> On 27 Apr 2017, at 23:24, Mandy Chung <mandy.ch...@oracle.com> wrote: > > >> On Apr 27, 2017, at 7:47 AM, Doug Simon <doug.si...@oracle.com> wrote: >> >> [1] http://cr.openjdk.java.net/~dnsimon/8177845.02 > > I reviewed the top repo and jdk repo change. For hotspot change, > I reviewed jdk.vm.ci.services/* and jdk.internal.vm.ci module-info.java. > > make/make/CompileJavaModules.gmk > jdk/make/launcher/Launcher-jdk.aot.gmk > Reading the original jdk.internal.vm.ci module-info.java [1], > jdk.aot only accesses a small set of packages. e.g. it does not > access jdk.vm.ci.aarch64, jdk.vm.ci.code.stack, jdk.vm.ci.common, etc.
I know but I agree with Vladimir that there's no harm in exporting everything to jdk.aot to simplify continued development of this tool as support for other platforms is added. As I learnt putting this patch together, missing these exports when they're actually needed leads to very confusing compiler error messages. > 115 jdk.internal.vm.compiler \ > > This should be removed from PLATFORM_MODULES list. $PLATFORM_MODULES > already includes $UPGRADEABLE_MODULES. Will do. > Otherwise, looks good to me. Thanks for the review. > There are jdk tests verifying upgradeable modules and qualified exports. > I suggest to run through JPRT -testset core options to catch any test failure. Will do. -Doug
