On Fri, 26 Jan 2024 19:33:49 GMT, Aleksey Shipilev <sh...@openjdk.org> wrote:
>> Current GHA runs produce lots of warnings: >> >> Node.js 16 actions are deprecated. Please update the following actions to >> use Node.js 20: actions/cache@v3, actions/download-artifact@v3, >> actions/upload-artifact@v3. For more information see: >> https://github.blog/changelog/2023-09-22-github-actions-transitioning-from-node-16-to-node-20/. >> >> We can upgrade to new actions to get the Node20. >> >> Release/migration notes: >> https://github.com/actions/cache#whats-new >> https://github.com/actions/upload-artifact/blob/main/docs/MIGRATION.md >> https://github.com/actions/download-artifact/blob/main/docs/MIGRATION.md >> https://github.com/actions/github-script#breaking-changes >> >> There is also msys2/setup-msys2, which was pinned by >> [JDK-8310259](https://bugs.openjdk.org/browse/JDK-8310259). We need at least >> 2.21.0 to get Node 20: >> https://github.com/msys2/setup-msys2/blob/main/CHANGELOG.md. I think we can >> unpin msys2 at this point. >> >> I don't think any of the migration problems outlined in those notes apply to >> our workflows. >> >> Additional testing: >> - [x] 3x GHA with cleaned caches >> - [x] 3x GHA with populated caches (default) > > Aleksey Shipilev has updated the pull request incrementally with one > additional commit since the last revision: > > Also update github-script > [We pinned it](https://bugs.openjdk.org/browse/JDK-8310259) because there > were weird failures in jtreg builds on Windows due to msys2 bug. This is why > I tested GHA multiple times, every time clearing the caches to trigger jtreg > rebuild. msys2 had fixed lots of possibly related path conversion bugs > recently, but it is really hard to pin-point what went into latest release. > Example of msys2 fix: > [msys2/MSYS2-packages#4200](https://github.com/msys2/MSYS2-packages/pull/4200) > -- done in Nov 2023, so the current release hopefully has all these. Ah-ha! Here we go. This PR proposes updating to `v2.22.0` (Update base distribution to 20240113). So the easy test for the hypothesis that November fixes got there is trying with `v2.21.0` (Update base distribution to 20231026). Tried it right now, and Windows jtreg build [immediately failed](https://github.com/shipilev/jdk/actions/runs/7694685053/job/20965966992): /bin/cp /d/a/jdk/jdk/jtreg/src/build/deps/jtharness/build/binaries/lib/javatest.jar D:/a/jdk/jdk/jtreg/src/build/images/jtreg/lib/javatest.jar CLASSPATH="D:/a/jdk/jdk/jtreg/src/build/classes;D:/a/jdk/jdk/jtreg/src/build/deps/jtharness/build/binaries/lib/javatest.jar;" \ C:/hostedtoolcache/windows/Java_Temurin-Hotspot_jdk/17.0.10-7/x64/bin/javac --release 11 -Xlint:all,-options,-deprecation -Werror \ -d D:/a/jdk/jdk/jtreg/src/build/classes \ -encoding ASCII \ ../src/share/classes/com/sun/javatest/diff/Diff.java ../src/share/classes/com/sun/javatest/diff/DiffReader.java ../src/share/classes/com/sun/javatest/diff/Fault.java ../src/share/classes/com/sun/javatest/diff/Help.java ../src/share/classes/com/sun/javatest/diff/HTMLReporter.java ../src/share/classes/com/sun/javatest/diff/HTMLWriter.java ../src/share/classes/com/sun/javatest/diff/Main.java ../src/share/classes/com/sun/javatest/diff/MultiMap.java ../src/share/classes/com/sun/javatest/diff/Reporter.java ../src/share/classes/com/sun/javatest/diff/ReportReader.java ../src/share/classes/com/sun/javatest/diff/SimpleReporter.java ../src/share/classes/com/sun/javatest/diff/StandardDiff.java ../src/share/classes/com/sun/javatest/diff/StatusComparator.java ../src/share/classes/com/sun/javatest/diff/SuperDiff.java ../src/share/classes/com/sun/javatest/diff/WorkDirectoryReader.java /bin/sh: line 1: C:/hostedtoolcache/windows/Java_Temurin-Hotspot_jdk/17.0.10-7/x64/bin/javac: Bad address make: *** [jtdiff.gmk:36: D:/a/jdk/jdk/jtreg/src/build/classes.com.sun.javatest.diff.ok] Error 126 This gives me even more confidence in msys2 unpinning :) ------------- PR Comment: https://git.openjdk.org/jdk/pull/17572#issuecomment-1914418168
