I am curious why you didn't make it conditional based upon DEBUG_LEVEL like jdk and corba? It would be difficult/annoying to turn it on if needed.
The name "GENERATE_NEWBYTECODE_DEBUG" is also confusing since the bytecode is not tuned for debugging. Mike On Apr 15 2014, at 09:03 , Tim Bell <tim.b...@oracle.com> wrote: > Hello Erik: > >> The simplest way to get consistency is to just remove -g from jaxp and jaxws >> compile lines. Providing different levels of debug info for jre and jdk >> would certainly be nice but has a much bigger scope. >> >> Please review this small patch. > > Looks good to me. > > Tim > > >> Bug: https://bugs.openjdk.java.net/browse/JDK-8040267 >> Patch: >> diff -r 2d9f4166e0be make/BuildJaxws.gmk >> --- a/make/BuildJaxws.gmk >> +++ b/make/BuildJaxws.gmk >> @@ -38,7 +38,7 @@ >> $(eval $(call SetupJavaCompiler,GENERATE_NEWBYTECODE_DEBUG, \ >> JVM := $(JAVA), \ >> JAVAC := $(NEW_JAVAC), \ >> - FLAGS := -XDignore.symbol.file=true $(DISABLE_JAXWS_WARNINGS) -g, \ >> + FLAGS := -XDignore.symbol.file=true $(DISABLE_JAXWS_WARNINGS), \ >> SERVER_DIR := $(SJAVAC_SERVER_DIR), \ >> SERVER_JVM := $(SJAVAC_SERVER_JAVA))) >> >> diff -r 3b360a77658e make/BuildJaxp.gmk >> --- a/make/BuildJaxp.gmk >> +++ b/make/BuildJaxp.gmk >> @@ -38,7 +38,7 @@ >> $(eval $(call SetupJavaCompiler,GENERATE_NEWBYTECODE_DEBUG, \ >> JVM := $(JAVA), \ >> JAVAC := $(NEW_JAVAC), \ >> - FLAGS := -XDignore.symbol.file=true $(DISABLE_JAXP_WARNINGS) -g, \ >> + FLAGS := -XDignore.symbol.file=true $(DISABLE_JAXP_WARNINGS), \ >> SERVER_DIR := $(SJAVAC_SERVER_DIR), \ >> SERVER_JVM := $(SJAVAC_SERVER_JAVA))) >> >> /Erik >> >> On 2014-04-11 16:40, Erik Joelsson wrote: >>> Hello, >>> >>> While converting the build to the new build-infra makefiles, one thing that >>> annoyed me was the fact that we aren't consistently compiling with or >>> without -g for java code. In the new makefiles we just emulate the same >>> behavior, but I would like to sort it out properly now. >>> >>> Currently langtools, jaxp and jaxws repos build with -g always, while corba >>> and jdk only build with -g when DEBUG_LEVEL is fastdebug or slowdebug. >>> >>> How would we really like this to work? Is there a reason not to ship with >>> -g enabled? I know from personal experience that I get very annoyed when I >>> can't step into the jdk classes and look at local variable values when >>> debugging my own java applications. >>> >>> /Erik >> >