Re: java/openjdk19 fails to build on aarch64
- In reply to: Jonathan Chen : "Re: java/openjdk19 fails to build on aarch64"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Thu, 20 Mar 2025 10:15:21 UTC
On Thu, Mar 20, 2025 at 04:33:15PM +1300, Jonathan Chen wrote: > On 19/03/25 22:24, Harald Eilertsen wrote: > > libraries.m4 should be patched before configure is run, which again > > produces spec.gmk. Otherwise it would be no point in patching it to > > begin with... > > You are correct. I only realised this just after I pressed the "Send" > button. So I dug into this a bit deeper, and it appears that > make/autoconf/build-aux directory has a later timestamp than spec.gmk after > the configure phase. This was from a clean work directory in the port? I see that gonfig.guess and config.sub for some reason is touched during the port build, but I'm not sure why. (This does not seem to happen when just building the jdk normally.) As you also observe this still happens before the spec is generated. Other than that I don't know what else that can happen to the dir during build. I'm not able to replicate this no matter what shenanigans I do to the work dir in the port. > To force the build to ignore the warning, I applied the following patch: > > diff --git a/java/openjdk19/Makefile b/java/openjdk19/Makefile > index 76e8910459..4b84b82092 100644 > --- a/java/openjdk19/Makefile > +++ b/java/openjdk19/Makefile > @@ -48,7 +48,7 @@ MAKE_ENV= LANG="C" \ > CC=${CC} \ > CXX=${CXX} \ > CPP=${CPP} \ > - MAKEFLAGS="" > + MAKEFLAGS="CONF_CHECK=ignore " What happens if you set CONF_CHECK=auto instead? Do you get a build cycle? Take care! H!