Closed pshipton closed 5 years ago
@andrew-m-leonard I'll restart it just in case. https://ci.eclipse.org/openj9/job/Build_JDKnext_x86-64_windows_OpenJDK/64/
OpenJ9: 896a36e OMR: c30a477 OpenJDKnext: 4127695 - openj9-openjdk-jdk
previous build OpenJ9: 9514041 OMR: 7e9584e OpenJDKnext: da32c68 - openj9-openjdk-jdk
https://github.com/ibmruntimes/openj9-openjdk-jdk/compare/da32c68...4127695
The restarted job failed the same way.
The auto-merge conflicts for jdkNext were resolved yesterday, hence quite a lot of backlog'd changes would have been pulled in. It's at jdk-14+16 now.
@andrew-m-leonard ok, but Windows doesn't build any more, so we can't promote openj9-staging. It's not clear from your comment, are you looking at the problem?
@pshipton Adam Thorpe in my team is looking at this...
These test builds are using JDK12 to build next (14), which is causing at least this issue?
Looking at other acceptance jobs, they are all using JDK12. My understanding is that you cannot build JDK 14 with anything below 13
The java.desktop problem occurred first, while jdk12 was still permitted as a bootjdk for jdk14. Then the bootjdk requirements were updated to require jdk13. We'll get the machines and builds updated to use jdk13, and see if this problem still occurs.
Okay, because I've been setting up a windows environment and I am able to get to a further step in the build process locally, so I am not sure what the issue is just yet.
@adam-thorpe we've gotten jdk13 installed and updated the jdk14 builds to use it. The java.desktop<cr>
problem still occurs.
https://ci.eclipse.org/openj9/job/Build_JDKnext_x86-64_windows_OpenJDK/70
Thats odd but okay I'll look into it and see if I can spot the difference between this and my local machine
I can't seem to recreate the issue locally, and cannot test on another jenkins instance since they do not have boot JDK 13 so cannot determine if this is an isolated problem or not. One difference I did note was that eclipse is using cygwin v2.10.0, whereas the latest version is 3.0.7. Since cygwin does not support old versions, I cannot install this locally to test whether something in the older version is causing a problem.
Testing on hyc runtimes jenkins with the first failing SHAs does not replicate the issue: https://hyc-runtimes-jenkins.swg-devops.com/job/Build_JDKnext_x86-64_windows_Personal/17/console (using cygwin 2.11.0).
This commit caused the issue: https://github.com/ibmruntimes/openj9-openjdk-jdk/commit/fcbc07f47ad02b560345459bafe0e071ecc980b3 I'm suspecting a combination of this and possibly the significance of the \<cr> in the dependency, so 'java.desktop' is defined, but not 'java.desktop\<cr>' and I would hazzard a guess on non-windows platforms the \<cr> is ignored so they match...?
The generated module-deps.gmk for openj9 modules has \<cr>'s in it:
22:17:22 DEPS_openj9.dtfj := java.base java.desktop
java.logging
java.xml
openj9.traceformat
22:17:22 TRANSITIVE_MODULES_openj9.dtfj := java.base java.desktop
java.logging
22:17:22 DEPS_openj9.dtfjview := java.base openj9.dtfj
java.logging
22:17:22 TRANSITIVE_MODULES_openj9.dtfjview := java.base
22:17:22 DEPS_openj9.gpu := java.base java.base
openj9.cuda
There is a new blocking problem now, #7500
The blocker is resolved.
https://ci.eclipse.org/openj9/job/Build_JDKnext_x86-64_windows_OpenJDK/63