Re: someone could port libgc

[Roland McGrath <roland@frob.com>]

I'm still undecided about the advisability of this (I'm not checking it in),
but you might try this patch to hurd/exec.c to put the stack at high address.

Index: exec.c
===================================================================
RCS file: /home/cvs/hurd/exec/exec.c,v
retrieving revision 1.72
diff -u -b -p -r1.72 exec.c
--- exec.c	1999/07/18 03:01:29	1.72
+++ exec.c	2000/01/23 20:45:54
@@ -1,5 +1,5 @@
 /* GNU Hurd standard exec server.
-   Copyright (C) 1992,93,94,95,96,98,99 Free Software Foundation, Inc.
+   Copyright (C) 1992,93,94,95,96,98,99,2000 Free Software Foundation, Inc.
    Written by Roland McGrath.
 
    Can exec ELF format directly.
@@ -1656,11 +1656,25 @@ do_exec (file_t file,
     goto out;
 
   /* Start up the initial thread at the entry point.  */
-  boot->stack_base = 0, boot->stack_size = 0; /* Don't care about values.  */
+#ifndef STACK_GROWS_UP		/* ? */
+  /* First try for a stack (of some arbitrary "maximal" size, with red zone)
+     at the top of the address space, since some things just like that better.
+     If we can't get that part of the address space, then just go anywhere.  */
+  boot->stack_size = (16 * 1024 * 1024); /* 16MB, arbitrary.  */
+  boot->stack_base = VM_MAX_ADDRESS - boot->stack_size - vm_page_size;
   e.error = mach_setup_thread (newtask, thread,
 			       (void *) (e.interp.section ? interp.entry :
 					 e.entry),
 			       &boot->stack_base, &boot->stack_size);
+  if (e.error == KERN_NO_SPACE)
+#endif
+    {
+      boot->stack_base = 0, boot->stack_size = 0;
+      e.error = mach_setup_thread (newtask, thread,
+				   (void *) (e.interp.section ? interp.entry :
+					     e.entry),
+				   &boot->stack_base, &boot->stack_size);
+    }
   if (e.error)
     goto out;