OSDN Git Service

2002-10-23 David Carlton <carlton@math.stanford.edu>
authorDavid Carlton <carlton@bactrian.org>
Wed, 23 Oct 2002 16:40:55 +0000 (16:40 +0000)
committerDavid Carlton <carlton@bactrian.org>
Wed, 23 Oct 2002 16:40:55 +0000 (16:40 +0000)
* dwarf2read.c (dwarf_tag_name): Add DWARF 3 names.
(dwarf_attr_name): Ditto.
(dwarf_type_encoding_name): Ditto.
(scan_partial_symbols): Descend into DW_TAG_namespace entries.
(process_die): Handle DW_TAG_namespace,
DW_TAG_imported_declaration, DW_TAG_imported_module.
(read_namespace): New function.

gdb/ChangeLog
gdb/dwarf2read.c

index fec6b76..bb555e1 100644 (file)
@@ -1,3 +1,13 @@
+2002-10-23  David Carlton  <carlton@math.stanford.edu>
+
+       * dwarf2read.c (dwarf_tag_name): Add DWARF 3 names.
+       (dwarf_attr_name): Ditto.
+       (dwarf_type_encoding_name): Ditto.
+       (scan_partial_symbols): Descend into DW_TAG_namespace entries.
+       (process_die): Handle DW_TAG_namespace,
+       DW_TAG_imported_declaration, DW_TAG_imported_module.
+       (read_namespace): New function.
+
 2002-10-22  Joel Brobecker  <brobecker@gnat.com>
 
        * configure.in: Define NEW_PROC_API on Interix too.
index d6de7ef..29fc164 100644 (file)
@@ -848,6 +848,9 @@ static void read_structure_scope (struct die_info *, struct objfile *,
 static void read_common_block (struct die_info *, struct objfile *,
                               const struct comp_unit_head *);
 
+static void read_namespace (struct die_info *die, struct objfile *objfile,
+                           const struct comp_unit_head *cu_header);
+
 static void read_enumeration (struct die_info *, struct objfile *,
                              const struct comp_unit_head *);
 
@@ -1338,6 +1341,17 @@ scan_partial_symbols (char *info_ptr, struct objfile *objfile,
 
   int nesting_level = 1;
 
+  /* We only want to read in symbols corresponding to variables or
+     other similar objects that are global or static.  Normally, these
+     are all children of the DW_TAG_compile_unit die, so are all at
+     level 1.  But C++ namespaces give ries to DW_TAG_namespace dies
+     whose children are global objects.  So we keep track of what
+     level we currently think of as referring to file scope; this
+     should always equal 1 plus the number of namespaces that we are
+     currently nested within.  */
+  
+  int file_scope_level = 1;
+
   *lowpc = ((CORE_ADDR) -1);
   *highpc = ((CORE_ADDR) 0);
 
@@ -1360,7 +1374,7 @@ scan_partial_symbols (char *info_ptr, struct objfile *objfile,
                    {
                      *highpc = pdi.highpc;
                    }
-                 if ((pdi.is_external || nesting_level == 1)
+                 if ((pdi.is_external || nesting_level == file_scope_level)
                      && !pdi.is_declaration)
                    {
                      add_partial_symbol (&pdi, objfile, cu_header);
@@ -1373,46 +1387,65 @@ scan_partial_symbols (char *info_ptr, struct objfile *objfile,
            case DW_TAG_structure_type:
            case DW_TAG_union_type:
            case DW_TAG_enumeration_type:
-             if ((pdi.is_external || nesting_level == 1)
+             if ((pdi.is_external || nesting_level == file_scope_level)
                  && !pdi.is_declaration)
                {
                  add_partial_symbol (&pdi, objfile, cu_header);
                }
              break;
            case DW_TAG_enumerator:
-             /* File scope enumerators are added to the partial symbol
-                table.  */
-             if (nesting_level == 2)
+             /* File scope enumerators are added to the partial
+                symbol table.  They're children of the enumeration
+                type die, so they occur at a level one higher than we
+                normally look for.  */
+             if (nesting_level == file_scope_level + 1)
                add_partial_symbol (&pdi, objfile, cu_header);
              break;
            case DW_TAG_base_type:
              /* File scope base type definitions are added to the partial
                 symbol table.  */
-             if (nesting_level == 1)
+             if (nesting_level == file_scope_level)
                add_partial_symbol (&pdi, objfile, cu_header);
              break;
+           case DW_TAG_namespace:
+             /* FIXME: carlton/2002-10-16: we're not yet doing
+                anything useful with this, but for now make sure that
+                these tags at least don't cause us to miss any
+                important symbols.  */
+             if (pdi.has_children)
+               file_scope_level++;
            default:
              break;
            }
        }
 
-      /* If the die has a sibling, skip to the sibling.
-         Do not skip enumeration types, we want to record their
-         enumerators.  */
-      if (pdi.sibling && pdi.tag != DW_TAG_enumeration_type)
+      /* If the die has a sibling, skip to the sibling.  Do not skip
+         enumeration types, we want to record their enumerators.  Do
+         not skip namespaces, we want to record symbols inside
+         them.  */
+      if (pdi.sibling
+         && pdi.tag != DW_TAG_enumeration_type
+         && pdi.tag != DW_TAG_namespace)
        {
          info_ptr = pdi.sibling;
        }
       else if (pdi.has_children)
        {
-         /* Die has children, but the optional DW_AT_sibling attribute
-            is missing.  */
+         /* Die has children, but either the optional DW_AT_sibling
+            attribute is missing or we want to look at them.  */
          nesting_level++;
        }
 
       if (pdi.tag == 0)
        {
          nesting_level--;
+         /* If this is the end of a DW_TAG_namespace entry, then
+            decrease the file_scope_level, too.  */
+         if (nesting_level < file_scope_level)
+           {
+             file_scope_level--;
+             gdb_assert (nesting_level == file_scope_level);
+           }
        }
     }
 
@@ -1712,6 +1745,19 @@ process_die (struct die_info *die, struct objfile *objfile,
       break;
     case DW_TAG_common_inclusion:
       break;
+    case DW_TAG_namespace:
+      read_namespace (die, objfile, cu_header);
+      break;
+    case DW_TAG_imported_declaration:
+    case DW_TAG_imported_module:
+      /* FIXME: carlton/2002-10-16: Eventually, we should use the
+        information contained in these.  DW_TAG_imported_declaration
+        dies shouldn't have children; DW_TAG_imported_module dies
+        shouldn't in the C++ case, but conceivably could in the
+        Fortran case, so we'll have to replace this gdb_assert if
+        Fortran compilers start generating that info.  */
+      gdb_assert (!die->has_children);
+      break;
     default:
       new_symbol (die, NULL, objfile, cu_header);
       break;
@@ -2927,6 +2973,27 @@ read_common_block (struct die_info *die, struct objfile *objfile,
     }
 }
 
+/* Read a C++ namespace.  */
+
+/* FIXME: carlton/2002-10-16: For now, we don't actually do anything
+   useful with the namespace data: we just process its children.  */
+
+static void
+read_namespace (struct die_info *die, struct objfile *objfile,
+               const struct comp_unit_head *cu_header)
+{
+  if (die->has_children)
+    {
+      struct die_info *child_die = die->next;
+      
+      while (child_die && child_die->tag)
+       {
+         process_die (child_die, objfile, cu_header);
+         child_die = sibling_die (child_die);
+       }
+    }
+}
+
 /* Extract all information from a DW_TAG_pointer_type DIE and add to
    the user defined type vector.  */
 
@@ -5512,6 +5579,22 @@ dwarf_tag_name (register unsigned tag)
       return "DW_TAG_variable";
     case DW_TAG_volatile_type:
       return "DW_TAG_volatile_type";
+    case DW_TAG_dwarf_procedure:
+      return "DW_TAG_dwarf_procedure";
+    case DW_TAG_restrict_type:
+      return "DW_TAG_restrict_type";
+    case DW_TAG_interface_type:
+      return "DW_TAG_interface_type";
+    case DW_TAG_namespace:
+      return "DW_TAG_namespace";
+    case DW_TAG_imported_module:
+      return "DW_TAG_imported_module";
+    case DW_TAG_unspecified_type:
+      return "DW_TAG_unspecified_type";
+    case DW_TAG_partial_unit:
+      return "DW_TAG_partial_unit";
+    case DW_TAG_imported_unit:
+      return "DW_TAG_imported_unit";
     case DW_TAG_MIPS_loop:
       return "DW_TAG_MIPS_loop";
     case DW_TAG_format_label:
@@ -5656,7 +5739,30 @@ dwarf_attr_name (register unsigned attr)
       return "DW_AT_virtuality";
     case DW_AT_vtable_elem_location:
       return "DW_AT_vtable_elem_location";
-
+    case DW_AT_allocated:
+      return "DW_AT_allocated";
+    case DW_AT_associated:
+      return "DW_AT_associated";
+    case DW_AT_data_location:
+      return "DW_AT_data_location";
+    case DW_AT_stride:
+      return "DW_AT_stride";
+    case DW_AT_entry_pc:
+      return "DW_AT_entry_pc";
+    case DW_AT_use_UTF8:
+      return "DW_AT_use_UTF8";
+    case DW_AT_extension:
+      return "DW_AT_extension";
+    case DW_AT_ranges:
+      return "DW_AT_ranges";
+    case DW_AT_trampoline:
+      return "DW_AT_trampoline";
+    case DW_AT_call_column:
+      return "DW_AT_call_column";
+    case DW_AT_call_file:
+      return "DW_AT_call_file";
+    case DW_AT_call_line:
+      return "DW_AT_call_line";
 #ifdef MIPS
     case DW_AT_MIPS_fde:
       return "DW_AT_MIPS_fde";
@@ -6093,6 +6199,8 @@ dwarf_type_encoding_name (register unsigned enc)
       return "DW_ATE_unsigned";
     case DW_ATE_unsigned_char:
       return "DW_ATE_unsigned_char";
+    case DW_ATE_imaginary_float:
+      return "DW_ATE_imaginary_float";
     default:
       return "DW_ATE_<unknown>";
     }