A race condition, possibly leading to a failed build on Solaris, was introduced with JDK-8072842.
The two Solaris helper script fix_empty_sec_hdr_flags and add_gnu_debuglink are built several times, in the same location. If we are unlucky then thread A does: A1: compile add_gnu_link.c A2: link add_gnu_link and thread B does: B1: compile add_gnu_link.c after A1 but before A2, so that when linking is about to start, the .o file has been truncated, and then linking will fail. A typical error message looks like this: Main.gmk:264: recipe for target 'build-test-hotspot-jtreg-native' failed ld: fatal: file /opt/jprt/T/P1/002839.magnusi/s/build/solaris-sparcv9-normal-server-release/buildtools/objs/fix_empty_sec_hdr_flags/fix_empty_sec_hdr_flags.o: not an ELF object ld: fatal: file processing errors. No output written to /opt/jprt/T/P1/002839.magnusi/s/build/solaris-sparcv9-normal-server-release/buildtools/bin/fix_empty_sec_hdr_flags gmake[3]: *** [/opt/jprt/T/P1/002839.magnusi/s/build/solaris-sparcv9-normal-server-release/buildtools/bin/fix_empty_sec_hdr_flags] Error 2 The proper long-term solution is to fix JDK-8064808. However, since this issue causes a build break, a quicker and less risky fix is to let build-test-*-jtreg-native depend on buildtools-jdk. It is by definition hard to test that a race condition has been resolved, but I have tried this at least a dozen times with no failures. https://bugs.openjdk.java.net/browse/JDK-8074072 Fix inline: diff --git a/make/Main.gmk b/make/Main.gmk --- a/make/Main.gmk +++ b/make/Main.gmk @@ -433,6 +433,10 @@ test-make: clean-test-make + build-test-hotspot-jtreg-native: buildtools-jdk + + build-test-jdk-jtreg-native: buildtools-jdk + test-image-hotspot-jtreg-native: build-test-hotspot-jtreg-native test-image-jdk-jtreg-native: build-test-jdk-jtreg-native /Magnus
