From owner-freebsd-java Fri Sep 3 8: 7: 1 1999 Delivered-To: freebsd-java@freebsd.org Received: from heresy.is.titech.ac.jp (heresy.is.titech.ac.jp [131.112.35.182]) by hub.freebsd.org (Postfix) with ESMTP id E223814C08; Fri, 3 Sep 1999 08:06:57 -0700 (PDT) (envelope-from fuyuhik8@is.titech.ac.jp) Received: from localhost (localhost.is.titech.ac.jp [127.0.0.1]) by heresy.is.titech.ac.jp (8.9.3/8.9.2) with ESMTP id AAA93605; Sat, 4 Sep 1999 00:06:06 +0900 (JST) (envelope-from fuyuhik8@is.titech.ac.jp) To: java-port@FreeBSD.org Cc: freebsd-java@FreeBSD.org Subject: Some bugs exist in JDK for FreeBSD. From: Maruyama Fuyuhiko X-Mailer: Mew version 1.94pre4 on XEmacs 21.2 (Demeter) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-Id: <19990904000606R.fuyuhik8@is.titech.ac.jp> Date: Sat, 04 Sep 1999 00:06:06 +0900 X-Dispatcher: imput version 990903(IM127) Lines: 49 Sender: owner-freebsd-java@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Hi, all! I'm a developer of Java JIT compiler. And I found some bugs in JDK-1.1.8 for FreeBSD. *First When you try division of two special int values, x86 cpu causes SIGFPE. You can easily see it. Try this "Test.java" compile with FreeBSD's javac. public class Test { public static void main(String[] args) { System.out.println("0x80000000 / -1 = " + (0x80000000/-1)); } } I think only this combination (i.e. 0x80000000/-1) causes unexcepted SIGFPE, so JavaVM may easily handle it. Our JIT compiler can handle it in support functions for JIT compiled methods, but I think JavaVM should handle it by itself! This bug may exist all version of JDK for FreeBSD (may be in Linux version), but I have checked only these versions. jdk1.1.8_AOUT.V99-7-19.tar.gz jdk1.1.8_ELF.V99-7-19.tar.gz *Second Radio buttons of first window of SPECjvm98 don't appear! We always use SPECjvm98 to measure the performance of our JIT. SPECjvm98 uses AWT, and it causes visual problem only in *ELF version* of jdk1.1.8 for FreeBSD, I beleave AOUT version works correctly. So, I cannot measure our JIT with ELF version in SPEC-confirmed style. Sorry, I have not figured out the precise condition, and I don't have enough time to do it. Are there someone can check the SPECjvm98 and figure out the bug? Thanks in advance. -- Fuyuhiko Maruyama Matsuoka laboratory, Department of Mathematical and Computing Sciences, Tokyo Institute of Technology. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message