2003-01-26 Havoc Pennington <hp@pobox.com>

        The unit tests pass, but otherwise untested.  If it breaks, the
	tests should have been better. ;-)

	* bus/driver.c (bus_driver_handle_hello): return if we disconnect
	the connection.

	* dbus/dbus-message.c: redo everything so we maintain
	message->header as the only copy of the various fields.
	This avoids the possibility of out-of-memory in some cases,
	for example dbus_message_lock() can't run out of memory anymore,
	and avoids extra copying. Figured I may as well go ahead and do
	this since it was busted for dbus_message_lock to not return
	failure on OOM, and dbus_message_write_header was totally
	unchecked for OOM. Also fixed some random other bugs.

	* dbus/dbus-marshal.c (_dbus_marshal_get_field_end_pos): verify
	that strings are nul-terminated. Also, end_pos can be equal
	to string length just not greater than, I think.
	(_dbus_marshal_set_int32): new function
	(_dbus_marshal_set_uint32): new function
	(_dbus_marshal_set_string): new function

	* dbus/dbus-connection.c (_dbus_connection_new_for_transport): fix
	a warning, init timeout_list to NULL
	(dbus_connection_send_message): don't use uninitialized variable
	"serial"

	* dbus/dbus-string.c (_dbus_string_replace_len): new function

1 files changed, 1 insertions, 0 deletions

diff --git a/bus/driver.c b/bus/driver.c
index 1fd26a26..9f159a22 100644
--- a/bus/driver.c
+++ b/bus/driver.c
@@ -147,6 +147,7 @@ bus_driver_handle_hello (DBusConnection *connection,
     {
       dbus_free (name);
       dbus_connection_disconnect (connection);
+      return;
     }
   
   _DBUS_HANDLE_OOM (_dbus_string_init (&unique_name, _DBUS_INT_MAX));