Author: markt
Date: Wed Jun 24 19:45:58 2009
New Revision: 788158

URL: http://svn.apache.org/viewvc?rev=788158&view=rev
Log:
Withdraw the AccessLogValve patches. Updated patch to follow.

Modified:
    tomcat/current/tc5.5.x/STATUS.txt

Modified: tomcat/current/tc5.5.x/STATUS.txt
URL: 
http://svn.apache.org/viewvc/tomcat/current/tc5.5.x/STATUS.txt?rev=788158&r1=788157&r2=788158&view=diff
==============================================================================
--- tomcat/current/tc5.5.x/STATUS.txt (original)
+++ tomcat/current/tc5.5.x/STATUS.txt Wed Jun 24 19:45:58 2009
@@ -39,32 +39,6 @@
   +1: markt, kkolinko, fhanik
   -1:
 
-* Fix https://issues.apache.org/bugzilla/show_bug.cgi?id=38483
-  Make access log valves thread safe
-  http://people.apache.org/~markt/patches/2009-02-04-bug38483.patch
-  +1: markt, rjung
-  +1: kkolinko (ok, but I propose an additional patch below)
-  -1: fhanik - sounds like a use case for ThreadLocal, synchronized this on a 
valve is bad
-      markt - This is post response having been sent to the client so the 
syncs are less of
-              an issue. Further, the micro-benchmarks suggests that the syncs 
are a little
-              quicker.
-
-* Fix https://issues.apache.org/bugzilla/show_bug.cgi?id=38483
-  Make access log valves thread safe (part 2)
-  (It is an addition to markt's 2009-02-04-bug38483.patch that is proposed 
above)
-  The dateFormatter that is used for log file rotation also needs its syncs.
-  Also added volatile modifier to rotationLastChecked, because
-  long assignments are not atomic. (see Java Language Specification 3rd ed, 
ch. 17.7)
-  http://people.apache.org/~kkolinko/patches/2009-06-07_bug38483_2.patch
-  +1: kkolinko
-  -1: fhanik - I don't see how this is more thread safe than before. 
-               A thread can still call writer.println(message) while another 
thread has called close() but not yet open
-      markt - Strictly, the sync is required to protect access to the 
fileDateFormatter object
-              Based on the microbenchmarks for the other syncs, I suspect a 
sync here will be
-              faster than a ThreadLocal.
-              The patch appears to use tabs rather than spaces
-              The issue Filip identified with writing messages when the file 
is being rotated is still an issue
-
 * Fix use of setLoopbackMode() for tribes multicast.
   in 
container/modules/cluster/src/share/org/apache/catalina/cluster/mcast/McastServiceImpl.java
   and 
container/modules/groupcom/src/share/org/apache/catalina/tribes/membership/McastServiceImpl.java



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

Reply via email to