Seems reasonable. I personally have a preference for using the bootjdk for JT_JAVA but that's mostly a consequence of the bits that I work on--testing a HashMap change is more difficult when jtreg itself may not be working correctly.
Mike On Jun 5 2014, at 10:37 , Maurizio Cimadamore <maurizio.cimadam...@oracle.com> wrote: > Hi, > I've spent some time investigating why jck compiler tests could not be > launched from toplevel make - i.e. using > > make test TEST=langtools_jck-compiler > > Turns out that langtools makefile expects certain variables, while the build > provide others, and those are not correctly wired up. On top of that, > manually setting an extra make variable (for JCK_HOME) would have no effect, > as such variable would not be propagated by the make pipeline (toplevel make > -> test/Makefile -> langtools/test/Makefile). > > The following patch seems to fix the issue - should I charge forward with it? > > diff -r 7d27dbb50d95 test/Makefile > --- a/test/Makefile Wed Jun 04 10:03:17 2014 -0700 > +++ b/test/Makefile Thu Jun 05 15:48:54 2014 +0100 > @@ -58,7 +58,7 @@ > > # Test targets > langtools_% : > - @$(NO_STOPPING)$(call SUBDIR_TEST, $(LANGTOOLS_DIR), > JT_JAVA=$(PRODUCT_HOME) JTREG_HOME=$(JT_HOME) TEST="$(subst langtools_,,$@)" > $(subst langtools_,,$@)) > + @$(NO_STOPPING)$(call SUBDIR_TEST, $(LANGTOOLS_DIR), > JCK_HOME=$(JCK_HOME) TESTJAVA=$(PRODUCT_HOME) JT_JAVA=$(PRODUCT_HOME) > JCK_CONCURRENCY=$(CONCURRENCY) JTREG_HOME=$(JT_HOME) TEST="$(subst > langtools_,,$@)" $(subst langtools_,,$@)) > > jdk_% core_%s svc_%: > @$(NO_STOPPING)$(call SUBDIR_TEST, $(JDK_DIR), TEST="$@" $@) > > > Maurizio
