summaryrefslogtreecommitdiffstats
path: root/dbus/dbus-marshal-recursive.h
diff options
context:
space:
mode:
Diffstat (limited to 'dbus/dbus-marshal-recursive.h')
-rw-r--r--dbus/dbus-marshal-recursive.h13
1 files changed, 10 insertions, 3 deletions
diff --git a/dbus/dbus-marshal-recursive.h b/dbus/dbus-marshal-recursive.h
index 79ce52c2..064bf156 100644
--- a/dbus/dbus-marshal-recursive.h
+++ b/dbus/dbus-marshal-recursive.h
@@ -46,6 +46,8 @@
* - implement has_next()
* - the all-in-one-block array accessors
* - validation
+ *
+ * - remember to try a HAVE_INT64=0 build at the end
*/
typedef struct DBusTypeMark DBusTypeMark;
@@ -54,7 +56,12 @@ typedef struct DBusTypeWriter DBusTypeWriter;
typedef struct DBusTypeReaderClass DBusTypeReaderClass;
/* The mark is a way to compress a TypeReader; it isn't all that
- * successful though.
+ * successful though. The idea was to use this for caching header
+ * fields in dbus-message.c. However now I'm thinking why not cache
+ * the actual values (e.g. char*) and if the field needs to be set or
+ * deleted, just linear search for it. Those operations are uncommon,
+ * and getting the values should be fast and not involve all this type
+ * reader nonsense.
*/
struct DBusTypeMark
{
@@ -145,6 +152,8 @@ void _dbus_type_reader_get_signature (const DBusTypeReader *
const DBusString **str_p,
int *start_p,
int *len_p);
+dbus_bool_t _dbus_type_reader_set_basic (DBusTypeReader *reader,
+ const void *value);
void _dbus_type_writer_init (DBusTypeWriter *writer,
int byte_order,
@@ -169,6 +178,4 @@ dbus_bool_t _dbus_type_writer_unrecurse (DBusTypeWriter *writer,
dbus_bool_t _dbus_type_writer_write_reader (DBusTypeWriter *writer,
DBusTypeReader *reader);
-
-
#endif /* DBUS_MARSHAL_RECURSIVE_H */