util/AsciiString.java", "microbench/src/main/java/io/netty/microbench/http2/internal/hpack/EncoderBenchmark.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/internal/hpack/TestCase.java" ]

[ "timestamp(timedelta=38.0, similarity=0.8489490830934392)" ]

[ "yes the equals implementation does seem suspect ... @veebs wdyt?\n", "Fixed by https://github.com/netty/netty/pull/5527\n" ]

[ "nit: kill line\n", "do we have any context for how this was intended to be used? It seems strange to compare the length, but not compare the actual content. If we don't do this consider just making this method `return upload1.getName().compareToIgnoreCase(upload2.getName());`\n", "@Scottmitch no context here.. Let me just remove the TODO and change to what you suggested.\n" ]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/multipart/DiskFileUpload.java", "codec-http/src/main/java/io/netty/handler/codec/http/multipart/MemoryFileUpload.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/multipart/DiskFileUpload.java", "codec-http/src/main/java/io/netty/handler/codec/http/multipart/FileUploadUtil.java", "codec-http/src/main/java/io/netty/handler/codec/http/multipart/MemoryFileUpload.java" ]

[ "codec-http/src/test/java/io/netty/handler/codec/http/multipart/DiskFileUploadTest.java", "codec-http/src/test/java/io/netty/handler/codec/http/multipart/MemoryFileUploadTest.java" ]

[ "timestamp(timedelta=158290.0, similarity=0.8964804195974887)" ]

[ "@olix0r nope... A PR would be awesome :)\n", "@olix0r actually @Scottmitch made a good point. Can you wrap the code into a try / catch block and fail the promise on error ?\n" ]

[ "@buchgr should we remove the `exceptionCaught(...)` here ? I think if we propagate to the promise we should not also call `exceptionCaught(...)`. Also @nmittler @Scottmitch WDYT ?\n" ]

[]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameCodec.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameCodec.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2FrameCodecTest.java" ]

[ "timestamp(timedelta=19.0, similarity=0.8874743946475105)" ]

[ "BTW, I have tried the 4.1.3.Final-SNAPSHOT version, it came out the same error too.\n", "@xsir let us know... I have not much knowledge in terms of osgi.\n", "I have figured it out finally. The root cause is: the `netty-tcnative` library is loaded by a ClassLoader that is fully independent of its bundle ClassLoader.\n\nSpecifically, the `netty-tcnative` is a bundle, which is loaded by OSGi framework with a separated ClassLoader. And the `netty-handler` is a bundle too, which contains the `OpenSsl` utility class. When the `OpenSsl` class is accessed, the static initialization will try to load the `netty-tcnative` library, which is embedded in `netty-tcnative` bundle. The OpenSsl loads the `tcnative` library through the `NativeLibraryLoader.loadFirstAvailable()` helper method, which will try to find the library in **PATH**, then in the JAR if failed. As we know, the library file is embedded in the bundle JAR, the `NativeLibraryLoader` will extract it to a tmp file, then call `System.load()` with the tmp file absolute path. The point is: in the SUN implementation, both `System.load()` and `System.loadLibrary()` are **CallerSensitive**, that means the native library is loaded by the caller class's ClassLoader! That's it, the `netty-tcnative` library is loaded by `NativeLibraryLoader.class`'s ClassLoader -- the netty-comon bundle's ClassLoader. But when the native method is called, it using the method definition class's ClassLoader, it can NOT find the native library in itself, then the `java.lang.UnsatisfiedLinkError: org.apache.tomcat.jni.Library.version(I)I` will be threw.\n\nYes, the root cause is the abused ClassLoader, it is discussed by @jjpp in [netty-tcnative issue 136](https://github.com/netty/netty-tcnative/issues/136). \n\nThen, the solution is clear: try to use the same ClassLoader to load the `netty-tcnative` library as the ClassLoader to initialize the native method. \n\nI hope I have made myself clear:) I'll submit a PR soon, thank you.\n", "Fixed by https://github.com/netty/netty/pull/5545\n" ]

[ "does this need to be public? can this just be `private static` and live in `NativeLibraryLoader`?\n", "can we do the reflection initialization parts bits statically when initializing `NativeLibraryLoader`?\n", "Hi, @Scottmitch ! That was a good idea. But, the helper class's duty is to delegate the calling to System.load() and it will be injected into the native library's ClassLoader, then when its method is invoked, the native library would be loaded into the native library's ClassLoader, NOT the netty-common's ClassLoader.\nSo, **the helper class should be as simple as possible, and should not be a inner class, otherwise we have to load the outer class and its dependencies**.\nBut, the helper class can be package.\n", "The key point of this method is to load the helper class into the specified ClassLoader **if it not found in the ClassLoader**. Yes, the define class process is invoked only once if every thing is fine.\n", "@xsir might be worth adding a java doc explaining this?\n", "remove `/*package*/`\n", "`return absolute ? System.load(libName) : System.loadLibrary(libname);`\n", "`{@link System#load(String)}`, `{@link System#System.loadLibrary(String)}`\n", "2016\n", "put the `close()` calls into a `try { } catch (IOException ignore)` block so we not swallow other exceptions.\n", "`{@link Class}` ... `{@link ClassLoader}` \n", "`IllegalStateException(...)` ?\n", "@jasontedor can you comment about the impact of `SecurityManager` here ? Seems like you know this stuff in and out :)\n", "is NOT\n", "Use `{@link }`\n", "We usually use `//` for comments.\n", "IllegalStateException ?\n", "Throwable ?\n", "The System.load() will throw a UnsatisfiedLinkError if the library is not found. If we catch Throwable here, the System.load() will be called twice. That is why I catch Exception here. WDYT?\n", "The System.load() is void too.\n", "@xsir - Is it not possible / not worth it to move `Method defineClass` initialization here into a static context, and if so can you comment? It doesn't seem to depend upon any local state.\n", "env -> environment?\n", "To my view, the answer is not possible. Because the `NativeLibraryLoader` is a utility. Its method such as `load(String, ClassLoader)` is invoked with a **target** `ClassLoader` and we need to define the helper class into it, that is a dynamic process. For example, the `netty-tcnative` library is loaded by `OpenSsl` in `netty-handler` with the `SSL`'s `ClassLoader`, we have to inject the helper class into `SSL`'s `ClassLoader`. But for `netty-transport-native-epoll` library is loaded by `Native` with its `ClassLoader`, then we have to inject the helper class into `Native`'s `ClassLoader`.\nThe helper class injection process is nothing to do with `NativeLibraryLoader` itself . I have no idea how to make it static.\n\nBTW, there are two native library involved in netty, both have the same OSGi loading issue. And there has another solution to this issue:\n- `netty-tcnative` case, which provided a loading helper class `org.apache.tomcat.jni.Library`. Just like 9f2a2135, we should modify `OpenSsl`, instead of calling the `NativeLibraryLoader`, trying the helper class first (Depends on a properly `Bundle-NativeCode` manifest entry). _The helper class `Library` in `netty-tcnative` does NOT support loading library by absolute path._\n- `netty-transport-native-epoll` case, we should modify `Native`, instead of calling the `NativeLibraryLoader`, trying the `System.load()` first (Depends on a properly `Bundle-NativeCode` manifest entry).\n\nThis solution is more simple, but there has a drawback: the library loading process is diff between OSGi and non-OSGi. In the OSGi, the library in `$PATH` has a higher priority than the one in the bundle. The non-OSGi is just the reverse. That's why I submitted a more complicated solution:)\nThanks for all of the comments!\n", "thanks for the explanation ... fine as is\n", "nit: consider restructuring to reduce the conditionals\n\n``` java\nint r;\nwhile ((r = in.read(buf)) != -1) {\n out.write(buf, 0, r);\n}\n```\n", "should we use a larger default size?\n", "what I meant was using a bigger size for `ByteArrayOutputStream` (e.g `ByteArrayOutputStream(4096)`).\n", "The security manager implications here are fine. \n", "@jasontedor so I understand you right that you think its fine to pull in in terms of `SecurityManager` stuff ? \n" ]

[ "defect" ]

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

[ "common/src/main/java/io/netty/util/internal/NativeLibraryLoader.java", "common/src/main/java/io/netty/util/internal/NativeLibraryUtil.java" ]

[]

[ "timestamp(timedelta=11.0, similarity=0.968899664563478)" ]

[ "@rkapsi good catch! Let me come up with a fix\n", "@rkapsi PTAL https://github.com/netty/netty/pull/5557\n", "Fixed by https://github.com/netty/netty/pull/5557\n" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=20.0, similarity=0.8538622961285909)" ]

[ "I added a PR for this: #5567 \n", "Fixed by https://github.com/netty/netty/pull/5567\n" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=17.0, similarity=0.9533093199440464)" ]

[ "Fixed by https://github.com/netty/netty/pull/5571\n" ]

[ "Shouldnt this stay outside of the rose block?\n", "I don't understand what you mean.\n", "Outside of the else block.... Thanks to autocorrect\n", "no, it's the same behavior than in the foreach loop that returns immediately after calling internalResolve.\n", "got it.\n", "call resolver.close() at the end of the test ?\n" ]

[ "defect" ]

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

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

[ "resolver-dns/src/test/java/io/netty/resolver/dns/SearchDomainTest.java" ]

[ "timestamp(timedelta=63.0, similarity=0.8996140167226422)" ]

[]

[]

[]

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

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

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

[ "timestamp(timedelta=385.0, similarity=0.8761801680318148)" ]

[]

[]

[ "defect" ]

[ "buffer/src/main/java/io/netty/buffer/PoolThreadCache.java", "common/src/main/java/io/netty/util/Recycler.java", "common/src/main/java/io/netty/util/ResourceLeakDetector.java", "common/src/main/java/io/netty/util/internal/MathUtil.java", "common/src/main/templates/io/netty/util/collection/KObjectHashMap.template" ]

[ "buffer/src/main/java/io/netty/buffer/PoolThreadCache.java", "common/src/main/java/io/netty/util/Recycler.java", "common/src/main/java/io/netty/util/ResourceLeakDetector.java", "common/src/main/java/io/netty/util/internal/MathUtil.java", "common/src/main/templates/io/netty/util/collection/KObjectHashMap.template" ]

[]

[ "timestamp(timedelta=50.0, similarity=0.8688967097123617)" ]

[ "Thanks will check\n", "thanks for reporting @tkaitchuck !\n", "Fixed by https://github.com/netty/netty/pull/5605\n" ]

[]

[ "defect" ]

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

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

[ "timestamp(timedelta=134149.0, similarity=1.0000000000000002)" ]

[ "To show the issue in vert.x the following unit test should show the error message when running it inside the current master of eclipse/vert.x\n\nhttps://gist.github.com/alexlehm/d4d4ffdf95c97c78dff7ba91de1b0d09\n\nStarting test: UnknownHostExceptionTest#testResolveExceptionHostname \njava.net.UnknownHostException: failed to resolve 'sdfsdfsdfsdsdffd.internal.example.com'. Exceeded max queries per resolve 3\n", "@alexlehm makes sense... interested in proving a PR ?\n", "https://github.com/netty/netty/pull/5608\n" ]

[]

[ "defect" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolverContext.java" ]

[ "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolverContext.java" ]

[ "resolver-dns/src/test/java/io/netty/resolver/dns/SearchDomainTest.java" ]

[ "timestamp(timedelta=75.0, similarity=0.8588913995068724)" ]

[ "I agree, I prefer `SLF4J-API` and `Logback`\n", "it is causing any harm? maybe we can just mark it as deprecated, comment it will be removed in future releases, and recommend using an alternative (e.g. slf4j).\n", "> it is causing any harm?\n\nNo harm. It is very outdated and not used within netty. Deprecation also fine.\n", "+1 for deprecated.\n", "Fixed\n" ]

[]

[ "cleanup" ]

[ "common/src/main/java/io/netty/util/internal/logging/CommonsLogger.java", "common/src/main/java/io/netty/util/internal/logging/CommonsLoggerFactory.java" ]

[ "common/src/main/java/io/netty/util/internal/logging/CommonsLogger.java", "common/src/main/java/io/netty/util/internal/logging/CommonsLoggerFactory.java" ]

[]

[ "timestamp(timedelta=20.0, similarity=0.9641005524722586)" ]

[ "Do you want to provide a pr to fix it?\n", "I'm on company time - I'll ask (+day of turnaround).\n\nSupposing that I get an OK to fix/re test.\nDo you want to:\n- have the DefaultSctpServerChannelConfig brought to match DefaultSctpChannelConfig (where the static import is `import static io.netty.channel.sctp.SctpChannelOption.*;`)\n- just add ChannelOption to SCTP_INIT_MAXSTREAMS\n\n?\n", "Yes sounds good\n", "I've never got an answer back about it (working on this issue on company time). \nIf I where to guess that means NO.\nSorry.\n", "God damn, so that's what \"close and comment\" does.\nReopening.\n", "@normanmaurer I will do the fix. In which branch I have to open a PR?\n", "4.1 would be best\n\n> Am 22.01.2016 um 18:37 schrieb Jestan Nirojan [email protected]:\n> \n> @normanmaurer I will do the fix. In which branch I have to open a PR?\n> \n> —\n> Reply to this email directly or view it on GitHub.\n", "@jestan did you fix this ?\n", "Fixed by https://github.com/netty/netty/pull/5638\n" ]

[ "nit: newline\n", "fixed\n" ]

[ "defect" ]

[ "transport-sctp/src/main/java/io/netty/channel/sctp/DefaultSctpChannelConfig.java", "transport-sctp/src/main/java/io/netty/channel/sctp/DefaultSctpServerChannelConfig.java" ]

[ "transport-sctp/src/main/java/io/netty/channel/sctp/DefaultSctpChannelConfig.java", "transport-sctp/src/main/java/io/netty/channel/sctp/DefaultSctpServerChannelConfig.java" ]

[ "transport-sctp/src/main/test/io/netty/channel/sctp/SctpLimitStreamsTest.java", "transport-sctp/src/main/test/io/netty/channel/sctp/nio/NioSctpLimitStreamsTest.java", "transport-sctp/src/main/test/io/netty/channel/sctp/oio/OioSctpLimitStreamsTest.java" ]

[ "timestamp(timedelta=2.0, similarity=0.9480561467821279)" ]

[ "@trustin so you will look into it ?\n", "Yes\n", "Ok cool thx\n\n> Am 09.08.2016 um 16:22 schrieb Trustin Lee [email protected]:\n> \n> Yes\n> \n> —\n> You are receiving this because you commented.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "I believe a vert.x user observed this issue when running his program in google cloud (and trying to connect to www.googleapis.com) https://groups.google.com/d/topic/vertx/7xTo_a_lOWk/discussion\n", "would be worth mentioning this issue in the group thread for the record @alexlehm \n", "and perhaps create a corresponding issue in Vert.x project ?\n", "Fixed by https://github.com/netty/netty/pull/5659\n" ]

[ " Make this array \"private\". [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1873)\n", "False positive? It's referenced by `DnsNameResolverBuilder`.\n", "@trustin I think the rule only checks the local scope, the suggestion from Sonarqube is that the elements are mutable even though the field is final and so its better to make it private if it is used only in the class or provide a getter that copies, but this is not applicable to this example\n" ]

[ "defect" ]

[ "common/src/main/java/io/netty/util/NetUtil.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java" ]

[ "common/src/main/java/io/netty/util/NetUtil.java", "resolver-dns/src/main/java/io/netty/resolver/dns/DnsNameResolver.java" ]

[]

[ "timestamp(timedelta=53920.0, similarity=0.9559465783424206)" ]

[ "thanks for reporting @Spikhalskiy ! will look soon\n", "@Spikhalskiy - do you get any requests of type 100? also do you get any requests that exceed the maximum content length? are you able to reproduce the conditions that cause the leak?\n", "> do you get any requests of type 100?\n\nYes, we have some % of requests of type \"100 Continue\".\n\n> do you get any requests that exceed the maximum content length?\n\nNo, we don't get TooLongFrameException in mainHandler.\n\n> are you able to reproduce the conditions that cause the leak?\n\nNo, I don't know which specific request or condition causing this leak. If it's hard to get into this problem with already provided info I would think about a way of collecting more data.\n", "BTW, more info, that could be related. We had a bug that we returned\n\n```\nif (HttpHeaders.is100ContinueExpected(request)) {\n send100Continue(ctx);\n}\n```\n\nin the mainHandler AFTER HttpObjectAggregator, which already does this. And looks like our clients failed after receiving this second \"continue\". After fixing this bug (removing this code from mainHandler) I didn't see leaks anymore, so it looks like something possibly related, or just by chance and I will see leak soon (this bug fixed just yesterday).\n\nOn the same time, I don't get how it could be related because we sent this second continue in mainHandler, which have release() inside channelRead and we should see this release in advanced logs, but we don't. So, it shouldn't be that cases when we send this second continue.\n\nJust a stream of thoughts, that probably could be useful for you.\n", "The `HttpObjectAggregator` owns the responsibility to send a 100 continue response (because it aggregates and knows the limit on how much data you want to accept). Your peer may get confused if you send 2 continue responses. I will throw together a PR which tightens up the release related stuff in `HttpObjectAggregator` and ping you when ready.\n", "@Scottmitch yep, I already got it and bug with double continue fixed now in our server. Thanks for your work, waiting.\n", "Closed by https://github.com/netty/netty/pull/5676\n" ]

[ "This rework doesn't make too much sense because this invocation will always fall to code\n\n```\n private void AbstractChannelHandlerContext#invokeChannelInactive() {\n if (invokeHandler()) {\n try {\n ((ChannelInboundHandler) handler()).channelInactive(this);\n } catch (Throwable t) {\n notifyHandlerException(t);\n }\n } else {\n fireChannelInactive();\n }\n }\n```\n\nSo, looks like the current call `invokeChannelInactive` here is safe in terms of throwing exceptions.\n", "super#handlerRemoved is empty, so should be safe without `try {} catch` too.\n", "practically there should be no difference, but there is no need to make assumptions about what the super class does.\n" ]

[]

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

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

[]

[ "timestamp(timedelta=959.0, similarity=0.841766743990749)" ]

[ "ups... let me fix it.\n", "thanks for reporting @ferrybig !\n", "Fixed by https://github.com/netty/netty/pull/5725\n" ]

[]

[ "cleanup" ]

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

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

[]

[ "timestamp(timedelta=35.0, similarity=0.93808538678266)" ]

[ "@gustavonalle which version exactly ?\n", "@normanmaurer we are on 4.1.0.CR7\n", "Csn you try 4.1.4.Final as well?\n\n> Am 19.08.2016 um 12:25 schrieb Gustavo Fernandes [email protected]:\n> \n> @normanmaurer we are on 4.1.0.CR7\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "@normanmaurer Looking at 4.1.4.Final, the static initializers are unchanged, so I looks like the issue will happen as well\n", "@gustavonalle sorry this was fixed in the 4.1 branch and will be part of 4.1.5.Final once released. See https://github.com/netty/netty/commit/b427a8c8bdb453b45d9f3e39f3fdb15927356e42\n", "@normanmaurer cool thanks\n", "@normanmaurer \n\nA closer look at branch 4.1 shows the cycle may still the there:\n\n`SystemPropertyUtil` requires `InternalLoggerFactory` requires `ThreadLocalRandom` requires `SystemPropertyUtil`\n", "@gustavonalle doh... you are right. Let me fix it.\n", "Fixed by https://github.com/netty/netty/pull/5730\n" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=7555.0, similarity=0.9075777944986547)" ]

[ "@adityakishore so you basically say both should return 1. Right ?\n", "@adityakishore ups -1 I mean :)\n", "Let me fix this.\n", "Yes, both should return -1 if we want to retain the memcmp/strcmp behavior.\n\nFor that, both buffer should be compared as BIG_ENDIAN if we still want to retain the performance optimization of comparing 4 bytes at a time.\n", "@adityakishore PTAL https://github.com/netty/netty/pull/5732\n", "fixed by #5732\n", "Thanks for fixing this.\n" ]

[ "nit: remove space before `=`\n", "if you change `uintCount` to something like `aEnd` you could get rid of `i` and a decrement on every iteration.\n\n``` java\nfor (;aIndex < aEnd; aIndex += 4, bIndex +=4) {\n..\n}\n```\n", "consider removing `va` and `vb` for the following ... trade an aritmatic operation for potentially a conditional\n\n``` java\nlong cmp = bufferA.getUnsignedInt(aIndex) - bufferB.getUnsignedInt(bIndex);\nif (cmp == 0) {\n continue;\n}\nreturn cmp;\n```\n", "I think this could overflow ...\n\n> Am 23.08.2016 um 19:36 schrieb Scott Mitchell [email protected]:\n> \n> In buffer/src/main/java/io/netty/buffer/ByteBufUtil.java:\n> \n> > ```\n> > }\n> > \n> > return aLen - bLen;\n> > }\n> > ```\n> > - private static int compareUintBigEndian(\n> > - ByteBuf bufferA, ByteBuf bufferB, int aIndex, int bIndex, int uintCount) {\n> > - for (int i = uintCount; i > 0; --i, aIndex += 4, bIndex += 4) {\n> > - long va = bufferA.getUnsignedInt(aIndex);\n> > - long vb = bufferB.getUnsignedInt(bIndex);\n> > - if (va == vb) {\n> > - // fast-path\n> > - continue;\n> > - }\n> > - if (va > vb) {\n> > consider removing va and vb for the following ... trade an aritmatic operation for potentially a conditional\n> \n> long cmp = bufferA.getUnsignedInt(aIndex) - bufferB.getUnsignedInt(bIndex);\n> if (cmp == 0) {\n> continue;\n> }\n> return cmp;\n> —\n> You are receiving this because you were assigned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "we are using `long` to represent an \"unsigned int\" ... seems like we should not overflow in this situation?\n", "is `aEnd` correct here? does `uintCount` need to be multiplied by 4 to be in unit of `bytes`?\n", "maybe consider passing in `uintCountIncrement` from outside?\n", "@Scottmitch good catch... Let me fix it.\n" ]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=2141.0, similarity=0.9192202716298955)" ]

[ "Yep, looks like a bug to me. The issue lays in `io.netty.handler.codec.http.cookie.DefaultCookie#equals` implementation that compares names with `equalsIgnoreCase`.\n\nWould you like to contribute a fix?\n", "yes, I'd like\n", "https://github.com/netty/netty/pull/5737\n", "Can be closed, as of #5737\n" ]

[ "I know the spec doesn't say anything here, but it feels very weird to me that domain would be case sensitive. After all, it's more or less the host name of the server that originally set the cookie. And host names are definitively case insensitive. I would vote to keep case insensitivity here.\n", "same as above\n", "ping @normanmaurer @Scottmitch \n", "revert it back\n" ]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=1555.0, similarity=0.8681263283293253)" ]

[ "thanks for reporting @rkapsi ! let me come up with a PR.\n", "Thanks @Scottmitch and @rkapsi \n" ]

[ "Maybe even fire a custom `SniHandlerException(hostname, context, cause)` to give the user every possible opportunity not to lose track of the SslContext.\n", " Remove this unused method parameter \"hostname\". [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1172)\n", "Firing exceptions with reference counted objects can create confusion in my experience. IMO lets avoid this and push folks toward `replaceHandler`.\n" ]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=21.0, similarity=0.902178681582907)" ]

[ "@rkapsi oversight... let me fix it.\n", "Fixed by https://github.com/netty/netty/pull/5782\n" ]

[]

[]

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

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

[]

[ "timestamp(timedelta=14.0, similarity=0.8610196742337676)" ]

[ "Will have a look\n\n> Am 30.08.2016 um 18:22 schrieb Nick Pordash [email protected]:\n> \n> After updating from 4.1.4.Final to 4.1.5.Final I have failing tests because of the following:\n> \n> java.lang.IndexOutOfBoundsException: index: 21, length: 1 (expected: range(0, 21))\n> at io.netty.buffer.AbstractByteBuf.checkIndex0(AbstractByteBuf.java:1359)\n> at io.netty.buffer.AbstractByteBuf.checkIndex(AbstractByteBuf.java:1354)\n> at io.netty.buffer.AbstractByteBuf.checkIndex(AbstractByteBuf.java:1349)\n> at io.netty.buffer.CompositeByteBuf.findComponent(CompositeByteBuf.java:1407)\n> at io.netty.buffer.CompositeByteBuf._getByte(CompositeByteBuf.java:751)\n> at io.netty.buffer.AbstractByteBuf.forEachByteDesc0(AbstractByteBuf.java:1297)\n> at io.netty.buffer.AbstractByteBuf.forEachByteDesc(AbstractByteBuf.java:1277)\n> This seems to be specific to CompositeByteBuf:\n> \n> // Doesn't throw\n> Unpooled.copiedBuffer(\"test\", StandardCharsets.UTF_8)\n> .forEachByteDesc(ByteProcessor.FIND_CR);\n> \n> // Throws (forEachByteDesc)\n> Unpooled.compositeBuffer()\n> .addComponent(true, Unpooled.copiedBuffer(\"test\", StandardCharsets.UTF_8))\n> .forEachByteDesc(ByteProcessor.FIND_CR);\n> \n> // Doesn't throw (forEachByte)\n> Unpooled.compositeBuffer()\n> .addComponent(true, Unpooled.copiedBuffer(\"test\", StandardCharsets.UTF_8))\n> .forEachByte(ByteProcessor.FIND_CR);\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "@npordash I think I have an idea... stay tuned\n", "thanks for the repro @npordash \n", "See https://github.com/netty/netty/pull/5783\n", "Fixed by https://github.com/netty/netty/pull/5783\n" ]

[]

[ "defect" ]

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

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

[ "buffer/src/test/java/io/netty/buffer/AbstractByteBufTest.java", "buffer/src/test/java/io/netty/buffer/SlicedByteBufTest.java", "buffer/src/test/java/io/netty/buffer/WrappedUnpooledUnsafeByteBufTest.java" ]

[ "timestamp(timedelta=12.0, similarity=0.8946751578239116)" ]

[ "/cc @fredericBregier \n", "Let me take care\n", "@normanmaurer Sorry to be late again (very busy those last 9 months). I look at your commit, fine for me.\n", "Fixed by https://github.com/netty/netty/pull/5791\n", "Thanks guys! I could've sent you a poll request but wasn't sure if this is the right design. \n" ]

[]

[ "improvement" ]

[ "handler/src/main/java/io/netty/handler/traffic/AbstractTrafficShapingHandler.java", "handler/src/main/java/io/netty/handler/traffic/GlobalChannelTrafficShapingHandler.java" ]

[ "handler/src/main/java/io/netty/handler/traffic/AbstractTrafficShapingHandler.java", "handler/src/main/java/io/netty/handler/traffic/GlobalChannelTrafficShapingHandler.java" ]

[]

[ "timestamp(timedelta=15.0, similarity=0.9568037118728966)" ]

[ "@buchgr will look into it.\n", "Should be fixed by https://github.com/netty/netty/pull/5798\n", "Fixed by https://github.com/netty/netty/pull/5798\n" ]

[ "This will first cast `capacity` to int, then do the division. I think you need to wrap the division with parantheses?\n", "doh! right... \n", "This cast doesn't seem safe, for `maxHeaderTableSize` gt `MAX_INT`?\n", "@buchgr why ? Shouldn't it just be converted to a signed int ?\n", "@normanmaurer what if it doesn't fit a signed int?\n", "@buchgr good point... Maybe we should do a range check in `setMaxHeaderTableSize(...)` and check its not bigger then `0xFFFFFFFFL`?\n", "@normanmaurer I think that would be good.\n", "done\n", "Does this need to check if capacity is greater than 1<<32 ?\n", "yep let me check for uint as well\n", "Does here need to check as well?\n", "done\n", "done\n", "@normanmaurer You should probably also add a comment to this line, explaining why it's safe.\n", "@normanmaurer - it seems like this may be problematic if `maxHeaderTableSize > Integer.MAX_INT`. Do we have a unit test covering this case?\n", "@normanmaurer - should this be of type `long` now?\n", "should we instead check around the bounds `MAX_HEADER_TABLE_SIZE` and `MAX_HEADER_TABLE_SIZE+1`?\n" ]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2HeadersDecoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2HeadersEncoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2HeaderTable.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Settings.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Decoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/DynamicTable.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2HeadersDecoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/DefaultHttp2HeadersEncoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2CodecUtil.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2HeaderTable.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Settings.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Decoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/DynamicTable.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2SettingsTest.java" ]

[ "timestamp(timedelta=24.0, similarity=0.8764989131870482)" ]

[ "@akurilov let me fix this.\n", "also FYI Netty 5.x has been discontinued.\n", "Fixed by https://github.com/netty/netty/pull/5825\n" ]

[ "src.limit() -> limit\n", "Can you add a comment why you explicitly use `Unpooled` here instead of just using `alloc`?\n", "on line 402 ... can you change `region.transfered()` -> `region.transferred()` while you are here?\n", " Add the missing @deprecated Javadoc tag. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AMissingDeprecatedCheck)\n", "actually I should use the allocator.\n", "false positive\n", "fixed\n", "fixed\n" ]

[ "defect" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java" ]

[ "testsuite/src/main/java/io/netty/testsuite/transport/socket/SocketFileRegionTest.java" ]

[ "timestamp(timedelta=70.0, similarity=0.9335052059039113)" ]

[ "@henrik-lindqvist maybe you can provide a PR ?\n", "or provide a link to an RFC or specification which describes expected behavior and an input which doesn't produce the expected output.\n", "NetUtil.bytesToIpAddress(new byte[] { 83, -1, -20, 16 }, 0, 4) returns \"0.255.255.16\" not \"83.255.236.16\" as expected. See initial post for cause.\n" ]

[ "shouldnt this be 16 ?\n", "I think 15 should be correct. `\"000.000.000.000\".length() == 15`\n", "doh... right\n" ]

[ "defect" ]

[ "codec-socks/src/main/java/io/netty/handler/codec/socksx/v5/Socks5AddressDecoder.java", "common/src/main/java/io/netty/util/NetUtil.java" ]

[ "codec-socks/src/main/java/io/netty/handler/codec/socksx/v5/Socks5AddressDecoder.java", "common/src/main/java/io/netty/util/NetUtil.java" ]

[ "codec-socks/src/test/java/io/netty/handler/codec/socksx/v5/Socks5CommandRequestDecoderTest.java", "codec-socks/src/test/java/io/netty/handler/codec/socksx/v5/Socks5CommandResponseDecoderTest.java", "common/src/test/java/io/netty/util/NetUtilTest.java" ]

[ "timestamp(timedelta=30.0, similarity=0.8635042964389493)" ]

[ "@bryce-anderson interested to provide a PR + unit test ?\n", "@normanmaurer Do you mean PR to fix or PR to demonstrate? I'm happy to provide a PR either way, I just want to be clear on what PR we want. 😄\n\nIf we are talking about a fix, the easy solution is to adopt the netty3 behavior of resetting the encoder on every new `HttpMessage`, but before making the change I think it's worth quickly discussing if that is an acceptable solution: in my minds eye someone introduced that `IllegalStateException` for a reason.\n", "let me fix this...\n", "@bryce-anderson actually I not understand how this happens. Can you provide a unit test that shows it ?\n", "@bryce-anderson nevermind... I think I understand it now\n", "Ah, and I just hacked together an example! Just in case, here it is: https://github.com/bryce-anderson/netty/commit/0fdfcacb67eff649699d266ea6425cb2b93b1218\n\nI apologize in advance, there is probably a cleaner way to do it.\n", "@bryce-anderson PTAL at https://github.com/netty/netty/pull/5846\n", "Fixed by https://github.com/netty/netty/pull/5846" ]

[ "move inner `if` to this level to make an `} else if { ... } else { ... }`\n", "done\n", "Could we define an overridable protected method in `HttpObjectEncoder` rather than duplicating HEAD check in `encode*Content()` methods? That way, `HttpObjectEncoder` could keep `encode*Content()` private. i.e.\n\n``` java\nprotected boolean isContentAlwaysEmpty(HttpMessage msg) { // subclasses will override.\n return false;\n}\n```\n\nActually, this is what we are doing for `HttpObjectDecoder`, so it's probably good to be consistent?\n", "@trustin actually I think this will not work as we need to map request to response which we only want to do in the `HttpServerCodec`. Or I am missing something ?\n", "@trustin ping\n", "@trustin ping again... sorry for being a PITA.", "@Scottmitch as @trustin seems to be busy wdyt ? I think what @trustin suggested can't be done so I think we should pull it in as it is.", "@normanmaurer - I think what @trustin was suggesting was to do something like this:\r\n\r\n\r\n```java\r\nclass HttpObjectEncoder {\r\n final boolean encodeNonChunkedContent(ByteBuf buf, Object msg, long contentLength, List<Object> out) {\r\n // use isContentAlwaysEmpty, release the msg && add empty buff if true\r\n }\r\n\r\n final boolean encodeChunkedContent(ChannelHandlerContext ctx, Object msg, long contentLength, List<Object> out) {\r\n // use isContentAlwaysEmpty, release the msg && add empty buff if true\r\n }\r\n\r\n protected boolean isContentAlwaysEmpty(HttpMessage msg) {\r\n return false;\r\n }\r\n}\r\n\r\n// In HttpServerCodec.java\r\nprivate final class HttpServerResponseEncoder extends HttpResponseEncoder {\r\n @Override\r\n protected boolean isContentAlwaysEmpty(HttpMessage msg) {\r\n if (msg instanceof LastHttpContent) {\r\n HttpMethod method = queue.poll();\r\n return HttpMethod.HEAD.equals(method);\r\n }\r\n return false;\r\n }\r\n}\r\n```", "@Scottmitch Ah got it! Let me update the PR, this is a good idea.", " Remove this unused method parameter \"msg\". [](https://garage.netty.io/sonarqube/coding_rules#rule_key=squid%3AS1172)\n", "we have a bunch of conditionals on `state` ... can we make this a `switch`?", "the `List` interface allows base classes to do some interesting things which may make this iteration not do exactly what we want. Maybe its OK to assume the super class behavior only adds to the list, and doesn't change indexes though bcz its an internal class? This method is also used in HttpClientCodec too. We have discussed before but it seems like a good idea to remove ambiguity and overhead by removing the the list interface (for future releases).", "add a call to `assertFalse(ch.finishAndReleaseAll());` or something similar to assert there is no more data and we clean everything up.", "yep for future releases, for now its ok.", "+1" ]

[ "defect" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpObjectEncoder.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpServerCodec.java" ]

[ "codec-http/src/main/java/io/netty/handler/codec/http/HttpObjectEncoder.java", "codec-http/src/main/java/io/netty/handler/codec/http/HttpServerCodec.java" ]

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

[ "timestamp(timedelta=15.0, similarity=0.9115880020539914)" ]

[ "Let me fix it\n", "Fixes by https://github.com/netty/netty/pull/5876\n" ]

[]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2EventAdapter.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2EventAdapter.java" ]

[]

[ "timestamp(timedelta=18.0, similarity=0.9318689056094328)" ]

[ "@ewie sounds like a bug... let me fix it.\n", "@ewie should be fixed by https://github.com/netty/netty/pull/5877\n", "Fixed by https://github.com/netty/netty/pull/5877\n" ]

[]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=62.0, similarity=0.8592571639964799)" ]

[ "Adding some new details. Turns out we only need to issue one read to trigger the handshake (no matter how many messages it takes to complete for HTTP and SOCKS5) given that [`ProxyHandler.readComplete()`](https://github.com/netty/netty/blob/4.1/handler-proxy/src/main/java/io/netty/handler/proxy/ProxyHandler.java#L387-L389) will keep reading for us.\n", "Fixed by https://github.com/netty/netty/pull/5939\n" ]

[ "ThreadLocalRandom.current().nextBoolean()\n", "Thanks @normanmaurer! Just updated the branch.\n" ]

[ "defect" ]

[ "handler-proxy/src/main/java/io/netty/handler/proxy/ProxyHandler.java" ]

[ "handler-proxy/src/main/java/io/netty/handler/proxy/ProxyHandler.java" ]

[ "handler-proxy/src/test/java/io/netty/handler/proxy/ProxyHandlerTest.java" ]

[ "timestamp(timedelta=4.0, similarity=0.87246764186294)" ]

[ "Fixed by https://github.com/netty/netty/pull/5956\n" ]

[ "@mosesn please add missing `@Override`\n", "private final static\n", "private final static\n", "private final static\n", "@mosesn please add missing `@Override`\n", "assert return value\n", "assert return value\n", "assert return value\n", "assert return value\n", "ensure you call `channel.finish()` and assert the return value. Also ensure you consume all messages and call `release()` on them.\n", "same comments as above\n", "same comments as above\n", "@mosesn just a NIT but I think the methodname is kind of misleading as we not really check anything but just return the event. \n", "Isnt this a `FullHttpResponse` ? If so you should also call `release()` on it.\n", "Isnt this a `FullHttpRequest` ? If so you should also call `release()` on it.\n", "Isnt this a `FullHttpRequest` ? If so you should also call `release()` on it.\n", "`twitter.com` -> `netty.io` ;)\n", "call `last.release()`\n", "Isnt this a `FullHttpRequest` ? If so you should also call `release()` on it.\n", "whoops (◕ ω ◕✿)\n" ]

[ "improvement" ]

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

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

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

[ "timestamp(timedelta=98856.0, similarity=0.9859878293651412)" ]

[ "Maybe want to provide a PR with a fix?\n\n> Am 26.10.2016 um 21:25 schrieb ichaki5748 [email protected]:\n> \n> netty 4.1.6\n> \n> io.netty.handler.ssl.ReferenceCountedOpenSslEngine.OpenSslSession#initPeerCerts\n> \n> final byte[] clientCert;\n> ...\n> int len = clientCert.length + 1; // <------ BUG: raw certificate byte array size is used\n> peerCerts = new Certificate[len];\n> x509PeerCerts = new X509Certificate[len];\n> \n> But should probably be\n> \n> int len = chain.length + 1; // <------ so we have first entry as client cert + chain.length\n> \n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "Yes, sure, will try to do it this week \n", "Thx\n\n> Am 26.10.2016 um 23:54 schrieb ichaki5748 [email protected]:\n> \n> Yes, sure, will try to do it this week\n> \n> —\n> You are receiving this because you commented.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "Also `io.netty.handler.ssl.OpenSslJavaxX509Certificate.*`\ncould theoretically blow in concurrent usage because of lazy init in `io.netty.handler.ssl.OpenSslJavaxX509Certificate#unwrap` because \n`io.netty.handler.ssl.OpenSslJavaxX509Certificate#wrapped` is not `volatile` (certificate would not be fully initialized)? \n", "@iainmcgin please open separate PRs for the separate issues and ping me once you are ready. I will review then\n", "You probably meant to reference @ichaki5748\n", "@iainmcgin ups yes... sorry I need more ☕️ \n", "@normanmaurer Hi, please see https://github.com/netty/netty/pull/5967 https://github.com/netty/netty/pull/5966\n", "Fixed by https://github.com/netty/netty/pull/5967\n" ]

[]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslJavaxX509Certificate.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslX509Certificate.java" ]

[ "handler/src/main/java/io/netty/handler/ssl/OpenSslJavaxX509Certificate.java", "handler/src/main/java/io/netty/handler/ssl/OpenSslX509Certificate.java" ]

[]

[ "timestamp(timedelta=12.0, similarity=0.9859878293651412)" ]

[ "Maybe want to provide a PR with a fix?\n\n> Am 26.10.2016 um 21:25 schrieb ichaki5748 [email protected]:\n> \n> netty 4.1.6\n> \n> io.netty.handler.ssl.ReferenceCountedOpenSslEngine.OpenSslSession#initPeerCerts\n> \n> final byte[] clientCert;\n> ...\n> int len = clientCert.length + 1; // <------ BUG: raw certificate byte array size is used\n> peerCerts = new Certificate[len];\n> x509PeerCerts = new X509Certificate[len];\n> \n> But should probably be\n> \n> int len = chain.length + 1; // <------ so we have first entry as client cert + chain.length\n> \n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "Yes, sure, will try to do it this week \n", "Thx\n\n> Am 26.10.2016 um 23:54 schrieb ichaki5748 [email protected]:\n> \n> Yes, sure, will try to do it this week\n> \n> —\n> You are receiving this because you commented.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "Also `io.netty.handler.ssl.OpenSslJavaxX509Certificate.*`\ncould theoretically blow in concurrent usage because of lazy init in `io.netty.handler.ssl.OpenSslJavaxX509Certificate#unwrap` because \n`io.netty.handler.ssl.OpenSslJavaxX509Certificate#wrapped` is not `volatile` (certificate would not be fully initialized)? \n", "@iainmcgin please open separate PRs for the separate issues and ping me once you are ready. I will review then\n", "You probably meant to reference @ichaki5748\n", "@iainmcgin ups yes... sorry I need more ☕️ \n", "@normanmaurer Hi, please see https://github.com/netty/netty/pull/5967 https://github.com/netty/netty/pull/5966\n", "Fixed by https://github.com/netty/netty/pull/5967\n" ]

[ "@ichaki5748 can you make this method static\n", "please make static\n", "sure\n", " Consider using varargs for methods or constructors which take an array the last parameter. [](https://garage.netty.io/sonarqube/coding_rules#rule_key=pmd%3AUseVarargs)\n" ]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=20.0, similarity=0.9229476379369648)" ]

[ "Fixed by https://github.com/netty/netty/pull/5983\n" ]

[ "just make this one listener which always close after logging the error ? No need to have to listeners and at worst have both call close. After this I will merge.\n\n@bryce-anderson thanks!\n", "Done. Let me know if I misunderstood your comment. 😄 \n", "@bryce-anderson remove empty line and we are good :)\n", "Done. Sorry for the noise.\n" ]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=34.0, similarity=0.909561793045737)" ]

[ "One of the fields in PlatformDependent references PlatformDependent0 directly without being guarded by hasUnsafe.\n", "Is the issue you don't want to see the warning message, or is this causing a functional issue?\n\n@johnou - which field do you think is an issue?\n", "@johnou - your commit https://github.com/johnou/netty/commit/cff208eb2895134a487424b4b2dcd9c5cafc9f99 changes the initialization of `BYTE_ARRAY_BASE_OFFSET` ... is their a problem with just using the default as set by `PlatformDependent0` (currently `-1`)? IIRC it is assumed whenever this value is used it is guarded by an unsafe check ... did we miss a case?\n", "@Scottmitch BYTE_ARRAY_BASE_OFFSET is set to -1 in PlatformDependent0 if unsafe is unavailable, and if unsafe is unavailable, so is PlatformDependent0, therefore we need to use indirection when accessing it (cannot be directly how it is now otherwise PlatformDependent cannot be loaded)..\n\njava.lang.NoClassDefFoundError: Could not initialize class io.netty.util.internal.PlatformDependent0\n at io.netty.util.internal.PlatformDependent.<clinit>(PlatformDependent.java:101)\n", "@Scottmitch: this causes a functional issue in environments where sun.misc is not available, which is the default in OSGi frameworks\nI can make a pull request tomorrow, extracting the BYTE_ARRAY_BASE_OFFSET initialization into a method like addressSize0 does for ADDRESS_SIZE. Would that be something you could consider including?\n\nEdit: [johnou@cff208e](https://github.com/johnou/netty/commit/cff208eb2895134a487424b4b2dcd9c5cafc9f99) should fix the issue\n", "@jellenelis see https://github.com/netty/netty/pull/5988\n", "Thanks for the quick responses, BTW!\n", "thanks for clarifying ... we want to avoid loading `PlatformDependent0` if the `Unsafe` interface is not available.\n", "Fixed by https://github.com/netty/netty/pull/5988\n", "Hello I'm using netty version 4.1.8.Final on Apache Felix and still suffer from the same problem!\r\nlb:\r\n0|Active | 0|System Bundle (5.4.0)|5.4.0\r\n 1|Active | 1|Guava: Google Core Libraries for Java (18.0.0)|18.0.0\r\n 2|Active | 1|Jackson-annotations (2.7.0)|2.7.0\r\n 3|Active | 1|Jackson-core (2.7.4)|2.7.4\r\n 4|Active | 1|jackson-databind (2.7.4)|2.7.4\r\n 5|Active | 1|com.github.shumy.leffe-wg (1.0.0)|1.0.0\r\n 6|Active | 1|Netty/Buffer (4.1.8.Final)|4.1.8.Final\r\n 7|Active | 1|Netty/Codec (4.1.8.Final)|4.1.8.Final\r\n 8|Active | 1|Netty/Codec/DNS (4.1.8.Final)|4.1.8.Final\r\n 9|Active | 1|Netty/Codec/HTTP (4.1.8.Final)|4.1.8.Final\r\n 10|Active | 1|Netty/Codec/HTTP2 (4.1.8.Final)|4.1.8.Final\r\n 11|Active | 1|Netty/Codec/Socks (4.1.8.Final)|4.1.8.Final\r\n 12|Active | 1|Netty/Common (4.1.8.Final)|4.1.8.Final\r\n 13|Active | 1|Netty/Handler (4.1.8.Final)|4.1.8.Final\r\n 14|Active | 1|Netty/Handler/Proxy (4.1.8.Final)|4.1.8.Final\r\n 15|Active | 1|Netty/Resolver (4.1.8.Final)|4.1.8.Final\r\n 16|Active | 1|Netty/Resolver/DNS (4.1.8.Final)|4.1.8.Final\r\n 17|Active | 1|Netty/Transport (4.1.8.Final)|4.1.8.Final\r\n 18|Active | 1|Apache Felix Configuration Admin Service (1.8.14)|1.8.14\r\n 19|Active | 1|Apache Felix Gogo Command (0.16.0)|0.16.0\r\n 20|Active | 1|Apache Felix Gogo Runtime (0.16.2)|0.16.2\r\n 21|Active | 1|Apache Felix Gogo Shell (0.12.0)|0.12.0\r\n 22|Active | 1|Apache Felix Declarative Services (2.0.8)|2.0.8\r\n 23|Active | 1|Xtend Runtime Library (2.11.0.v20170124-1424)|2.11.0.v20170124-1424\r\n 24|Active | 1|Xtend Macro Interfaces (2.11.0.v20170124-1424)|2.11.0.v20170124-1424\r\n 25|Active | 1|Xbase Runtime Library (2.11.0.v20170124-1424)|2.11.0.v20170124-1424\r\n 26|Active | 1|OPS4J Pax Logging - API (1.9.1)|1.9.1\r\n 27|Active | 1|Vert.x Core (3.4.1)|3.4.1\r\n", "@shumy please open a new issue with an updated stacktrace.", "@shumy also a reproducer would be nice.", "New issue #6548", "I run netty4.1.42 under jdk11.0.6 and reports java.lang.UnsupportedOperationException: sun.misc.Unsafe unavailable, I don't know how to solve it?" ]

[]

[ "defect" ]

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

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

[]

[ "timestamp(timedelta=40.0, similarity=0.8763541037290001)" ]

[ "@nhnFreespirit so just to ensure I understand you right you suggest that we should for example allow to pass it in into the constructor ?\n", "Yeah, having it in a constructor of Http2Codec would solve my issues nicely.\n\nWhen constructing an http/2 client, it is possible to specify non default initial settings by subclassing AbstractHttp2ConnectionHandlerBuilder and invoking the initialSettings method, but unless I am missing something, there is no reasonable way to accomplish the same when using Http2Codec to build an http/2 server.\n", "I have a version of Netty with this added.\n\nOnly problem is I don't really like it, as to keep the initial settings optional, I have to add 2 new constructors to Http2Codec, so hesitant to create a pull request for this.\n", "created https://github.com/netty/netty/pull/5991\n", "Fixed by https://github.com/netty/netty/pull/5991#issuecomment-259270529\n" ]

[]

[ "improvement" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Codec.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameCodec.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2Codec.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/Http2FrameCodec.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/Http2FrameCodecTest.java" ]

[ "timestamp(timedelta=31.0, similarity=0.8573709676194836)" ]

[ "Related? https://github.com/netty/netty/pull/4254\n", "Yes, seems so. Thanks for the pointer.\n", "@jrudolph would be possible to submit a small unit test that works with JDK SSLEngine and fails with the OpenSslEngine ?\n", "@normanmaurer yes, will try to build one, shouldn't be too hard.\n", "Thanks a lot!\n\n> Am 03.11.2016 um 15:19 schrieb Johannes Rudolph [email protected]:\n> \n> @normanmaurer yes, will try to build one, shouldn't be too hard.\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "@jrudolph also when you say \"the JDK SSLEngine doesn't buffer encrypted data internally between calls to unwrap but \"puts them back\" into the src buffer\" do you mean that if you not completely drain stuff into the dst buffer it will just adjust the position of the src buffer and so try to process it on the next call again ?\n", "Here's a test I added to SslEngineTest: https://github.com/jrudolph/netty/tree/w/5975-SslEngine-test\n\nIt seems I was wrong about my first point. The JDK SSLEngine also only decrypts only one frame per unwrap call. Still it doesn't consume data that it hasn't read so far.\n\n> @jrudolph also when you say \"the JDK SSLEngine doesn't buffer encrypted data internally between calls to unwrap but \"puts them back\" into the src buffer\" do you mean that if you not completely drain stuff into the dst buffer it will just adjust the position of the src buffer and so try to process it on the next call again ?\n\nHmm, no, I originally don't meant to imply that case. But, indeed, the difference can be seen here as well. JDK returns BUFFER_OVERFLOW without consuming the src buffer while openssl consumes the buffer. I added a commented out block in the test to check this behavior as well.\n", "@jrudolph let me look into it\n", "I think the test case is not much different than @ijuma's [here](https://github.com/ijuma/netty/commit/0f6eff24885bfbe965d0cabd61115067d14feedc#diff-08da41869dfc1d842df441440b1c605bR232), so it might indeed just be a duplicate of #4238. Hi, @ijuma btw ;)\n", "@jrudolph are you license the shared code as ASL2 so I can include it in netty when I have a fix ?\n", "@jrudolph so after some digging I found out that the SSLEngineImpl (the JDK one) always only consume if a full packet is present while we even consume when it is only partly present. Both implementations correctly return `BUFFER_OVERFLOW` tho. I wonder if what we do is permitted.. I could not see details in the javadocs of `SSLEngine` which state that you can not consume a partial record. I am missing anything ?\n\nThe other \"problem\" is that our implementation will consume multiple records in one unwrap call but not write multiple out to the dsts buffers. The JDK implementation only consume one one record at a time (per unwrap call). Again I am not sure if what we do is allowed in terms of the the API and is just an implementation detail. That said I think we should also try to fill more then one record into the dsts buffers if we also consume more then one. \n\n@jrudolph WDYT ?\n", "@jrudolph actually I think this sentence may state we are not allowed to consume and still return a `BUFFER_OVERFLOW`:\n\n```\nThe SSLEngine produces/consumes complete SSL/TLS packets only, and does not store application data internally between calls to wrap()/unwrap().\n```\n\nAgree ?\n", "@jrudolph ok nevermind.. I think I know how to \"fix\" the problems stated here. Will need to check what performance impact this has and if it is too much I may introduce a \"strict\" and non \"strict\" mode for the `OpenSslEngine`. Will keep you posted.\n", "> @jrudolph are you license the shared code as ASL2 so I can include it in netty when I have a fix ?\n\nYes, here as well ;)\n", "> @jrudolph actually I think this sentence may state we are not allowed to consume and still return a BUFFER_OVERFLOW:\n> \n> The SSLEngine produces/consumes complete SSL/TLS packets only, and does not store application data internally between calls to wrap()/unwrap().\n> Agree ?\n\nYes, exactly. (I quoted that sentence above as well.) Hard to say if this is a requirement or just a description of the implementation. It's probably impossible to take the javadoc as a water-tight specification (I guess you know best about that...).\n\nIMO it's nice if you can rely on the openssl engine not to keep user data in internal buffers (if possible). I haven't looked into it in detail but how does the openssl engine provide backpressure? What happens if you always call `unwrap` e.g. with input buffers of size 30000 and output buffers of size 20000? Will data start to accumulate internally in the openssl engine or will it stop consuming the input buffer at some point?\n", "I have a fix locally here so it works as the JDK impl :) will open a PR later today 👍\n", "@normanmaurer with the strict / non-strict as you previously mentioned? or did you manage to keep it simple?\n", "Was able to keep it simple \n\n> Am 08.11.2016 um 15:51 schrieb Johno Crawford [email protected]:\n> \n> @normanmaurer with the strict / non-strict as you previously mentioned? or did you manage to keep it simple?\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n", "@jrudolph should be fixed by https://github.com/netty/netty/pull/5995\n", "Fixed by https://github.com/netty/netty/pull/5995\n", "@jrudolph - I know it's not nice to comment on an ancient closed topic especially when it's off-topic, but I will like to say a BIG TANK YOU. Your findings and description solved my 2 days no-sleep pain with SSLEngine, where the unconsumed data was copied back in the src, leaving my implementation hanging and waiting for more data. You are a life saver, so again a big THANKS!" ]

[ "@normanmaurer Have you verified that doing this each time will not impact performance? SslHandler would cache the packetLength so that it wouldn't recalculate it each time. I suspect it's not an issue.\n", "Is it correct to use a constant here or should we get this value from the session? Or does the session always return this constant?\n", "It always return the constants\n", "I ran a benchmark and the perf was the same\n" ]

[ "defect" ]

[ "handler/src/main/java/io/netty/handler/ssl/ReferenceCountedOpenSslEngine.java", "handler/src/main/java/io/netty/handler/ssl/SslHandler.java", "handler/src/main/java/io/netty/handler/ssl/SslUtils.java" ]

[ "handler/src/main/java/io/netty/handler/ssl/ReferenceCountedOpenSslEngine.java", "handler/src/main/java/io/netty/handler/ssl/SslHandler.java", "handler/src/main/java/io/netty/handler/ssl/SslUtils.java" ]

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

[ "timestamp(timedelta=13.0, similarity=0.8963185923985323)" ]

[ "Sounds ok to me... @Scottmitch ?\n", "sgtm", "@derbylock could you provide a PR ?", "Fixed by https://github.com/netty/netty/pull/6059" ]

[]

[]

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

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

[]

[ "timestamp(timedelta=20.0, similarity=0.8506734955420215)" ]

[ "Will check later today\n\n> Am 25.11.2016 um 04:45 schrieb byeongguk.gim <[email protected]>:\n> \n> Netty version: 4.1.0.Final\n> \n> Context: I found bug on EpollServerDomainSocketChannel class. Its isActive() method always return false. It is required to set isActive = true in doBind() operation. (same as EpollServerSocketChannel)\n> \n> Steps to reproduce:\n> \n> Use EpollServerDomainSocketChannel class to server bootstrap's channel class\n> Bind any unix domain socket.\n> Server Channel's channelActive() is not called due to EpollServerDomainSocketChannel.isActive() return false.\n> $ java -version\n> java version \"1.8.0_45\"\n> Java(TM) SE Runtime Environment (build 1.8.0_45-b14)\n> Java HotSpot(TM) 64-Bit Server VM (build 25.45-b02, mixed mode)\n> \n> Operating system: Cent OS 7.2 64-bit\n> \n> $ uname -a\n> Linux 3.10.0-327.22.2.el7.x86_64 #1 SMP Thu Jun 23 17:05:11 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux\n> \n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub, or mute the thread.\n> \n", "@byoengguk thanks for reporting... you are right. Let me fix this", "Fixed by https://github.com/netty/netty/pull/6078" ]

[]

[ "defect" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollServerDomainSocketChannel.java" ]

[ "transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollServerDomainSocketChannel.java" ]

[]

[ "timestamp(timedelta=20.0, similarity=0.8803081878243956)" ]

[ "@maseev - I had not heard about this ... but this sounds great thanks!" ]

[ "typo ihttp -> http", "Why not copy direct from http://netty.io/wiki/writing-a-commit-message.html? (I think the sentences under each header help a lot when thinking of what to write).", "I'm currently experimenting with the format. Markdown looks nice in UIs that support it and helps visually separate different sections/titles and still looks reasonable as plain text.", "nit caps? Minimal yet complete reproducer code", "as @johnou pointed out this is currently different from http://netty.io/wiki/writing-a-commit-message.html ... @normanmaurer @trustin @nmittler do you have a preference or objection to this (I will update the wiki if we are OK with this). Markdown looks nice in UIs that support it and helps visually separate different sections/titles and still looks reasonable as plain text.", "Maybe something like this?\r\n\r\n```text\r\n**Motivation:**\r\n\r\nExplain here the context, and why you're making that change.\r\nWhat is the problem you're trying to solve.\r\n\r\n**Modifications:**\r\n\r\nDescribe the modifications you've done.\r\n\r\n**Result:**\r\n\r\nAfter your change, what will change.\r\n```", "not strong here but the intention is that this is a \"section header/title\" (not a sentence)", "I'll include the descriptions for each section, but I prefer the Markdown section header. That way extra \"hard coded\" newlines aren't necessary ... the spacing is determined by the styles.", "I would prefer to not use markdown in commit messages honestly.", "For Netty version, you can use something very similar to [this](https://github.com/orientechnologies/orientdb/blob/master/.github/ISSUE_TEMPLATE.md#orientdb-version-operating-system-or-hardware). You can also declare that Netty 3.x is EOL and no longer supported and a person needs to upgrade to 4.0.x or 4.1.x. WDYT?", "I would prefer to not update this template with every new release. I'm fine with leaving the field free form for now and updating it later if the field is being abused.\r\n\r\nWe could add a note that 3.x is EOL but it may be useful for folks to still report the issue for knowledge sharing, and in-case someone is motivated enough they can fix it.", "@normanmaurer - You don't like the way it looks in UIs, just think it is too noisy as plain text, or something else?", "Fair enough.", "I don't like to include markdown as not all tools support it and it looks ugly if they not do. For example I often use `git log` on the cmdline etc.", "To keep it consistent with the Netty wiki I would remove caps from every word even if it's a section header / title.", "@normanmaurer You're not a fan of hashtags? :)\r\n\r\nHonestly, the basic formatting used by this template wouldn't be that ugly with pure text-based tools. The only ugly bit would be if the user added a bunch of markdown within the comments. ", "I agree with @nmittler, but I don't want to get hung up on this cause I think its roughly the same either way ... @normanmaurer do you still want me to remove the markdown for section headers?", "sgtm", "I just like to keep it simple in commit messages which basically means no markdown ;) But if you guys really love it I am ok with it. ", "@normanmaurer - I'll just stay consistent with the wiki ... we can update later if we want.", "this extra vertical space hurts my eyes 😜 " ]

[ "documentation" ]

[]

[ ".github/CONTRIBUTING.md", ".github/ISSUE_TEMPLATE.md", ".github/PULL_REQUEST_TEMPLATE.md" ]

[]

[ "timestamp(timedelta=126.0, similarity=0.8475656572580857)" ]

[ "@jo-kin - Thanks for reporting. Stand by for a PR.", "Thanx a lot.\r\nDeeply impressed by speed and thoroughness. Will report how it works for me soon.\r\nWishing you a very Happy Christmas and New Year 2017. " ]

[ "@Scottmitch just remove the `@return` as it is the same as what is stated already ? ", "@Scottmitch `:-]` looks nice and friendly ;)", "sure" ]

[ "defect" ]

[ "common/src/main/java/io/netty/util/internal/MacAddressUtil.java", "transport/src/main/java/io/netty/channel/DefaultChannelId.java" ]

[ "common/src/main/java/io/netty/util/internal/MacAddressUtil.java", "transport/src/main/java/io/netty/channel/DefaultChannelId.java" ]

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

[ "timestamp(timedelta=59.0, similarity=0.9681198608469817)" ]

[ "Whether this is a bug or missing feature is open to interpretation ... however if we do want to provide this functionality I don't think we should change the existing default behavior for `IdleStateHandler` for backwards compatibility reasons.\r\n\r\n`ChannelOutboundBuffer` may be able to provide the tools we need...here are some random thoughts:\r\n- The ChannelOutboundBuffer provides a counter of \"pending bytes\" but it is currently only updated at the \"message\" (aka `ByteBuf`, `FileRegion`, etc...) granularity). Even if this was updated at a finer granularity it will not work to answer the \"has something changed\" between successive calls (ABA situation may occur).\r\n- An alternative approach of \"save a reference to the head of the queue, and current size then check these changed after the timeout\" may also be problematic because we pool objects (again ABA). However if we couple this second approach with the completion of the future we may be able to overcome this ABA type problem. The completion of the future would mean we are now looking at a \"different\" object (even if it happens to be the same object because of pooling).\r\n\r\nIt would be good to explore if we can pull this off with existing tools rather than add explicit support for it in the core (unless we think this will drive other useful features).", "@rkapsi @Scottmitch why can't this be fixed by using a `ChannelProgressivePromise` ?", "@normanmaurer it'd work at the expense of garbage. `IdleStateHandler` (or some other class) would have to create and swap out the promise in the `write(...)` method and a listener that will notify the original promise that was passed into `write(...)`. A technique I'd be fine using in my project where I know the nature of the ChannelPipeline. Adding something like that to Netty could be risky. Ideally I'd like to avoid the garbage though.\r\n\r\nI think for pure idless detection there is a difference in measuring \"change\" and \"progress\". And then there's also overall progress and progress in the context of a `write(...)`. The `ChannelProgressivePromise` covers the latter but it comes at an expense. I'd be happy with an overall progress indicator (a `long` value that says: \"This Channel has transferred this many bytes in its lifetime\"). Lastly something that indicates change would be fine too (for this purpose) but I don't think it'd be a good API addition.\r\n\r\n\r\n\r\n", "@rkapsi - does the approach I mentioned in https://github.com/netty/netty/issues/6150#issuecomment-268414509 work for you?", "@Scottmitch - I think No. 1 wouldn't work for the reason you described. No. 2 should work. We'd get the `ChannelOutboundBuffer` via `Channel#unsafe()`?", "> We'd get the ChannelOutboundBuffer via Channel#unsafe()?\r\n\r\nYip. Yah the second bullet is the approach I was referring to ... the first bullet was just to clarify that only using the existing \"pending bytes\" counter won't be sufficient.", "@Scottmitch want me to look into this? I think it can be added to `IdleStateHandler` with a configuration option (i.e. default will retain the current behavior).", "@rkapsi - Sounds good to me. Thanks!" ]

[ "maybe not when this method is currently called but it is possible the `ChannelOutboundBuffer` may be `null` (e.g. channel is closed) .. consider protecting against this or adding a comment as to why it isn't necessary.", "protect against `buf == null`.", "nit: put on previous line?", "consider just setting the variable unconditionally. conditionals can be expensive and this variable is not volatile.", "can we make this method always return NS granularity and do the conversion in the unit tests?", "use `channel.finishAndReleaseAll();` instead for `EmbeddedChannel`.", "use `channel.finishAndReleaseAll();` instead for `EmbeddedChannel`.", "use `channel.finishAndReleaseAll();` instead for `EmbeddedChannel`.", "we should take care to release any object returned by `channel.consume()`", "hah mathematical unit test proof done I guess", "Can be done but is there a downside with the current impl? `TimeUnit#convert(...)` will be a ~noop if source and destination units are the same (which is the case inside the IdleStateHandler class).", "My preference is to keep the production code minimal, and not have to think/worry about implementation details and/or what optimizations may or may not be made if possible. In this case it looks like the extra complexity can be isolated in the unit tests and thus simplify the production code, right?", "Certainly and fair enough. I'm personally of the opinion that time without units is among some of the most useless pieces of information one can have in computer programs and I never write code without stating it explicitly (encoding the unit in the field/method name being my least favorite but still better than noting stated at all).", "Agreed that time (or values in general) without units can be error prone. However not accepting `TimeUnit` as an argument doesn't mean you cannot indicate the units in other ways (as you indicated). For example in the function name (e.g. `timeNanos()`, `nanoTime()`) and javadocs. This function is also \"internal\" to this class and it not intended to be part of the public API of this class so the risk of confusion due to units is lower in this case. This is effectively equivalent to `System.nanoTime()`.", "Done." ]

[]

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

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

[ "handler/src/test/java/io/netty/handler/timeout/IdleStateHandlerTest.java" ]

[ "timestamp(timedelta=24.0, similarity=0.894930691689767)" ]

[ "@markdascher I think this is fixed in 4.1.7.Final-SNAPSHOT (current 4.1 branch). Can you please test and let us know ?", "Wow, wasn't expecting an answer so fast. 👍\r\n\r\nI just downloaded [netty-all-4.1.7.Final-20161230.120821-77.jar](https://oss.sonatype.org/content/repositories/snapshots/io/netty/netty-all/4.1.7.Final-SNAPSHOT/netty-all-4.1.7.Final-20161230.120821-77.jar), but the behavior is the same.", "Ok thx will have a look", "thanks for the detailed issue description @markdascher! seems reasonable to me and very helpful.", "Yay!\r\n\r\nThe best workaround I've found, for the moment, is to add a listener to `SslHandler.sslCloseFuture()` to close the TCP connection immediately. For example, replace [line 47](https://github.com/netty/netty/blob/netty-4.1.6.Final/example/src/main/java/io/netty/example/securechat/SecureChatServerInitializer.java#L47) of SecureChatServerInitializer.java with:\r\n\r\n```\r\nSslHandler sslHnd = sslCtx.newHandler(ch.alloc());\r\nsslHnd.sslCloseFuture().addListener(\r\n\tnew GenericFutureListener<Future<Channel>>() {\r\n\t\t@Override\r\n\t\tpublic void operationComplete(Future<Channel> future) throws Exception {\r\n\t\tif (future.isSuccess()) {\r\n\t\t\tfuture.getNow().close();\r\n\t\t}\r\n\t}\r\n});\r\npipeline.addLast(sslHnd);\r\n```\r\n\r\nThat at least prevents the connection from sitting idle while the client waits for an answer, but it'll still be better to have the close_notify go out first. I'm pretty new to Netty, so please correct me if there's a better way to do it. (e.g. Would `.disconnect()` be better than `.close()`?)", "Looking now...", "Stay tuned for a PR... I think I know the best way to fix this.", "@markdascher https://github.com/netty/netty/pull/6189 should fix it... Can you check ?", "Looks like that does the trick! I built netty from the `fix_close_notify` branch, and now see the close_notify whether or not I use `netty-tcnative-boringssl-static-1.1.33.Fork25.jar`.\r\n\r\nThanks!", "@markdascher boom! Thanks for reporting back :)", "@markdascher and thanks again for the detailed description how to reproduce it easily. Without this it would have take a lot longer.", "Fixed by https://github.com/netty/netty/pull/6189" ]

[ "nit: revert to be consistent with other switch cases (and make merging easier :) )", "futher -> further\r\n\r\nor just `UNWRAP/WRAP are not expected after this point`", "fixed", "fixed" ]

[ "defect" ]

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

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

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

[ "timestamp(timedelta=12.0, similarity=0.8699828685304778)" ]

[ "@ZuluForce maybe you can come up with a PR that fixes the issue for you ?", "We should consider using [toAddressString](https://github.com/netty/netty/blob/4.1/common/src/main/java/io/netty/util/NetUtil.java#L948) for the address portion which follows the RFC 5952 formatting (last time I checked java didn't). We could also add a method to format a address+port which handles ipv6 and ipv4 correctly ... this is common enough.", "@Scottmitch +1 Want to take a stab ?", "Thanks for the replies @normanmaurer and @Scottmitch. I would need to get the time from my employer to work on this and I currently don't have the Netty development environment setup or approval for the CCLA. If I could submit something it wouldn't be until after the holidays (ie early to mid January).", "@ZuluForce @Scottmitch let me take care.", "@ZuluForce PTAL https://github.com/netty/netty/pull/6195", "Fixed by https://github.com/netty/netty/pull/6195" ]

[ "I wonder if it would should be more explicit about the string concatenation operations used here. For example it seems unlikely JIT will be able to combine all of these conditional concatenation operations into a single `StringBuilder`. Can we create a single `StringBuilder` and manually do the `append` calls (e.g. including what is now building `rhost`) to be sure we don't create multiple `String`/`StringBuilder` operations?", "`enought` -> `enough`", "also remove the space before the period?", "`enought` -> `enough`\r\n\r\nalso remove the space before the period?", "nit: no need for `else`", "I added the space before to make it clear it is not part of it... But I guess its ok to remove it as well." ]

[ "defect" ]

[ "common/src/main/java/io/netty/util/NetUtil.java", "handler-proxy/src/main/java/io/netty/handler/proxy/HttpProxyHandler.java" ]

[ "common/src/main/java/io/netty/util/NetUtil.java", "handler-proxy/src/main/java/io/netty/handler/proxy/HttpProxyHandler.java" ]

[ "common/src/test/java/io/netty/util/NetUtilTest.java", "handler-proxy/src/test/java/io/netty/handler/proxy/HttpProxyHandlerTest.java" ]

[ "timestamp(timedelta=30.0, similarity=0.8704484361788252)" ]

[ "Thanks for reporting @rkapsi ... standby for a PR." ]

[]

[ "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/AbstractByteBufTest.java" ]

[ "timestamp(timedelta=57946.0, similarity=0.8625178401546824)" ]

[ "@Scottmitch @nmittler can you take a look ?", "I've been helping @exell-christopher with debugging this.\r\n\r\nOur current understanding is that there really is no safe way to recover from an exception thrown in the Decoder except for tearing down the connection as the index tables in client and server can have gotten out of synch with no way to safely bring them back in sync.", "This is even called out in the RFC. Makes sense since HPACK is stateful.\r\n\r\nhttps://tools.ietf.org/html/rfc7540#section-10.5.1\r\n> The header block MUST be processed to ensure a consistent\r\n connection state, unless the connection is closed.\r\n\r\nIntroducing a \"limit for how much we allow exceeding another limit\" is starting to get intense. To make things interesting we enforce the limit at 2 levels (during accumulating the header blocks, and during the decompression). So implementing the \"continue to process and bit bucket\" approach will not be limited in scope to `hpack.Decoder`. A simpler alternative is just to GO_AWAY/close the connection if the peer violates the settings we advertise.", "For our use case, sending a go-away as the behavior for any overflow is not desirable. The client in our case is a proxy and the inbound connection is heavily multiplexed. With what you are suggesting, one request that exceeds the limit by even one byte causes all other requests to be dropped. Having a reasonable overflow amount allows us to continue servicing the other requests and reject just the one that was oversized, but not particularly malicious.", "@exell-christopher - Yes I understand the implications. Just requires a bit more effort to implement, and the memory upper bound would not be set by `SETTINGS_MAX_HEADER_LIST_SIZE` but some other implementation dependent limit. This is not necessarily a bad thing but just needs to be communicated clearly (via javadocs, docs, etc...).\r\n\r\nJust curious in your scenario what do you anticipate your \"malicious threshold\" to be vs. `SETTINGS_MAX_HEADER_LIST_SIZE`?", "I would likely use MAX_HEADER_LIST_SIZE + (MAX_HEADER_LIST_SIZE * .25) I would expect that the .25 would be user configurable", "I'll take a stab at this...there is some other related cruft that can be cleaned up too.", "Thanks. Happy to submit a PR as well.", "@exell-christopher - Already in progress :) I'll ping you for review when I have a PR ready.", "Some information about summarize the header length,\r\nIn [rfc7540](https://tools.ietf.org/html/rfc7540#section-6.5.2)\r\n> SETTINGS_MAX_HEADER_LIST_SIZE (0x6): This advisory setting informs a\r\n peer of the maximum size of header list that the sender is\r\n prepared to accept, in octets. The value is based on the\r\n uncompressed size of header fields, including the length of the\r\n name and value in octets plus an overhead of 32 octets for each\r\n header field.\r\n\r\nI think that means when we sum up all headers, we should plus 32 byte for each header.\r\nMaybe it could be implemented at the same time! [Decoder.jara#L404](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Decoder.java#L404)", "Sorry, there is a better reference [rfc7541](https://tools.ietf.org/html/rfc7541#section-4.1)\r\n> The size of an entry is the sum of its name's length in octets (as\r\n defined in Section 5.2), its value's length in octets, and 32.", "@Scottmitch Let me know if there's anything you need on this. ", "@exell-christopher - All good for now thanks. I will ping you once the PR is pushed (should be soon).", "@chhsiao90 - We can consider this as a followup PR if you want. The `SETTINGS_MAX_HEADER_LIST_SIZE` is not required to be strictly enforced. We account for the `32` bits of overhead on the `encoding` size, but not on the decode size. So we are currently following the \"be liberal in what you accept and conservative in what you send\" principle which I'm not yet convinced is a problem in this case.", "@Scottmitch \r\nSure, I will rebase once the PR is merged.\r\n\r\nAbout the 32 bytes overhead,\r\nI found that encoder will calculate the header size by add 32 bytes for each header. [Encoder](https://github.com/netty/netty/blob/4.1/codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java#L122)\r\nBut the decoder didn't add the 32 bytes in current 4.1 branch.\r\nOr maybe I missed that the 32 bytes is added in somewhere when encoding.", "> I found that encoder will calculate the header size by add 32 ... But the decoder didn't add the 32 bytes\r\n\r\nYip this reflects the current state. This is what I described in https://github.com/netty/netty/issues/6209#issuecomment-273365577.", "OK, sorry, I missed read that.... :(\r\nSo, the 32 bytes overhead could implement or not depends on need!\r\nThanks for your clarify!" ]

[]

[ "defect" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Decoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HuffmanEncoder.java" ]

[ "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Decoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/Encoder.java", "codec-http2/src/main/java/io/netty/handler/codec/http2/internal/hpack/HuffmanEncoder.java", "microbench/src/main/java/io/netty/microbench/http2/internal/hpack/DecoderULE128Benchmark.java" ]

[ "codec-http2/src/test/java/io/netty/handler/codec/http2/internal/hpack/DecoderTest.java", "codec-http2/src/test/java/io/netty/handler/codec/http2/internal/hpack/EncoderTest.java" ]