* printcmd.c (print_address), values.c (value_as_pointer): Don't

	use ADDR_BITS_REMOVE.

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index c85a369..798363e 100644 (file)
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,5 +1,8 @@
 Wed Aug 18 12:03:00 1993  Jim Kingdon  (kingdon@lioth.cygnus.com)
 
+       * printcmd.c (print_address), values.c (value_as_pointer): Don't
+       use ADDR_BITS_REMOVE.
+
        * blockframe.c (block_innermost_frame): Uncomment.
        Return NULL if passed NULL.
        * frame.h: Declare it.

diff --git a/gdb/printcmd.c b/gdb/printcmd.c
index cfe2e9b..8c5dc36 100644 (file)
--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -587,7 +587,9 @@ print_address (addr, stream)
      CORE_ADDR addr;
      FILE *stream;
 {
-#ifdef ADDR_BITS_REMOVE
+#if 0 && defined (ADDR_BITS_REMOVE)
+  /* This is wrong for pointer to char, in which we do want to print
+     the low bits.  */
   fprintf_filtered (stream, local_hex_format(), ADDR_BITS_REMOVE(addr));
 #else
   fprintf_filtered (stream, local_hex_format(), addr);
@@ -1628,7 +1630,14 @@ typedef struct {
   MAKEVA_EXTRA_INFO
 #endif
 
-  char arg_bytes[1];
+  /* Some systems (mips, pa) would like this to be aligned, and it never
+     will hurt.  */
+  union
+    {
+      char arg_bytes[1];
+      double force_double_align;
+      LONGEST force_long_align;
+    } aligner;
 } makeva_list;
 
 /* Tell the caller how many bytes to allocate for a makeva_list with NARGS
@@ -1639,9 +1648,7 @@ makeva_size (nargs, max_arg_size)
      unsigned int nargs;
      unsigned int max_arg_size;
 {
-  return sizeof (makeva_list) + nargs * max_arg_size
-    /* The PA might need up to this much for alignment.  */
-    + max_arg_size - 1;
+  return sizeof (makeva_list) + nargs * max_arg_size;
 }
 
 /* Start working on LIST with NARGS arguments and whose largest
@@ -1671,7 +1678,7 @@ makeva_arg (list, argaddr, argsize)
 #if defined (MAKEVA_ARG)
   MAKEVA_ARG (list, argaddr, argsize);
 #else
-  memcpy (&list->arg_bytes[list->argindex], argaddr, argsize);
+  memcpy (&list->aligner.arg_bytes[list->argindex], argaddr, argsize);
   list->argindex += argsize;
 #endif
 }
@@ -1686,7 +1693,7 @@ makeva_end (list)
   MAKEVA_END (list);
 #else
   /* This works if a va_list is just a pointer to the arguments.  */
-  return (va_list) list->arg_bytes;
+  return (va_list) list->aligner.arg_bytes;
 #endif
 }
 \f

diff --git a/gdb/values.c b/gdb/values.c
index 337dd10..65ecb23 100644 (file)
--- a/gdb/values.c
+++ b/gdb/values.c
@@ -552,7 +552,14 @@ value_as_pointer (val)
 {
   /* Assume a CORE_ADDR can fit in a LONGEST (for now).  Not sure
      whether we want this to be true eventually.  */
+#if 0
+  /* ADDR_BITS_REMOVE is wrong if we are being called for a
+     non-address (e.g. argument to "signal", "info break", etc.), or
+     for pointers to char, in which the low bits *are* significant.  */
   return ADDR_BITS_REMOVE(value_as_long (val));
+#else
+  return value_as_long (val);
+#endif
 }
 \f
 /* Unpack raw data (copied from debugee, target byte order) at VALADDR