On Thu, 11 Jan 2024 12:36:31 GMT, Martin Doerr <mdo...@openjdk.org> wrote:
>> Regarding >> https://github.com/TheShermanTanker/jdk/actions/runs/7070564987/job/19247370401, >> could it be that the adlc build didn't get the correct C++ version flags? >> It doesn't look like a clang 13 specific problem. > >> @TheRealMDoerr >> >> > The only issue I see is requiring clang 14.0 on MacOS is not in sync with >> > "Other JDK 22 build platforms" >> > (https://wiki.openjdk.org/display/Build/Supported+Build+Platforms). >> >> That page is suppose to document what we actually do, not be a binding >> contract; so if we change stuff, we update the page to reflect it, rather >> than the other way around. >> >> Or maybe I misunderstood your comment? > > Correct, but raising requirements requires extra effort to change the build > environments, updating docs, etc. (It may even cause incompatibilities. > Probably not in this case.) While it may be better to use a newer Xcode on > Mac, I can't see sufficient reason for forcing the whole world to build with > clang 14. > @TheRealMDoerr The adlc build is notoriously problematic, since it does not > share the common flags set for JVM or JDK native compilation. :( So your > suggestion sounds highly likely to me. Running with LOG=cmdlines will confirm > this. > > (This can be done on GHA by manually starting a run, and setting the value of > "Additional make arguments" to `LOG=cmdlines` or possibly `LOG=info,cmdlines`) Thanks for the hint! The command line is also shown here: make-support/failure-logs/hotspot_variant-server_tools_adlc_objs_adlparse.o.cmdline The -std option is not passed. That seems to be the issue. So, this is not a clang 13 vs 14 thing. ------------- PR Comment: https://git.openjdk.org/jdk/pull/14988#issuecomment-1887095112
