| Re: [mdt-papyrus.dev] Broken test builds |
|
Hi all, The Extra-Tests build is now back to normal [1], except that there are 4 test failures (Related to the new SysML 1.4 tests) [1] 470367: [Releng] Extra-Tests builds fail on Hudson https://bugs.eclipse.org/bugs/show_bug.cgi?id=470367 Regards, Camille De : mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi all, The builds are mostly back to normal, with a few changes: -
Gerrit now builds everything in a single pass (Rather than 4 separate builds)
o
Pro: It is a little bit faster (~38 minutes vs ~42 minutes)
o
Con: It takes 25 minutes to resolve *all* dependencies before actually compiling anything.
This means that, in case of error, you’ll be notified a little bit later. In case of success, you’ll be notified a little bit earlier -
We now have a “Papyrus-Full” job which builds and tests everything in a single pass. This was a prerequisite for Sonar -
Tests-Failures is red again (Tycho Surefire 0.22/0.23 bug, it works fine with 0.21, so we may consider rolling back to Tycho 0.21) -
Extra-Tests is red for an unknown reason. To be investigated. Works fine locally for me, works fine in the Full job, but fails on Hudson when executed as a subset of the build -
Some extra parameters have been added to skip the tests (Required by the removal of Tycho 0.23 <<TestSuite>> and the new build layout. It may be possible to improve this to avoid using custom parameters) Additionally, the Gerrit plug-in configuration has been changed to skip older patch sets. If you push two changes on the same Gerrit contribution, the build for the first one will be automatically canceled (If
it is still running). This should reduce the load on the server, a little bit. That’s still not an excuse for pushing 5 patch sets in a row, but at least the consequences will be less severe :) Sorry for all the trouble and several days of red builds; I ran into several unexpected issues [1] and it took much longer than expected to get the new build(s) running... The work is not complete yet, but at
least you can start using Gerrit again. Please rebase any change on the latest master commit to get a proper Gerrit Validation. I will create the new maintenance branches tomorrow (RC4a and streams/1.1-maintenance, and maybe a specific Neon branch for the more disruptive changes, while master should remain ‘stable’ as much as possible) [1] 469953: [Releng] Update the build to support a global POM https://bugs.eclipse.org/bugs/show_bug.cgi?id=469953 PS: The Gerrit build is actually still running, so I can’t really *guarantee* that it works. But I’m still confident :) Regards, Camille De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi, Test builds use a private repository to avoid any conflict, and I cleaned the Extra-Tests workspace several times during the build refactoring. I’ve restarted the Main-Tests with a clean workspace as well, to
see if it starts failing, too (Build in progress) Locally, I’ve also wiped my m2 repository and both Main and Extra Tests still work Camille De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx
[mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LE FEVRE FRANCOIS Hello Perhaps it is related to :
http://dev.eclipse.org/mhonarc/lists/cross-project-issues-dev/msg11953.html https://bugs.eclipse.org/bugs/show_bug.cgi?id=467449 Did you clear all repository in hudson to be sure not to have two artifacts (group id, artifact id, version id) that holds the same jar. Francois De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi, Is anyone able to run the Extra Tests pom (from releng/extra-tests)? It requires the following parameters (If you’re not using the provided Launch Config): papyrus.repo.extra=https://hudson.eclipse.org/papyrus/job/Papyrus-Master-Extra/lastSuccessfulBuild/artifact/repository/ papyrus.repo.dev=https://hudson.eclipse.org/papyrus/job/Papyrus-Master-Developer/lastSuccessfulBuild/artifact/repository/ papyrus.repo.main=https://hudson.eclipse.org/papyrus/job/Papyrus-Master/lastSuccessfulBuild/artifact/repository/ It works for me in my local environment, but not on Hudson anymore: !SESSION Fri Jun 12 10:09:05 EDT 2015 ------------------------------------------ !ENTRY org.eclipse.equinox.launcher 4 0 2015-06-12 10:09:05.837 !MESSAGE Exception launching the Eclipse Platform: !STACK java.lang.SecurityException: SHA-256 digest error for org/eclipse/osgi/framework/log/FrameworkLogEntry.class at sun.security.util.ManifestEntryVerifier.verify(ManifestEntryVerifier.java:220) at java.util.jar.JarVerifier.processEntry(JarVerifier.java:241) at java.util.jar.JarVerifier.update(JarVerifier.java:228) at java.util.jar.JarVerifier$VerifierStream.read(JarVerifier.java:483) at sun.misc.Resource.getBytes(Resource.java:124) at java.net.URLClassLoader.defineClass(URLClassLoader.java:444) at java.net.URLClassLoader.access$100(URLClassLoader.java:71) at java.net.URLClassLoader$1.run(URLClassLoader.java:361) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:425) at java.lang.ClassLoader.loadClass(ClassLoader.java:358) at org.eclipse.core.runtime.adaptor.EclipseStarter.run(EclipseStarter.java:241) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:606) at org.eclipse.equinox.launcher.Main.invokeFramework(Main.java:669) at org.eclipse.equinox.launcher.Main.basicRun(Main.java:608) at org.eclipse.equinox.launcher.Main.run(Main.java:1515) at org.eclipse.equinox.launcher.Main.main(Main.java:1488) I don’t understand this error. Other test jobs work fine on Hudson, with a very similar configuration. All plug-ins are properly compiled, but the Eclipse Runtime can’t start, due to this exception. Any idea? Thanks, Camille De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi all, I’ve reported Bug 469953 [1] to track this. The main thing is that we now have different POMs, meaning that the Hudson jobs need to be reconfigured accordingly. This means that any commit which is not (re)based
on the refactoring commits will now fail on Gerrit or Master builds. This also means that Gerrit won’t work for Luna anymore (We might add a Luna-specific Gerrit job, but I’d like to avoid that) Feature builds will also need to be updated (or deleted if they are not used anymore, which is true for most of our feature builds) For example, jobs should build releng/main/pom.xml instead of releng/top-pom-main.xml [1] 469953: [Releng] Update the build to support a global POM https://bugs.eclipse.org/bugs/show_bug.cgi?id=469953 Camille De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi all, More precisions about this build refactoring. The idea is to provide a main pom (Which we currently don’t have) to build all components of Papyrus at the same time (Especially useful for Gerrit and Sonar) However, we made several bad choices in the current build implementation, which means that simply adding a top-level pom is not sufficient. An important build refactoring is required -
Top poms will be moved to a different location (Because we can’t have several Poms in the same folder) -
Top poms will be renamed (Because they all had the same artifact ID) -
Shared artifacts (Especially the test framework) will have to be built/deployed separately Additionally, some changes are required in the test framework, to be able to run tests for each test plug-in separately. We also need to distinguish plug-ins which provide helpers for the tests (diagram.tests,
junit.framework, ...), and actual test plug-ins These changes will be more disruptive than I expected, so i’d like to push them on both Mars maintenance and Neon, to keep builds and Hudson jobs as consistent as possible between branches. Feature branches and
existing Gerrit contributions affecting POMs/Releng will need to be updated accordingly. The mars-maintenance branch will be created when this refactoring is complete (Soon after, but we still need to clarify the next steps) Regards, Camille De :
mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de LETAVERNIER Camille Hi all,
I will fix this, along with a general build refactoring/improvement, but it might take a few hours Sorry for the inconvenience, Camille |