Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
2.0.20
-
None
-
None
Description
I'm updated from a MINA 2.0.7 to the 2.0.20 and am a user of CompressionFilter. Writing of Messages fails with a InvalidMarkException.
Reproducible Test is:
- Copy MINA Core Test org.apache.mina.core.service.AbstractIoServiceTest to MINA Compression Filter org.apache.mina.filter.compression Test Packages package.
- Adapt package statement to org.apache.mina.filter.compression.
- Add Compression to acceptor and connector
- acceptor.getFilterChain().addLast("compression", new CompressionFilter());
acceptor.getFilterChain().addLast("logger", new LoggingFilter());
acceptor.getFilterChain().addLast("codec",
new ProtocolCodecFilter(new TextLineCodecFactory(StandardCharsets.UTF_8))); - connector.getFilterChain().addLast("compression", new CompressionFilter());
connector.getFilterChain().addLast("logger", new LoggingFilter());
connector.getFilterChain().addLast("codec",
new ProtocolCodecFilter(new TextLineCodecFactory(StandardCharsets.UTF_8))); - Set a Breakpoint to java.nio.Buffer.reset() method, where the InvalidMarkException is thrown.
- Run Debug Testfile on org.apache.mina.filter.compression.AbstractIoServiceTest
After the Exception the session is immediatelly scheduled for disconnect.
It seems that there is a discrepancy between the mark() and reset() calls on the underlaying Buffer. In case of compression, a Buffer with the compressed content is created and is wrapped with the original Buffer in a FilteredWriteRequest, because CompressionFilter is a WriteRequestFilter. This is in WriteRequestFilter.filterWrite()
In DefaultIoFilterChain$HeadFilter.filterWrite() is then the mark() call, which is done on the compressed Buffer.
In AbstractPollingIoProcessor.writeBuffer() is the reset() call, which is in this case done on the original Buffer, leading to the Exception.
It seems that the change at date 16.02.2016
SHA-1: 44b58469f84ce991074cdc187b1c1f23b94cf445
- Don't try to reset a message when it's not a IoBuffer
which reassignes the buf before reset() is called broke it. The buf before reassign looks much better as the right to reset() in this case.
java.nio.InvalidMarkException
at java.nio.Buffer.reset(Buffer.java:306)
at org.apache.mina.core.buffer.AbstractIoBuffer.reset(AbstractIoBuffer.java:425)
at org.apache.mina.core.polling.AbstractPollingIoProcessor$Processor.writeBuffer(AbstractPollingIoProcessor.java:1131)
at org.apache.mina.core.polling.AbstractPollingIoProcessor$Processor.flushNow(AbstractPollingIoProcessor.java:994)
at org.apache.mina.core.polling.AbstractPollingIoProcessor$Processor.flush(AbstractPollingIoProcessor.java:921)
at org.apache.mina.core.polling.AbstractPollingIoProcessor$Processor.run(AbstractPollingIoProcessor.java:688)
at org.apache.mina.util.NamePreservingRunnable.run(NamePreservingRunnable.java:64)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)