I think in AjpNioProcessor the socketClosing flag is reversed. It is not
a functional problem, because reccyle() of AjpNioProcessor and of super
ignores it. But for consistency it should be switched. I'm not
committing straight away, because we are so close to tagging.
Mark: feel free to apply the patch if you think it's correct and harmless.
Otherwise I'll apply after the tag if there's no negative feedback until
then.
Index: java/org/apache/coyote/ajp/AjpNioProcessor.java
===================================================================
--- java/org/apache/coyote/ajp/AjpNioProcessor.java (revision 1154852)
+++ java/org/apache/coyote/ajp/AjpNioProcessor.java (working copy)
@@ -128,14 +128,14 @@
} catch (IOException e) {
error = true;
}
- recycle(true);
+ recycle(false);
continue;
} else if(type != Constants.JK_AJP13_FORWARD_REQUEST) {
// Usually the servlet didn't read the previous
request body
if(log.isDebugEnabled()) {
log.debug("Unexpected message: "+type);
}
- recycle(false);
+ recycle(true);
continue;
}
request.setStartTime(System.currentTimeMillis());
@@ -217,7 +217,7 @@
socket.setTimeout(keepAliveTimeout);
}
- recycle(true);
+ recycle(false);
}
rp.setStage(org.apache.coyote.Constants.STAGE_ENDED);
Regards,
Rainer
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
