still problems with exec() in jdk16

[Michiel Boland]

Hi. Sorry to be a bore, but there are still problems with exec() in the 
current java/jdk16 port. I noticed this after the subversion module 
in netbeans broke. I narrowed it down to the following.

import java.io.IOException;

public class ExecTest {
     public static void main(String[] args) throws IOException, 
InterruptedException {
         Process p = Runtime.getRuntime().exec("ls", new String[] { } );
         int exitcode = p.waitFor();
         System.out.println("exitcode = " + exitcode);
     }
}

This program, when run with the jdk16 java will generate a hot spot error.

Stack trace from hs_err_pid*.log:

C  [libc.so.7+0xc6d30]  strcmp+0x60
C  [libjava.so+0x20317]  Java_java_lang_UNIXProcess_forkAndExec+0x367

The only reference to strcmp in UNIXProcess_md.c that makes sense here is 
in execvpe

         /* Parent and child PATH the same?  Use child PATH. */
         || (strcmp(parentPath, effectivePath()) == 0)

It appears that parentPath is 0. The only way that could happen is that 
Java_java_lang_UNIXProcess_initIDs is never called.

I believe the following patch needs to be made to UNIXProcess.java.bsd. 
(In effect, UNIXProcess.java.bsd should be identical to 
UNIXProcess.java.linux.)

--- UNIXProcess.java.bsd.orig	2007-08-31 20:43:03.000000000 +0200
+++ UNIXProcess.java.bsd	2007-08-31 21:24:40.000000000 +0200
@@ -195,4 +195,10 @@
          }
      }

+    /* This routine initializes JNI field offsets for the class */
+    private static native void initIDs();
+
+    static {
+	initIDs();
+    }
  }