Need to re-vamp the build system so that it always uses GCJ, but also need to ensure correctness of the code using ECJ. A compromise is to make the ECJ pass part of make check :-/
ajocksch cagney: ah, I see where you're going. I've got my alternatives pointing to sun right now, I'll change it to gcj and see if that fixes it cagney ajocksch: yea, but sun or ibm? strace showed it sucking ibm's .so's ajocksch cagney: yep, that fixes it ajocksch cagney: sun cagney really confused ajocksch isn't much clearer cagney waits for ajocksch to notice eclipse behaving wierd ;-0 ajocksch hehe, not yet, we'll see though cagney ajocksch: check http://sourceware.org/bugzilla/show_bug.cgi?id=1852 ; given this scrambled behavior perhaphs it should be implemented cagney ajocksch: fyi, what happens is that the compiler looks at its Note from IRC: turns out that on FC-5 gcj is a shells script that invokes the current java-alternatives interpreter and, when it isn't gij, can't find things like gnu.gij.RawData. Below outlines solution: containing JVM's navel to find its own lint^Djar files ajocksch cagney: yeah, that makes sense to me cagney ajocksch: can you try something like: make 'JAVAC=gij -classpath /usr/share/java/eclipse-ecj.jar org.eclipse.jdt.internal.compiler.batch.Main' cagney ajocksch: antlr is run using something similar : :-P ajocksch in frysk-sys? hang ong ajocksch *on cagney ongkay ajocksch cagney: that works ajocksch (with alternatives set to sun's jvm) cagney ajocksch: thanks! I'll add it to the bug
Made optional part of the build ... Index: common/ChangeLog 2006-01-28 Andrew Cagney <cagney@redhat.com> * acinclude.m4: General cleanup. (JAVAC, JAVA): Always set to GCJ, and JAVA. (ECJ_JAVA, ECJ, HAVE_ECJ): Define. * Makefile.gen.sh: Delete code generating rules for $(GEN_CLASSDIR)/files, ${GEN_DIRNAME}.jar. * Makefile.rules (JAVAC, JAVACFLAGS): Delete. (files.list, ${GEN_DIRNAME}.jar, ecj/checked): Moved to here from Makefile.gen.sh. (all-local): Add ecj/checked, when HAVE_ECJ.