[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #16 from Mark Thomas ma...@apache.org --- Looks good to me. The failing tests pass. I haven't done a full test run though. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #8 from Mladen Turk mt...@apache.org --- Hmm, reading the code it seems that a socket is member of two pollsets (timeouts and connections). Is this correct. Note that socket can be member of only one pollset and 1.1.26 kind of enforces that. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #9 from Mark Thomas ma...@apache.org --- (In reply to comment #8) Hmm, reading the code it seems that a socket is member of two pollsets (timeouts and connections). Is this correct. Note that socket can be member of only one pollset and 1.1.26 kind of enforces that. Neither of those are pollsets. There are local lists of sockets. There are a handful of bugs I have found in the new AprEndpoint. Fixes are on the way. In the meantime, I would be helpful if you could look at the return value of Poll.pollset. The use of n++ twice inside the loop looks like a bug to me. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #10 from Mladen Turk mt...@apache.org --- Not sure what you mean by double increment. Is it? p-set[n++] = (jlong)(fd-rtnevents); p-set[n++] = P2J(fd-client_data); That should be fine. The returned array consist of pair of values odd being event and even being the socket. It was always like that. But you are right, there is a bug. Returned number should be n/2 not n. I'll fix that. If you create a shortcut in Java which calls .pollset(...) and divide by two in case the value is 0 doesn't that help? Anyhow, I'll fix the native and create test 1.1.27-dev binaries for further test. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 Mark Thomas ma...@apache.org changed: What|Removed |Added Component|Library |Connectors Version|unspecified |trunk Product|Tomcat Native |Tomcat 8 Target Milestone|--- | --- Comment #11 from Mark Thomas ma...@apache.org --- (In reply to comment #10) Not sure what you mean by double increment. Is it? p-set[n++] = (jlong)(fd-rtnevents); p-set[n++] = P2J(fd-client_data); That should be fine. The returned array consist of pair of values odd being event and even being the socket. It was always like that. But you are right, there is a bug. Returned number should be n/2 not n. I'll fix that. Thanks. If you create a shortcut in Java which calls .pollset(...) and divide by two in case the value is 0 doesn't that help? Already doing this locally. That was how I moved past this issue and found the others. Anyhow, I'll fix the native and create test 1.1.27-dev binaries for further test. Perfect. I see your commit, so I'm going to move this back to Tomact 8 to track the remaining AprEndpoint issies. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 Mark Thomas ma...@apache.org changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #12 from Mark Thomas ma...@apache.org --- The root cause of the initial issue is the tc native bug that Mladen has now fixed. This affected Tomcat 7 and 8. The other issues (the poller threads not stopping, socket reads hanging) are Tomcat 8 specific and have been fixed in trunk. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #13 from Mladen Turk mt...@apache.org --- Please try with the binaries from http://people.apache.org/~mturk/native/1.1.27-dev-20130202/ -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #14 from Mark Thomas ma...@apache.org --- Sure. Just as soon as you make those files world readable ;) -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #15 from Mladen Turk mt...@apache.org --- Attrs should be fine now. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #5 from Mark Thomas ma...@apache.org --- I can repeat this on 64-bit Windows. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 Mark Thomas ma...@apache.org changed: What|Removed |Added Component|Connectors |Library Version|trunk |unspecified Product|Tomcat 8|Tomcat Native Target Milestone||--- --- Comment #6 from Mark Thomas ma...@apache.org --- I believe I have found the root cause of this. There is a bug in tc native introduced in r1412919 in the implementation of Poll.pollset(). n is incremented twice inside the for each loop for each entry in the poller. That means the return value is 2 * number of entries rather than number of entries. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #7 from Mark Thomas ma...@apache.org --- Hmm. There may be more to this problem than just that. I still see issues even with a workaround for the Poll.poollset() return value problem. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #1 from Konstantin Kolinko knst.koli...@gmail.com --- Created attachment 29911 -- https://issues.apache.org/bugzilla/attachment.cgi?id=29911action=edit TEST-org.apache.catalina.loader.TestVirtualContext.APR.txt JUnit test result file -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #2 from Konstantin Kolinko knst.koli...@gmail.com --- Created attachment 29912 -- https://issues.apache.org/bugzilla/attachment.cgi?id=29912action=edit thread_dump_2013-02-01.txt Thread dump, taken with jvisualvm during the 15-minutes pause -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #3 from Konstantin Kolinko knst.koli...@gmail.com --- I suspect that wrapper is null here. Yes, it is null. I added the following code at line 1632: [[[ if (wrapper==null) throw new NullPointerException(wrapper is null, n: + n + , rv: +rv+, desc[n*2+1]: + desc[n*2+1]); ]]] It results in the following on console: [junit] WARNING: Unexpected poller error [junit] java.lang.NullPointerException: wrapper is null, n: 0, rv: 1, desc[n*2+1]: 52898296 [junit] at org.apache.tomcat.util.net.AprEndpoint$Poller.run(AprEndpoint.java:1632) [junit] at java.lang.Thread.run(Thread.java:722) -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
[Bug 54513] NullPointerException in AprEndpoint$Poller.run
https://issues.apache.org/bugzilla/show_bug.cgi?id=54513 --- Comment #4 from Konstantin Kolinko knst.koli...@gmail.com --- Reproduced this issue with Native 1.1.24, but a bit different. I was not able to reproduce the NPE, but SocketInputStream.socketRead0 in the test hung in the same way. There were two variants: a) no error message is printed b) Feb 01, 2013 H:MM:SS AM org.apache.tomcat.util.net.AprEndpoint$Poller run WARNING: Error processing socket timeout There were a pair of runs with 1.1.26 and with 1.1.24 when the issue did not reproduce. thread_dump_2013-02-01.txt (14.90 KB, text/plain) What is odd in the thread dump is that there are two threads: http-apr-127.0.0.1-auto-2-Poller http-apr-127.0.0.1-auto-1-Poller All other threads besides this poller one are auto-2. It is the second test, but the poller thread from the first test has not stopped. -- You are receiving this mail because: You are the assignee for the bug. - To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
