Displaying #maven-dev/2017-03-02.log:

Thu Mar 2 01:05:52 2017  theRealGent:Joined the channel
Thu Mar 2 03:44:40 2017  theRealGent:Joined the channel
Thu Mar 2 07:54:14 2017  tibor_:Joined the channel
Thu Mar 2 09:55:28 2017  olamy:Joined the channel
Thu Mar 2 17:07:52 2017  tibor_:Joined the channel
Thu Mar 2 17:20:36 2017  tibor_:Joined the channel
Thu Mar 2 18:51:25 2017  hboutemy:Joined the channel
Thu Mar 2 19:18:59 2017  rfscholte:Joined the channel
Thu Mar 2 21:00:21 2017  Michael-O:Joined the channel
Thu Mar 2 21:00:25 2017  Michael-O:hi tibor_
Thu Mar 2 21:00:31 2017  Michael-O:Did you read my mail?
Thu Mar 2 21:00:40 2017  tibor_:yes
Thu Mar 2 21:00:51 2017  tibor_:I can't believe it :)
Thu Mar 2 21:01:28 2017  Michael-O:that it worked?
Thu Mar 2 21:01:33 2017  tibor_:i pushed some changes to 2.19.2-experimental and 0.9.x branch too
Thu Mar 2 21:01:39 2017  Michael-O:today?
Thu Mar 2 21:01:40 2017  tibor_:yes
Thu Mar 2 21:01:46 2017  tibor_:that it did not hang
Thu Mar 2 21:01:59 2017  tibor_:yesterday
Thu Mar 2 21:02:30 2017  Michael-O:let me update
Thu Mar 2 21:03:04 2017  tibor_:I found something like JarOoutputStream entry was not closed and stream has new flush method in java7
Thu Mar 2 21:03:32 2017  tibor_:I have a phonecall..
Thu Mar 2 21:08:00 2017  Michael-O:Joined the channel
Thu Mar 2 21:08:08 2017  Michael-O:tibor_: running tests
Thu Mar 2 21:08:29 2017  tibor_:thx
Thu Mar 2 21:10:21 2017  Michael-O:side note: I ran the branch on all ITs. The tests freeze repeatedly after org.apache.maven.surefire.its.ClasspathScopeFilteringIT
Thu Mar 2 21:14:51 2017  Michael-O:tests finished
Thu Mar 2 21:16:01 2017  Michael-O:tibor_: updated tarball uploaded
Thu Mar 2 21:42:11 2017  tibor_:Michael-O: It has still the same problem. In principle forked jvm is not different from the sample. Can you please open ForkBooter.java and you will se main() method. Let's comment it's body completely and run mvn install. The tests will fail but the jvm should return and CommandLineUtils should not hang. This way we can select which part of the method is doing it.
Thu Mar 2 21:54:23 2017  Michael-O:just a second
Thu Mar 2 21:58:03 2017  Michael-O:tibor_: code changed, tests running
Thu Mar 2 21:58:23 2017  tibor_:I don't know how the OS is handling two subprocessing running in parallel but according to the logs it is obious that when both start via ProcessBuilder.start, both hang util one kills itself and then another starts.
Thu Mar 2 21:59:00 2017  tibor_:always those tests fail which use two processes
Thu Mar 2 21:59:30 2017  tibor_:Maybe I would be able to simulate this scenarion in our small sample code.
Thu Mar 2 22:02:16 2017  Michael-O:but that should be async as long as we don't do waitFor
Thu Mar 2 22:02:29 2017  Michael-O:or you have to consume the streams
Thu Mar 2 22:04:31 2017  tibor_:Yes, ForkStarter has ThreadPool and both forks are independently started in two Threads.
Thu Mar 2 22:04:57 2017  tibor_:bewfore doing waitFore yes both should be async
Thu Mar 2 22:08:20 2017  Michael-O:tests completed
Thu Mar 2 22:08:23 2017  Michael-O:tarball updated
Thu Mar 2 22:08:24 2017  Michael-O:have a look
Thu Mar 2 22:21:20 2017  tibor_:Thx, I see.
Thu Mar 2 22:21:47 2017  tibor_:Now both processes are nicely concurrent. The methods .start() and waitFor() as well.
Thu Mar 2 22:22:52 2017  tibor_:So the ForkedBooter must cause this.
Thu Mar 2 22:23:43 2017  Michael-O:That the process#start happen in the threadpool task or the ForkedBooter itself?
Thu Mar 2 22:23:57 2017  Michael-O:I guess the start()ing thread is blocked for reading
Thu Mar 2 22:26:52 2017  tibor_:process#start in ForkStarter located in Maven process
Thu Mar 2 22:27:07 2017  tibor_:ForkedBooter in located in forked jvm
Thu Mar 2 22:27:59 2017  Michael-O:I will leave now for some rest. Just drop me an email if you need further testing.
Thu Mar 2 22:28:01 2017  Michael-O:Good night
Thu Mar 2 22:28:17 2017  tibor_:Good night.

Comments