+2003-01-15 Havoc Pennington <hp@redhat.com>
+
+ * dbus/dbus-test.c (dbus_internal_symbol_do_not_use_run_tests):
+ fix build when --disable-tests
+
+ * Makefile.am (EXTRA_DIST): put HACKING in here
+
+ * HACKING: document procedure for making a tarball release.
+
+2003-01-14 Anders Carlsson <andersca@codefactory.se>
+
+ * bus/connection.c: (connection_error_handler),
+ (bus_connection_setup):
+ * bus/main.c: (main):
+ Make sure that the DBusConnectionData struct is NULLed
+ out to prevent a segfault.
+
+ * dbus/dbus-errors.c: (dbus_result_to_string):
+ * dbus/dbus-errors.h:
+ * dbus/dbus-message.c: (dbus_message_get_fields),
+ (dbus_message_get_fields_valist), (_dbus_message_test):
+ * dbus/dbus-message.h:
+ Make dbus_message_get_fields return a result code so we can
+ track invalid fields as well as oom.
+
+2003-01-11 Havoc Pennington <hp@pobox.com>
+
+ * configure.in: change --enable-test/--enable-ansi action-if-given
+ to enable_foo=$enableval instead of enable_foo=yes
+
+2003-01-08 Havoc Pennington <hp@pobox.com>
+
+ * dbus/dbus-string.c (_dbus_string_align_length): new function
+
+ * dbus/dbus-test-main.c: move main() for test app here
+ * dbus/dbus-test.c
+ (dbus_internal_symbol_do_not_use_run_tests): we have to export a
+ symbol to run tests, because dbus-test isn't in the main
+ library
+
+ Code review nitpicks.
+
+ * dbus/dbus-message.c (dbus_message_write_header): add newlines
+ for people with narrow emacs ;-). Assert client_serial was filled
+ in. Assert message->name != NULL.
+ (dbus_message_append_fields): have "first_field_type" arg separate
+ from va list, needed for C++ binding that also uses varargs IIRC
+ and helps with type safety
+ (dbus_message_new): add @todo about using DBusString to store
+ service/name internally
+ (dbus_message_new): don't leak ->service and ->name on OOM later
+ in the function
+ (dbus_message_unref): free the service name
+ (dbus_message_get_fields): same change to varargs
+ i.e. first_field_type
+ (_dbus_message_loader_return_buffer): assert that the message data
+ is aligned (if not it's a bug in our code). Put in verbose griping
+ about why we set corrupted = TRUE.
+ (decode_header_data): add FIXME that char* is evil. Was going to
+ add FIXME about evil locale-specific string.h strncmp, but just
+ switched to wacky string-as-uint32 optimization. Move check for
+ "no room for field name" above get_const_data_len() to avoid
+ assertion failure in get_const_data_len if we have trailing 2
+ bytes or the like. Check for service and name fields being
+ provided twice. Don't leak service/name on error. Require field
+ names to be aligned to 4 bytes.
+
+ * dbus/dbus-marshal.c: move byte swap stuff to header
+ (_dbus_pack_int32): uscore-prefix
+ (_dbus_unpack_int32): uscore-prefix
+ (_dbus_unpack_uint32): export
+ (_dbus_demarshal_string): add @todo complaining about use of
+ memcpy()
+ (_dbus_marshal_get_field_end_pos): add @todo about bad error
+ handling allowing corrupt data to go unchecked
+
2003-01-08 Havoc Pennington <hp@redhat.com>
* dbus/dbus-transport-unix.c (unix_do_iteration): add read/write