Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 336691

Summary: Possible wrong length returned by ChannelEndPoint.flush() in case of RandomAccessFileBuffer
Product: [RT] Jetty Reporter: Simone Bordet <simone.bordet>
Component: serverAssignee: Simone Bordet <simone.bordet>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: P3 CC: gregw, jetty-inbox
Version: 7.2.2   
Target Milestone: 7.3.x   
Hardware: PC   
OS: Linux   
Whiteboard:

Description Simone Bordet CLA 2011-02-09 05:16:51 EST 
The length reported is the length of the buffer being written, but since the write is non-blocking, it may be that less bytes are written.
Comment 1 Simone Bordet CLA 2011-02-09 05:32:50 EST 
Returning correct length.
Comment 2 Greg Wilkins CLA 2011-03-03 22:03:28 EST 
ChannelEndPoint appears to be double handling consumed length as it is both doing a header.skip(length); and a header.setGetIndex(bbuf0.position());
Comment 3 Greg Wilkins CLA 2011-03-03 22:05:17 EST 
ah this does not appear to be a new problem... but an old one uncovered by recent changes????
Comment 4 Greg Wilkins CLA 2011-03-04 00:03:28 EST 
It wasn't this issue... but it is fixed anyway.

The double setting of the buffer index has been removed. Only the skip is performed.