Updates.

Mon Mar 27 23:04:30 2006  Soeren Sandmann  <sandmann@redhat.com>

        * TODO: Updates.

        * sfile.c (handle_{begin,end}_element): Don't generate begin/end
        instructions for value elements.

        * sfile.c (build_instructions): Remove debugging spew.

ChangeLog

@@ -1,3 +1,12 @@
+Mon Mar 27 23:04:30 2006  Soeren Sandmann  <sandmann@redhat.com>
+
+	* TODO: Updates.
+
+	* sfile.c (handle_{begin,end}_element): Don't generate begin/end
+	instructions for value elements.
+
+	* sfile.c (build_instructions): Remove debugging spew.
+
 Mon Mar 27 21:44:04 2006  Soeren Sandmann  <sandmann@redhat.com>
 
 	* sformat.[ch]: New files containing a simplified and sanitized

TODO

@@ -128,16 +128,6 @@ Before 1.2:
 	
 	- split out dfa in its own generic class
 	
-	- Formats should become first-class, stand-alone objects that offers
- 	  help with parsing and nothing else.
-
-		ParseContext* format_get_parse_context (format, err);
-		gboolean parse_context_begin (parse_context, name, err);
-		gboolean parse_context_end (parse_format, name, err);
-		
-	  basically, a Format encapsulates a DFA, and a ParseContext encapsulates
-	  the current state. 
-
 	- make a generic representation of xml files with quarks for strings:
 		struct item { 
 			int begin/end/text;
@@ -313,6 +303,18 @@ When the interrupt happens,
 
 Later:
 
+- Find out if sysprof accurately reports time spent handling pagefaults.
+  There is evidence that it doesn't:
+	- a version of sysprof took 10 seconds to load a certain profile.
+	  Profiling itself it appeared that most of the time was spent
+	  in the GMarkup parser
+	- a newer version of sysprof with significantly more compact
+	  Instructions took about 5 seconds, but the profile looked
+	  about the same.
+  The difference between the two versions has to be in page faults/
+  memory speed, but the profiles looked identically. 
+      Try and reproduce this result in a more controlled experiment.
+
 - See if it is possible to group the X server activity under the process that
   generated it.
 
@@ -490,6 +492,16 @@ Later:
 
 DONE:
 
+- Formats should become first-class, stand-alone objects that offers
+  help with parsing and nothing else.
+
+	ParseContext* format_get_parse_context (format, err);
+	gboolean parse_context_begin (parse_context, name, err);
+	gboolean parse_context_end (parse_format, name, err);
+		
+  basically, a Format encapsulates a DFA, and a ParseContext encapsulates
+  the current state. 
+
 - make stackstash ref counted
 
 - Charge 'self' properly to processes that don't get any stack trace at all 

sfile.c

@@ -210,10 +210,6 @@ sfile_get_pointer (SFileInput  *file,
 {
     Instruction *instruction;
     
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_pointer (instruction->type) &&
-                      check_name (instruction, name));
-    
     instruction = file->current_instruction++;
     g_return_if_fail (stype_is_pointer (instruction->type));
     
@@ -228,12 +224,7 @@ sfile_get_pointer (SFileInput  *file,
 	
         g_hash_table_insert (file->instructions_by_location, location, instruction);
     }
-    
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_pointer (instruction->type) &&
-                      check_name (instruction, name));
-    
-}
+ }
 
 void
 sfile_get_integer      (SFileInput  *file,
@@ -242,19 +233,11 @@ sfile_get_integer      (SFileInput  *file,
 {
     Instruction *instruction;
 
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_integer (instruction->type) &&
-                      check_name (instruction, name));
-    
     instruction = file->current_instruction++;
     g_return_if_fail (stype_is_integer (instruction->type));
     
     if (integer)
         *integer = instruction->u.integer.value;
-    
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_integer (instruction->type) &&
-                      check_name (instruction, name));
 }
 
 void
@@ -264,19 +247,11 @@ sfile_get_string       (SFileInput  *file,
 {
     Instruction *instruction;
     
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_string (instruction->type) &&
-                      check_name (instruction, name));
-    
     instruction = file->current_instruction++;
     g_return_if_fail (stype_is_string (instruction->type));
     
     if (string)
         *string = g_strdup (instruction->u.string.value);
-    
-    instruction = file->current_instruction++;
-    g_return_if_fail (stype_is_string (instruction->type) &&
-                      check_name (instruction, name));
 }
 
 static void
@@ -386,10 +361,12 @@ handle_begin_element (GMarkupParseContext *parse_context,
         set_unknown_element_error (err, "<%s> unexpected here", element_name);
         return;
     }
-    
-    /* FIXME - not10: is there really a reason to add begin/end instructions for values? */
-    instruction.kind = BEGIN;
-    g_array_append_val (build->instructions, instruction);
+
+    if (stype_is_list (instruction.type) || stype_is_record (instruction.type))
+    {
+	instruction.kind = BEGIN;
+	g_array_append_val (build->instructions, instruction);
+    }
 }
 
 static void
@@ -408,9 +385,12 @@ handle_end_element (GMarkupParseContext *context,
         return;
     }
     
-    instruction.kind = END;
+    if (stype_is_list (instruction.type) || stype_is_record (instruction.type))
+    {
+	instruction.kind = END;
     
-    g_array_append_val (build->instructions, instruction);
+	g_array_append_val (build->instructions, instruction);
+    }
 }
 
 static gboolean
@@ -671,8 +651,6 @@ build_instructions (const char *contents, SFormat *format, int *n_instructions,
 
     *n_instructions = build.instructions->len;
 
-    g_print ("n instructions: %d\n", *n_instructions);
-    
     return (Instruction *)g_array_free (build.instructions, FALSE);
 }