> 30 jan 2015 kl. 13:47 skrev matchew <mkors...@gmail.com>: > > Do I need to set JAVA and JAVAC explicitly? What should be the values here?
No, on the contrary, the failure could perhaps be caused by having them defined. /Magnus > > 2015-01-30 11:41 GMT+01:00 Magnus Ihse Bursie <magnus.ihse.bur...@oracle.com>: >>> On 2015-01-30 10:40, matchew wrote: >>> $ grep JAVA /opt/openjdk/8/build/linux-x86_64-normal-server-release/spec.gmk >>> ENABLE_SJAVAC:=no >>> SJAVAC_SERVER_DIR:= >>> JAVA_FLAGS:= -Xms64M -Xmx1100M -XX:PermSize=32m -XX:MaxPermSize=160m >>> -XX:ThreadStackSize=1536 >>> JAVA= $(BOOT_JDK)/bin/java $(JAVA_FLAGS) >>> JAVAC= $(BOOT_JDK)/bin/javac >>> JAVAC_FLAGS?= >>> JAVAH= $(BOOT_JDK)/bin/javah >>> # You run the new javac using the boot jdk with $(BOOT_JDK)/bin/java >>> $(NEW_JAVAC) ... >>> BOOTSTRAP_JAVAC_JAR:=$(LANGTOOLS_OUTPUTDIR)/dist/bootstrap/lib/javac.jar >>> BOOTSTRAP_JAVAC_ARGS:="-Xbootclasspath/p:$(BOOTSTRAP_JAVAC_JAR)" -cp >>> $(BOOTSTRAP_JAVAC_JAR) >>> NEW_JAVAC = $(BOOTSTRAP_JAVAC_ARGS) com.sun.tools.javac.Main >>> NEW_JAVADOC = $(BOOTSTRAP_JAVAC_ARGS) com.sun.tools.javadoc.Main >>> SJAVAC_SERVER_JAVA:= /usr/lib/jvm/java-7-openjdk-amd64/bin/java -verbosegc >>> -d64 -Xms1000M -Xmx1500M >>> >>> Yes, I did fresh clone from http://hg.openjdk.java.net/jdk8u/jdk8u but I am >>> using the same build script/procedure I was using before and it was totally >>> fine with last tag from http://hg.openjdk.java.net/jdk8/jdk8. >> >> That is *really* weird. I'm sorry I can't provide much more assistance at >> this point. Some suggestions: >> * Try removing the ALT_ variables as the warnings suggests. I don't believe >> they should interfere but the problem lies with the boot jdk so the >> ALT_BOOTDIR is a suspect. >> * Check your environment. Do you have JAVA or JAVAC set there? >> * Try running with LOG=debug and see if you can figure out what command line >> is being run and why. >> * If that does not help, running with LOG=trace produces even more logging >> on how make builds the command lines. >> * Both of the logging suggestions above is easier to handle if you first >> figure out a minimal target that reproduces the problem. Perhaps >> "langtools-only"? >> >> /Magnus >
