Hey guys, We really do not properly read the e-mails. All the links you sent me was from master branch.
We are NOT working in master branch. Do not test master yet. Our focus is the branch named "SUREFIRE-1489". Please run only the build on branch "SUREFIRE-1489" and nothing else yet. We should proceed step by step. Please try to be patient and: $ git clone .... $ git fetch $ git checkout SUREFIRE-1489 $ mvn clean install -nsu -P run-its -Djdk.home=/path/to/jdk10 -Djacoco.skip=true I will be waiting. All my previous emails about ASF Jenkins was from the branch "SUREFIRE-1489" and not the master. With Best Regards Tibor On Sun, Feb 25, 2018 at 1:20 PM, Olivier Lamy <ol...@apache.org> wrote: > something very interesting on my side..... > It works with empty on ASF Jenkins: https://builds. > apache.org/view/M-R/view/Maven/job/maven-surefire-fs/ > But not in this Jenkins https://jenkins.webtide.net/job/sandbox/job/ > surefire-master-jdk8/5/console or using Travis https://travis-ci.org/ > olamy/maven-surefire/builds/345892147 > Sorry I cannot test locally with an empty repo as my internet is very slow > today. > but anyway we should not have a build with a need of an empty repo... > Funny as ironically the Maven team cannot have a stable build working > everywhere :-) > > On 25 February 2018 at 19:45, Robert Scholte <rfscho...@apache.org> wrote: > >> Hi all, >> >> I'm on *master* and simply running 'mvn clean verify' and the unittests >> of Surefire Bootstrap fail. >> I get 28 errors, which can be brought back into 2 exceptions: >> - java.lang.NoClassDefFoundError: Could not initialize class >> org.apache.maven.surefire.booter.SystemUtils >> - java.lang.NoClassDefFoundError: org/apache/commons/lang3/SystemUtils >> Caused by: java.lang.ClassNotFoundException: org.apache.commons.lang3. >> JavaVersion >> >> I've removed my local repository, same result. >> >> Robert >> >> On Sun, 25 Feb 2018 04:04:02 +0100, Tibor Digana <tibordig...@apache.org> >> wrote: >> >> Hello Olivier, Robert, >> >> I think this issue you saw was caused by the old hack with commons-lang3 >> in module surefire-booter/pom.xml. >> I reproduced it on WindowsXP but not on Windows7, funny. >> I removed the change in the pom.xml and committed (git push force) to our >> branch SUREFIRE-1489. Now the branch SUREFIRE-1489 contains changes from >> three JIRAs, but that's ok, we will create new ones according to Jiras. >> I run this command: >> >> mvn clean install -P run-its -Dintegration-test-port=8000 >> -Dintegration-test-stop-port=8001 "-Djdk.home=d:\Program Files\Java\jdk10" >> -Djacoco.skip=true >> >> I am overriding the branch, so it is better to clone it fresh. >> >> Please let me know if it is fine on your side. >> >> With Best Regards, >> Tibor >> >> >> >> >> On Sun, Feb 25, 2018 at 2:51 AM, Tibor Digana <tibordig...@apache.org> >> wrote: >> >>> Try to delete local maven repository on Travis and run the branch >>> SUREFIRE-1489 with --no-snapshot-updates. >>> I hope we will have good results. >>> Good night! >>> >>> On Sun, Feb 25, 2018 at 2:34 AM, Tibor Digana <tibordig...@apache.org> >>> wrote: >>> >>>> What branch is for the build /maven-surefire-fs? >>>> Any special build configuration or sources are different from master? >>>> >>>> On Sun, Feb 25, 2018 at 2:28 AM, Olivier Lamy <ol...@apache.org> wrote: >>>> >>>>> Did you read my emails? The thread with subject " [3/3] maven-surefire >>>>> git >>>>> commit: Revert "[SUREFIRE-1473] upgrade commons-lang3 to avoid NPE on >>>>> jdk10" " >>>>> I assume no so here few links >>>>> https://builds.apache.org/view/M-R/view/Maven/job/maven- >>>>> surefire-fs/2/console >>>>> https://jenkins.webtide.net/job/sandbox/job/surefire- >>>>> master-jdk8/4/console >>>>> https://travis-ci.org/olamy/maven-surefire/builds/345611495 >>>>> >>>>> There are all Jenkins or Travis builds with an empty maven repo. >>>>> >>>>> >>>>> >>>>> On 25 February 2018 at 11:11, Tibor Digana <tibordig...@apache.org> >>>>> wrote: >>>>> >>>>> > Hi Olivier, >>>>> > >>>>> > I am trying to be friendly and helpful. >>>>> > >>>>> > Maybe a hint: >>>>> > One problem I found weeks ago was in Maven Central which contained a >>>>> > SNAPSHOT version built with Java 1.7. >>>>> > I had to do this on my PC: >>>>> > $ mvn >>>>> > org.codehaus.mojo:build-helper-maven-plugin:3.0.0: >>>>> remove-project-artifact >>>>> > -Dbuildhelper.failOnError=false >>>>> > $ mvn install *-nsu* >>>>> > and try with JAVA_HOME=JDK 1.8 for first time and then with >>>>> jdk.home=... >>>>> > with other JDKs. >>>>> > >>>>> > I remember your mail and the stacktrace with commons-lang3-3.5.jar >>>>> but I >>>>> > could not reproduce identical symptoms on my PC and Jenkins lately. >>>>> > Therefore I have opened the master log [1] and I do not see that >>>>> stacktrace >>>>> > you had before. >>>>> > I am building locally and on Jenkins almost every day but the issue >>>>> with >>>>> > calling commons-lang3 has not occurred, sorry. >>>>> > >>>>> >>>>> See above. Maybe it's "it fail on my machine" or "it works on my >>>>> machine" >>>>> BUT I have produced a lot of evidences there is a problem so please do >>>>> not >>>>> ignore that >>>>> >>>>> >>>>> > >>>>> > I would like to help you but I do not know what's going on your PC. >>>>> Maybe >>>>> > on Sunday we should talk on IRC, I am open for that. >>>>> >>>>> >>>>> > Not everyone can do whatever. Everybody has to open a branch with a >>>>> name of >>>>> > Jira, kindly ask the developers to talk about long term concept, to >>>>> make a >>>>> > code-review and wait for a response. We are not always available and >>>>> > therefore we have to wait for someone else and sometimes we have to >>>>> find a >>>>> > person in commit list, go to the developer, track her/him on e-mail >>>>> and >>>>> > IRC. >>>>> > But I would prefer IRC during the day or evening. >>>>> > >>>>> >>>>> I don't think irc is a good place for decisions as it's limited number >>>>> of >>>>> participants. >>>>> I see our project as a community and not as a limited number of people >>>>> who >>>>> can be on irc at a certain time. >>>>> Mailing list are the basis of how Apache projects works because you >>>>> have >>>>> history (you can search), async response mode etc... >>>>> >>>>> Anyway I asked here fro having master jdk7 minimum with bumping >>>>> version to >>>>> 3.0.0. >>>>> we can have a maintenance branch 2.x >>>>> But let's move forward before jdk7 get deprecated (oh oops it's >>>>> already the >>>>> case). >>>>> Joke apart. I really want a release supporting jdk10 at least >>>>> something not >>>>> failing with CNFE.... >>>>> >>>>> >>>>> > >>>>> > [1]: >>>>> > https://builds.apache.org/job/maven-wip/job/maven-surefire/ >>>>> > job/master/20/consoleFull >>>>> > >>>>> > Cheers >>>>> > Tibor >>>>> > >>>>> > >>>>> > >>>>> > >>>>> > On Sun, Feb 25, 2018 at 12:47 AM, Olivier Lamy <ol...@apache.org> >>>>> wrote: >>>>> > >>>>> > > Please Tibor read the emails and I'm so sorry to repeat again >>>>> myself but >>>>> > > it looks you do not listen anyone... >>>>> > > MASTER DOESN'T BUILD!!! >>>>> > > using jdk7 or jdk8 or jdk9 or jdk10 >>>>> > > I send few links with failure and as far as I can understand from >>>>> the >>>>> > > thread neither Robert can build it!! >>>>> > > We must first fix that!!! >>>>> > > >>>>> > > >>>>> > > On 25 February 2018 at 09:17, Tibor Digana <tibordig...@apache.org >>>>> > >>>>> > wrote: >>>>> > > >>>>> > >> Sorry, I have to repeat myself because of undelivered message(s) >>>>> to >>>>> > Maven >>>>> > >> mailing list. >>>>> > >> >>>>> > >> For clarifying the Surefire goes with Java 6. If you want to run >>>>> your >>>>> > >> tests with JDK10 or lower you have to specify the system property >>>>> > >> -Djdk.home=/path/to/jdk10 and the you will see that the tests >>>>> pass. Go >>>>> > to >>>>> > >> target/surefire-reports and failsafe-reports and target/it/** and >>>>> check >>>>> > the >>>>> > >> "java.home" and you will see your /path/to/jdk10. >>>>> > >> >>>>> > > >>>>> > > >>>>> > > I KNOW HOW IT WORKS AND AGAIN MY POINT IS whenever jdk you use >>>>> MASTER >>>>> > > DOESN'T BUILD!! >>>>> > > >>>>> > > >>>>> > >> One more remark I got the Jenkins working [1]. The whole problem >>>>> was >>>>> > with >>>>> > >> the call withMaven() in Jenkinsfile. After I switched back to >>>>> withEnv() >>>>> > the >>>>> > >> logs, InterruptedException has gone! >>>>> > >> I reported several improvements that Olivier did not like to >>>>> improve but >>>>> > >> I think there is no any significant showstopper for 2.21.0. >>>>> > >> >>>>> > > >>>>> > > I have no idea what you're talking about? Do you have any email >>>>> threads >>>>> > or >>>>> > > jira where I say that? >>>>> > > >>>>> > > >>>>> > >> I will now talk about the Jiras improvements with Olivier in >>>>> GitHub and >>>>> > I >>>>> > >> hope we will make the release soon! >>>>> > >> >>>>> > >> I still do not understand what codeline you run and why you have >>>>> > problems >>>>> > >> with runtime if I did not have several months and I tested with >>>>> JDK8 >>>>> > u162, >>>>> > >> u132, u121. >>>>> > >> Please post a link to your codeline because I highly doubt that >>>>> you are >>>>> > >> testing the identical code with me! >>>>> > >> >>>>> > > >>>>> > > PLEASE READ ALL THE PREVIOUS EMAIL I SENT WITH BUILD FAILURE ON >>>>> PUBLIC >>>>> > > JENKINS >>>>> > > >>>>> > > >>>>> > >> >>>>> > >> [1]: https://builds.apache.org/job/maven-wip/job/maven-surefire/j >>>>> > >> ob/SUREFIRE-1489/ >>>>> > >> >>>>> > >> For better clarification 2.21.0 is using modulepath - we have >>>>> > integration >>>>> > >> test for Jigsaw modularity and modulepath is activated for JDK 9+. >>>>> > >> NPE was fixed in 2.21.0 - check the release notes. >>>>> > >> We are no more calling the problematic method from commons-lang3 >>>>> which >>>>> > >> caused NPE. This does not require any upgrade of commons-lang3 >>>>> till >>>>> > 3.0.0. >>>>> > >> Please follow the build [1]. >>>>> > >> >>>>> > >> Cheers >>>>> > >> Tibor >>>>> > >> >>>>> > >> >>>>> > >> On Sat, Feb 24, 2018 at 3:22 PM, Robert Scholte < >>>>> rfscho...@apache.org> >>>>> > >> wrote: >>>>> > >> >>>>> > >>> Hi, >>>>> > >>> >>>>> > >>> I've noticed some unsatisfying discussions regarding surefire, >>>>> so let >>>>> > me >>>>> > >>> try to help moving this forward. >>>>> > >>> >>>>> > >> >>>>> > > I'm just not happy about how people do not listen neither read >>>>> emails.... >>>>> > > Emails is the primary way of working at Apache because of different >>>>> > > timezone etc... >>>>> > > IRC is just a volatile dicussion channel when you disconnect you >>>>> cannot >>>>> > > read history, later reply from someone (perso I often shutdown my >>>>> laptop) >>>>> > > >>>>> > > >>>>> > >> >>>>> > >>> Here's the situation as I see it: >>>>> > >>> - Surefire 2.20.1 is the latest released version (2017-09-15) >>>>> > >>> >>>>> > >>> - 2.20.1 doesn't use the modulepath yet, that will be part of >>>>> the next >>>>> > >>> release. A lot of people are waiting for this even before the >>>>> official >>>>> > >>> release of JDK9. >>>>> > >>> >>>>> > >>> - Surefire 2.x is is still JDK 1.6 compatible, the requirement >>>>> of JDK >>>>> > >>> 1.7 is planned for surefire 3.0.0 >>>>> > >>> >>>>> > >> >>>>> > > I'm fine with that so we can have a branch 2.x with jdk6 compatible >>>>> > > And master 3.0.0 with jdk7 requirement. >>>>> > > Does it sound reasonnable? >>>>> > > >>>>> > > >>>>> > >> >>>>> > >>> - 2.20.1 doesn't run on JDK10, it gets an NPE caused by >>>>> commons-lang >>>>> > >>> when trying to get the Java version. The version of commons-lang >>>>> where >>>>> > this >>>>> > >>> is fixed required JDK 1.7 (which clashes with the previous >>>>> bullet) >>>>> > >>> >>>>> > >>> - After the next version (2.21), surefire will focus on 3.0.0 >>>>> > >>> >>>>> > >>> - There's no agreement that the current master works with JDK10. >>>>> One of >>>>> > >>> the things I noticed is that you need to run surefire with a >>>>> special >>>>> > set of >>>>> > >>> arguments. I haven't been able to build it locally, so I cannot >>>>> judge >>>>> > it. >>>>> > >>> >>>>> > >>> - Surefire is a huge, complex project and Tibor has taken the >>>>> > >>> responsibility for maintaining this project. >>>>> > >>> >>>>> > >> >>>>> > > I do not see such role at Apache. everyone can work on the project >>>>> with >>>>> > > the same right. >>>>> > > There is no BDFL rule and NO one can revert commits from others >>>>> without >>>>> > > discussion. >>>>> > > I guess that's the problem here I keep sending emails I have no >>>>> > response!! >>>>> > > >>>>> > > >>>>> > >> >>>>> > >>> What I would like to see is: >>>>> > >>> - What's blocking a release right now? Is it just CI? Knowing >>>>> that CI >>>>> > >>> has changed a lot the last couple of months, did it just expose >>>>> > *unknown >>>>> > >>> issues* compared to 2.20.1? >>>>> > >>> - Is the master at its current state already an improvement >>>>> compared to >>>>> > >>> 2.20.1? >>>>> > >>> - What would be the estimated release date for surefire. >>>>> > >>> - If this date is too unsure or takes too long, can Olivier >>>>> create a >>>>> > >>> branch and do a 21.1-JDK10 release? or beta1? >>>>> > >>> >>>>> > >>> My main issue is : we should work towards a new release, because >>>>> 2.20.1 >>>>> > >>> doesn't work with JDK10 and there's no workaround. >>>>> > >>> IMHO this doesn't have to be an official surefire release, we >>>>> can add >>>>> > >>> some special qualifier to at least help those who need to work >>>>> with >>>>> > JDK10. >>>>> > >>> >>>>> > >>> thanks, >>>>> > >>> Robert >>>>> > >>> >>>>> > >> >>>>> > >> >>>>> > > >>>>> > > >>>>> > > -- >>>>> > > Olivier Lamy >>>>> > > http://twitter.com/olamy | http://linkedin.com/in/olamy >>>>> > > >>>>> > >>>>> >>>>> >>>>> >>>>> -- >>>>> Olivier Lamy >>>>> http://twitter.com/olamy | http://linkedin.com/in/olamy >>>>> >>>> >>>> >>> >> >> >> >> > > > -- > Olivier Lamy > http://twitter.com/olamy | http://linkedin.com/in/olamy > -- > Olivier Lamy > http://twitter.com/olamy | http://linkedin.com/in/olamy >