* Throughout, grand renaming to strip out the use of "service",
just say "name" instead (or "bus name" when ambiguous). Did not
change the internal code of the message bus itself, only the
programmer-facing API and messages.
* doc/dbus-specification.xml: further update the message bus section
* bus/config-parser.c (all_are_equiv): fix bug using freed string
in error case
* dbus/dbus-types.h: remove 16-bit types since we don't use them
ever
* dbus/dbus-marshal-validate.c (_dbus_validate_path): disallow any
"invalid name character" not only non-ASCII
* doc/dbus-specification.xml: further update spec, message bus
parts are still out-of-date but the marshaling etc. stuff is now
accurate-ish
* Throughout, align variant bodies according to the contained
type, rather than always to 8. Should save a fair bit of space in
message headers.
* dbus/dbus-marshal-validate.c (_dbus_validate_body_with_reason):
fix handling of case where p == end
* doc/TODO: remove the dbus_bool_t item and variant alignment items
* dbus/dbus-types.h: hardcode dbus_bool_t to 32 bits
* Throughout: modify DBUS_TYPE_BOOLEAN to be a 32-bit type instead
of an 8-bit type. Now dbus_bool_t is the type to use whenever you
are marshaling/unmarshaling a boolean.
This is about it on what can be disabled/deleted from libdbus
easily, back below 150K anyhow. Deeper cuts are more work than
just turning the code off as I've done here.
* dbus/dbus-marshal-basic.c (_dbus_pack_int32): we don't need the
signed int convenience funcs
* dbus/dbus-internals.c (_dbus_verbose_real): omit when not in
verbose mode
* dbus/dbus-string-util.c, dbus/dbus-string.c: more breaking
things out of libdbus
* dbus/dbus-sysdeps.c, dbus/dbus-sysdeps-util.c: same
* dbus/dbus-hash.c: purge the TWO_STRINGS crap (well, make it
tests-enabled-only, though it should probably be deleted)
* dbus/dbus-message-util.c: same stuff
* dbus/dbus-auth-util.c: same stuff
* dbus/dbus-userdb-util.c: split out part of dbus-userdb.c
* dbus/dbus-sysdeps.c (_dbus_uid_from_string): move here to pave
way for stripping down dbus-userdb.c stuff included in libdbus.
Rename _dbus_parse_uid for consistency.
* dbus/dbus-internals.c (_dbus_real_assert): print the function
name the assertion failed in
* dbus/dbus-internals.h (_dbus_return_if_fail)
(_dbus_return_val_if_fail): assert that the name of the function
containing the check doesn't start with '_', since we only want to
use checks on public functions
* dbus/dbus-connection.c (_dbus_connection_ref_unlocked): change
checks to assertions
* dbus/dbus-marshal-header.c (_dbus_header_set_field_basic):
change checks to asserts for private function
* dbus/dbus-message.c (_dbus_message_set_serial): checks
to asserts for private function
* dbus/dbus-marshal-recursive.c (skip_one_complete_type): remove
broken assertion that was breaking make check
(_dbus_type_reader_array_is_empty): remove this rather than fix
it, was only used in assertions
* test/unused-code-gc.py: hacky script to find code that's used
only by the bus (not libdbus) or used only by tests or not used at
all. It has some false alarms, but looks like we can clean up a
lot of size from libdbus.
* dbus/dbus-sysdeps.c, dbus/dbus-sysdeps-utils.c,
dbus/Makefile.am: initially move 10K of binary size out of libdbus
* Add and fix docs according to Doxygen warnings throughout
source.
* dbus/dbus-marshal-recursive.c
(_dbus_type_reader_array_is_empty): change this to just call
array_reader_get_array_len() and make it static
* dbus/dbus-message.c (dbus_message_iter_get_element_type): rename
from get_array_type
(dbus_message_iter_init_append): rename from append_iter_init
* dbus/dbus-marshal-recursive.c
(_dbus_type_reader_get_element_type): rename from
_dbus_type_reader_get_array_type
* test/glib/test-profile.c (with_bus_server_filter): fix crash
* dbus/dbus-marshal-basic.c (_dbus_unpack_uint32): inline as macro
when DBUS_DISABLE_ASSERT
(_dbus_marshal_set_basic): be sure we align for the string length
* dbus/dbus-marshal-recursive.c (skip_one_complete_type): make
this look faster
* dbus/dbus-string.c (_dbus_string_get_const_data_len): add an
inline macro version
(_dbus_string_set_byte): provide inline macro version
* Land the new message args API and type system.
This patch is huge, but the public API change is not
really large. The set of D-BUS types has changed somewhat,
and the arg "getters" are more geared toward language bindings;
they don't make a copy, etc.
There are also some known issues. See these emails for details
on this huge patch:
http://lists.freedesktop.org/archives/dbus/2004-December/001836.htmlhttp://lists.freedesktop.org/archives/dbus/2005-January/001922.html
* dbus/dbus-marshal-*: all the new stuff
* dbus/dbus-message.c: basically rewritten
* dbus/dbus-memory.c (check_guards): with "guards" enabled, init
freed blocks to be all non-nul bytes so using freed memory is less
likely to work right
* dbus/dbus-internals.c (_dbus_test_oom_handling): add
DBUS_FAIL_MALLOC=N environment variable, so you can do
DBUS_FAIL_MALLOC=0 to skip the out-of-memory checking, or
DBUS_FAIL_MALLOC=10 to make it really, really, really slow and
thorough.
* qt/message.cpp: port to the new message args API
(operator<<): use str.utf8() rather than str.unicode()
(pretty sure this is right from the Qt docs?)
* glib/dbus-gvalue.c: port to the new message args API
* bus/dispatch.c, bus/driver.c: port to the new message args API
* dbus/dbus-string.c (_dbus_string_init_const_len): initialize the
"locked" flag to TRUE and align_offset to 0; I guess we never
looked at these anyhow, but seems cleaner.
* dbus/dbus-string.h (_DBUS_STRING_ALLOCATION_PADDING):
move allocation padding macro to this header; use it to implement
(_DBUS_STRING_STATIC): ability to declare a static string.
* dbus/dbus-message.c (_dbus_message_has_type_interface_member):
change to return TRUE if the interface is not set.
* dbus/dbus-string.[hc]: move the D-BUS specific validation stuff
to dbus-marshal-validate.[hc]
* dbus/dbus-marshal-basic.c (_dbus_type_to_string): move here from
dbus-internals.c
* dbus/Makefile.am: cut over from dbus-marshal.[hc]
to dbus-marshal-*.[hc]
* dbus/dbus-object-tree.c (_dbus_decompose_path): move this
function here from dbus-marshal.c
* mono/Makefile.am: Always build the dbus DLL with --debug.
* mono/Arguments.cs (GetDBusTypeConstructor): If the type
is an enum, get the enum's underlying type. Another mono
1.1.3 fix.
Patch from Tambet Ingo <tambet@ximian.com>
* mono/DBusType/Array.cs (Get): Get the underlying element type by
calling type.GetElementType(). The code previously depended on
broken Mono behavior, which was fixed in Mono 1.1.3.
* mono/DBusType/Dict.cs (constructor): Fix the parameters for
Activator.CreateInstance() so that the class's constructor is
called with the right parameters.
Patch from Timo Teräs <ext-timo.teras@nokia.com>
* dbus/dbus-connection.c
(_dbus_connection_queue_received_message_link): Call
_dbus_connection_remove_timeout() instead of the _locked()
variant, since it's always called from
_dbus_connection_handle_watch(), which handles the locking.
Removed the _locked() variant since it's no longer used.
* python/dbus_bindings.pyx.in: Change DBUS_ACTIVATION_REPLY_ACTIVATED
and DBUS_ACTIVATION_REPLY_ALREADY_ACTIVE to match the values in
dbus-protocol.h. Because they are defines and not enums they are not
autogenerated.
* python/dbus_bindings.pyx.in (bus_activate_service): Bind
dbus_bus_activate_service
* python/dbus.py (Bus.activate_service): activate a service on the
bus.
* test/decode-gcov.c: change to use .gcno and .gcda files, but the
file format has also changed and I haven't adapted to that yet
* Makefile.am: load .gcno files from latest gcc
* dbus/dbus-string.c (_dbus_string_insert_4_aligned)
(_dbus_string_insert_8_aligned): new functions
* dbus/dbus-string.c (_dbus_string_alloc_space): new function
* dbus/dbus-string.c (_dbus_string_validate_ascii): use ISASCII
macro
* dbus/dbus-message.c: fix a comment, and add a still-unused
not-implemented function
* dbus/dbus-marshal.h: fix comment
* dbus/dbus-internals.h (_DBUS_ISASCII): new macro
* mono/DBusType/Byte.cs, mono/DBusType/Int32.cs,
mono/DBusType/Int64.cs, mono/DBusType/UInt32.cs,
mono/DBusType/UInt64.cs: Use Enum.GetUnderlyingType() instead of
Type.UnderlyingSystemType to get the actual system type
underneath. This code previously depended on the broken Mono
behavior, which was fixed in 1.1.3.
* dbus/dbus-string.h (_dbus_string_get_byte): inline when asserts
are disabled
(_dbus_string_get_const_data): inline when asserts are disabled
* dbus/dbus-message.c: record the _dbus_current_generation of
creation so we can complain if dbus_shutdown() is used improperly.
Do this only if checks are enabled.
* dbus/dbus-connection.c: ditto
* test/glib/test-profile.c: add with_bus mode to profile echoes
that go through the bus.
* test/glib/run-test.sh: add ability to run test-profile
* bus/dbus-daemon-1.1.in: fix to say that SIGHUP causes partial
config file reload.
* test/glib/test-profile.c: tweak a bit, add support for some
made-up minimal malloc overhead with plain sockets, since in
real life some sort of buffers are unavoidable thus we could
count them in the theoretical best case
* dbus/dbus-message.c (struct DBusMessage): put the locked bit and
the "char byte_order" next to each other to save 4 bytes
(dbus_message_new_empty_header): reduce preallocation, since the
message cache should achieve a similar effect
(dbus_message_cache_or_finalize, dbus_message_get_cached): add a
message cache that keeps a few DBusMessage around in a pool,
another 8% speedup or so.
* dbus/dbus-dataslot.c (_dbus_data_slot_list_clear): new function
* dbus/dbus-transport-unix.c (unix_do_iteration): if we're going
to write, without reading or blocking, try it before the poll()
and skip the poll() if nothing remains to write. This is about a
3% speedup in the echo client/server
The primary change here is to always write() once before adding
the write watch, which gives us about a 10% performance increase.
* dbus/dbus-transport-unix.c: a number of modifications to cope
with removing messages_pending
(check_write_watch): properly handle
DBUS_AUTH_STATE_WAITING_FOR_MEMORY; adapt to removal of
messages_pending stuff
(check_read_watch): properly handle WAITING_FOR_MEMORY and
AUTHENTICATED cases
(unix_handle_watch): after writing, see if the write watch can be
removed
(unix_do_iteration): assert that write_watch/read_watch are
non-NULL rather than testing that they aren't, since they
aren't allowed to be NULL. check_write_watch() at the end so
we add the watch if we did not finish writing (e.g. got EAGAIN)
* dbus/dbus-transport-protected.h: remove messages_pending call,
since it resulted in too much inefficient watch adding/removing;
instead we now require that the transport user does an iteration
after queueing outgoing messages, and after trying the first
write() we add a write watch if we got EAGAIN or exceeded our
max bytes to write per iteration setting
* dbus/dbus-string.c (_dbus_string_validate_signature): add this
function
* dbus/dbus-server-unix.c (unix_finalize): the socket name was
freed and then accessed, valgrind flagged this bug, fix it
* dbus/dbus-message.c: fix several bugs where HEADER_FIELD_LAST was taken
as the last valid field plus 1, where really it is equal to the
last valid field. Corrects some message corruption issues.
* dbus/dbus-mainloop.c: verbosity changes
* dbus/dbus-keyring.c (_dbus_keyring_new_homedir): handle OOM
instead of aborting in one of the test codepaths
* dbus/dbus-internals.c (_dbus_verbose_real): fix a bug that
caused not printing the pid ever again if a verbose was missing
the newline at the end
(_dbus_header_field_to_string): add HEADER_FIELD_SIGNATURE
* dbus/dbus-connection.c: verbosity changes;
(dbus_connection_has_messages_to_send): new function
(_dbus_connection_message_sent): no longer call transport->messages_pending
(_dbus_connection_send_preallocated_unlocked): do one iteration to
try to write() immediately, so we can avoid the write watch. This
is the core purpose of this patchset
(_dbus_connection_get_dispatch_status_unlocked): if disconnected,
dump the outgoing message queue, so nobody will get confused
trying to send them or thinking stuff is pending to be sent
* bus/test.c: verbosity changes
* bus/driver.c: verbosity/assertion changes
* bus/dispatch.c: a bunch of little tweaks to get it working again
because this patchset changes when/where you need to block.
* test/glib/test-profile.c: modify to accept a plain_sockets
argument in which case it will bench plain sockets instead of
libdbus, for comparison purposes.
* test/glib/test-profile.c (N_CLIENT_THREADS): run multiple
threads for more time, so sysprof can get a grip on it.
* dbus/dbus-string.c (_dbus_string_validate_utf8): remove
pointless variable
* test/glib/test-profile.c: fix this thing up a bit
* dbus/dbus-message.c (dbus_message_new_empty_header): increase
preallocation sizes by a fair bit; not sure if this will be an
overall performance win or not, but it does reduce reallocs.
* dbus/dbus-string.c (set_length, reallocate_for_length): ignore
the test hack that forced constant realloc if asserts are
disabled, so we can profile sanely. Sprinkle in some
_DBUS_UNLIKELY() which are probably pointless, but before I
noticed the real performance problem I put them in.
(_dbus_string_validate_utf8): micro-optimize this thing a little
bit, though callgrind says it didn't help; then special-case
ascii, which did help a lot; then be sure we detect nul bytes as
invalid, which is a bugfix.
(align_length_then_lengthen): add some more _DBUS_UNLIKELY
superstition; use memset to nul the padding instead of a manual
loop.
(_dbus_string_get_length): inline this as a
macro; it showed up in the profile because it's used for loop
tests and so forth
