Displaying #maven-dev/2017-03-16.log:

Thu Mar 16 02:00:23 2017  jasonvanzyl:Joined the channel
Thu Mar 16 09:17:59 2017  Michael-O:Joined the channel
Thu Mar 16 10:12:45 2017  olamy:Joined the channel
Thu Mar 16 11:38:11 2017  jasonvanzyl:Joined the channel
Thu Mar 16 14:17:19 2017  Michael-O:Joined the channel
Thu Mar 16 16:04:31 2017  Michael-O:Joined the channel
Thu Mar 16 17:24:35 2017  jasonvanzyl:Joined the channel
Thu Mar 16 17:30:06 2017  rfscholte:Joined the channel
Thu Mar 16 19:59:23 2017  jasonvanzyl:Joined the channel
Thu Mar 16 20:00:50 2017  tibor_:Joined the channel
Thu Mar 16 20:31:38 2017  Michael-O:Joined the channel
Thu Mar 16 20:32:25 2017  Michael-O:tibor_, I have updated both machines. No avail. Extracted cmd args for isolated test ForkModeIT forCountTwoNoReuse
Thu Mar 16 20:32:43 2017  Michael-O:-DforkCount=1 works, as soon as I increase this value it fails
Thu Mar 16 20:40:03 2017  tibor_:Yes because 1 and 2 is different. 1: testset is not communicated with master process - test set is in props file. 2: communication with master process.
Thu Mar 16 20:41:24 2017  tibor_:What happens if you put a delay 60 seconds and -DforkCount=1?
Thu Mar 16 20:59:46 2017  Michael-O:Joined the channel
Thu Mar 16 21:06:37 2017  Michael-O:You mean -DsleepLength=60000, right?
Thu Mar 16 21:07:53 2017  Michael-O:Passes
Thu Mar 16 21:11:33 2017  tibor_:Michael-O: That's however different from my expectation, but ok.
Thu Mar 16 21:11:55 2017  tibor_:I thought I would fail with 60000 millis.
Thu Mar 16 21:12:11 2017  Michael-O:forkCount=1 and 60000 is fine
Thu Mar 16 21:12:14 2017  Michael-O:Why should it fail?
Thu Mar 16 21:13:45 2017  tibor_:I guess if communication with master is not available and the delay is longer than 2x10 sec the PING timeout elapsed and jvm was killed.
Thu Mar 16 21:14:33 2017  tibor_:What branch did you test here?
Thu Mar 16 21:14:38 2017  tibor_:master?
Thu Mar 16 21:14:46 2017  Michael-O:2.19.2-experimental
Thu Mar 16 21:15:59 2017  tibor_:The new branch we implemented ACK fixed race condition. How long the faulty test scenario took?
Thu Mar 16 21:16:54 2017  Michael-O:Let me check
Thu Mar 16 21:27:12 2017  Michael-O:Just ran with ACK
Thu Mar 16 21:27:21 2017  Michael-O:Unfortunately, it does not contain our debug logs
Thu Mar 16 21:27:33 2017  Michael-O:But it fails too, I will bump sleepLength
Thu Mar 16 21:31:26 2017  Michael-O:you won't believe it, it passes with 60000 ms
Thu Mar 16 21:32:03 2017  tibor_:You have checked log.txt in target?
Thu Mar 16 21:32:35 2017  Michael-O:I am running the test directly, copied the cmd lien from failsafe-reports
Thu Mar 16 21:32:45 2017  tibor_:The IT runs usually few seconds only.
Thu Mar 16 21:33:19 2017  tibor_:If it shows you in log.txt time execution close to 20 sec then it is bad.
Thu Mar 16 21:33:55 2017  Michael-O:Iam retrying, it is likely flaky
Thu Mar 16 21:34:46 2017  tibor_:As you said passed with 60000 and forkCount 2?
Thu Mar 16 21:35:19 2017  tibor_:or forkCount 1?
Thu Mar 16 21:35:21 2017  Michael-O:2
Thu Mar 16 21:35:26 2017  tibor_:hm
Thu Mar 16 21:36:08 2017  tibor_:It looks like many ITs crash it.
Thu Mar 16 21:36:20 2017  tibor_:This runs in embedded mode?
Thu Mar 16 21:38:07 2017  Michael-O:no, I cd'ed into target/ForkMode_testForkCountTwoNoReuse and issuesd the command as per Failsafe
Thu Mar 16 21:39:35 2017  tibor_:I know you run from command line that's fine, but I meant the ForkModeIT. The IT runs in embedded mode. WDYT?
Thu Mar 16 21:40:48 2017  Michael-O:What do yo mean by embedded mode?
Thu Mar 16 21:50:59 2017  tibor_:I mean that the Verifier runs the IT in target from current Maven build. The MavenLauncher.
Thu Mar 16 21:54:50 2017  tibor_: @Test public void testForkCountTwoNoReuse()
Thu Mar 16 21:55:03 2017  tibor_:Is it this test which fails?
Thu Mar 16 21:55:52 2017  tibor_:The previous test methods call .setForkJvm(). But testForkCountTwoNoReuse() does not.
Thu Mar 16 22:05:43 2017  Michael-O:Yes, MavenLauncher launches the test in embedded mode.
Thu Mar 16 22:05:49 2017  Michael-O: @Test public void testForkCountTwoNoReuse(): correct
Thu Mar 16 22:07:58 2017  tibor_:Let it call .setForkJvm(). It shoul be fine. All test methods should have it.
Thu Mar 16 22:08:05 2017  tibor_:should
Thu Mar 16 22:09:20 2017  Michael-O:How from the cmd line?
Thu Mar 16 22:09:35 2017  tibor_:not from cmd
Thu Mar 16 22:09:59 2017  tibor_:the cmd is ok as you said, or?
Thu Mar 16 22:13:56 2017  Michael-O:It was once only
Thu Mar 16 22:14:04 2017  Michael-O:All subsequent iterations failed
Thu Mar 16 22:14:20 2017  hboutemy:Joined the channel
Thu Mar 16 23:05:46 2017  Michael-O:tibor_, http://stackoverflow.com/a/39720983/696632
Thu Mar 16 23:07:01 2017  tibor_:I am reading..
Thu Mar 16 23:11:53 2017  tibor_:What means "subshell"? No forked process? Embedded in current process?
Thu Mar 16 23:13:43 2017  Michael-O:It is a separete process but fork() != execve()
Thu Mar 16 23:13:55 2017  Michael-O:https://www.freebsd.org/cgi/man.cgi?query=execve+&sourceid=opensearch
Thu Mar 16 23:14:04 2017  Michael-O:https://www.freebsd.org/cgi/man.cgi?query=fork&apropos=0&sektion=0&manpath=FreeBSD+11.0-RELEASE+and+Ports&arch=default&format=html
Thu Mar 16 23:25:26 2017  Michael-O:stdio seems to very compliated
Thu Mar 16 23:36:16 2017  jasonvanzyl:Joined the channel
Thu Mar 16 23:42:56 2017  Michael-O:tibor_, I just made another test with freebsd-issue, I think that /bin/sh -c is absolutely redundant
Thu Mar 16 23:43:03 2017  Michael-O:It is just overhead
Thu Mar 16 23:52:23 2017  Michael-O:tibor_, this is every interesting: http://stackoverflow.com/a/38045043/696632
Thu Mar 16 23:52:50 2017  Michael-O:Especially two projects referenced: https://github.com/brettwooldridge/NuProcess and https://github.com/zeroturnaround/zt-exec
Thu Mar 16 23:53:27 2017  Michael-O:No pumper branfuck
Thu Mar 16 23:53:31 2017  Michael-O:brainfuck*

Comments