Ensure we will call read() if nothing was decoded and AUTO_READ is no… #88
Conversation
…t used Motivation: We should ensure we call `ctx.read()` if the codec received a message to decode but did not produce anything and AUTO_READ is set to false. Failing to do so might result in stales as the user will not know that more reads are needed to make progress. This is inline with what ByteToMessageDecoder does under the hood Modifications: - Change OHttpClientCodec and OHttpServerCodec to issue reads to make progress - Add unit test Result: Ensure reads are submitted when needed and so no stales happen
There was a problem hiding this comment.
Change looks good, except the missing piece i pointed out in the comment.
I am wondering if this logic should be moved to MessageToMessageDecoder and MessageToMessageCodec (which could just override channelReadComplete() and call decoder.channelReadComplete()). Is there a reason why ByteToMessageDecoder is doing it and not MessageToMessageDecoder? Is it because the former is using a cumulation buffer?
codec-ohttp/src/main/java/io/netty/incubator/codec/ohttp/OHttpClientCodec.java
Show resolved
Hide resolved
The problem is that people might have extended these and marked them '@sharable'. When adding the same instance it will not work correctly |
codec-ohttp/src/main/java/io/netty/incubator/codec/ohttp/OHttpServerCodec.java
Outdated
Show resolved
Hide resolved
…t used
Motivation:
We should ensure we call
ctx.read()if the codec received a message to decode but did not produce anything and AUTO_READ is set to false. Failing to do so might result in stales as the user will not know that more reads are needed to make progress. This is inline with what ByteToMessageDecoder does under the hoodModifications:
Result:
Ensure reads are submitted when needed and so no stales happen