- hold a lock while accessing the tree of delayed values
- use weak reference counts with the owner object to avoid doing
g_object_notify on a dead object
- dispatch the "has-unapplied" notify to the proper main context
This commit fixes up a few race conditions in the GSettingsBackend, mostly with
respect to change notifications occuring at the same time as the last reference
count on a GSettings is dropped. With GDBus feeding us our incoming signals in
a separate thread, this is something that could easily happen.
There is currently no way (near as I can tell) to ensure that a message
has been sent when using GDBus. If we exit() before we are sure, then
it is very possible that the message isn't sent at all. This behaviour
was observed when using the GSettings commandline tool with dconf.
A quick and dirty workaround for now.
Don't define __USE_GNU, thats a glibc-internal macro, and
don't use SOL_SOCKET when not including sys/socket.h.
Maybe this file should be called glinuxcredentialsmessage.c...
Bug #618730
Rename the --schema-files option to --schema-file, since it only
accepts one file at a time. Change the GSETTINGS_CHECK_RULE to
use it that way, too. And also make it work better with !srcdir
builds.
Bugs #616731 and #616864
Without this fix, we segfault if the exported object returned an error
on all get_property() calls (in reality, this never happens).
Signed-off-by: David Zeuthen <davidz@redhat.com>
Fix an unintentional double free introduced in commit
4ad4c306c3.
This bug manifested itself when trying to complete this
$ gdbus introspect --system --dest <tab>
From valgrind running gdbus-peer test:
==20513== 32 bytes in 1 blocks are definitely lost in loss record 1 of 15
==20513== at 0x4024E4C: realloc (vg_replace_malloc.c:429)
==20513== by 0x4079BB1: g_realloc (gmem.c:174)
==20513== by 0x4099472: g_string_maybe_expand (gstring.c:396)
==20513== by 0x409A42A: g_string_insert_c (gstring.c:1050)
==20513== by 0x42169AC: g_string_append_c_inline (gstring.h:153)
==20513== by 0x421682C: _my_g_input_stream_read_line_safe (gdbusauth.c:336)
==20513== by 0x421843E: _g_dbus_auth_run_server (gdbusauth.c:1265)
==20513== by 0x4222B94: initable_init (gdbusconnection.c:1783)
==20513== by 0x41CF8D5: g_initable_init (ginitable.c:106)
==20513== by 0x41CFA8D: g_initable_new_valist (ginitable.c:219)
==20513== by 0x41CF920: g_initable_new (ginitable.c:139)
==20513== by 0x4223479: g_dbus_connection_new_sync (gdbusconnection.c:2046)
Bug #618650.
==6279== 21,615 (4,708 direct, 16,907 indirect) bytes in 169 blocks are
definitely lost in loss record 12 of 13
==6279== at 0x4024D2E: malloc (vg_replace_malloc.c:207)
==6279== by 0x4079A90: g_malloc (gmem.c:135)
==6279== by 0x4079DC8: g_malloc_n (gmem.c:252)
==6279== by 0x4097E66: g_strsplit (gstrfuncs.c:2434)
==6279== by 0x42169A2: g_dbus_address_get_stream_sync
(gdbusaddress.c:875)
Bug #618622.
Free the bus address after creating the singleton.
==26308== 39,736 (10,517 direct, 29,219 indirect) bytes in 388 blocks
are definitely lost in loss record 14 of 15
==26308== at 0x4024D2E: malloc (vg_replace_malloc.c:207)
==26308== by 0x4079A90: g_malloc (gmem.c:135)
==26308== by 0x4079DC8: g_malloc_n (gmem.c:252)
==26308== by 0x4095607: g_strdup (gstrfuncs.c:102)
==26308== by 0x4216B9A: g_dbus_address_get_for_bus_sync
(gdbusaddress.c:961)
==26308== by 0x422A7AE: get_uninitialized_connection
(gdbusconnection.c:5241)
Bug #618622.
There's no need to re-build the a{sv} array, just get it right out of
the parameters. Also avoid some string copies.
Signed-off-by: David Zeuthen <davidz@redhat.com>
This is to match g_dbus_connection_new(). This extension allows us to
extend GDBusAuthObserver to also be used in client-side authentication
in the future (right now it's only used on the server-side).
This is needed to e.g. allow encoding maybe types (once we add
G_DBUS_CAPABILITY_FLAGS_MAYBE_TYPES) if, and only if, that capability
has been negotiated with the peer (via authentication).
This uncovered a bug in name watching if the name wasn't activatable.
Also provoked the need for on_connection variants of g_bus_watch_name
(added g_bus_watch_proxy's variant as well).
Also make the gdbus-example-server include some example
annotations. The output looks like this:
$ gdbus introspect --session --dest org.gtk.GDBus.TestServer --object-path /org/gtk/GDBus/TestObject
node /org/gtk/GDBus/TestObject {
interface org.freedesktop.DBus.Properties {
methods:
Get(in s interface_name,
in s property_name,
out v value);
GetAll(in s interface_name,
out a{sv} properties);
Set(in s interface_name,
in s property_name,
in v value);
signals:
PropertiesChanged(s interface_name,
a{sv} changed_properties);
};
interface org.freedesktop.DBus.Introspectable {
methods:
Introspect(out s xml_data);
};
interface org.freedesktop.DBus.Peer {
methods:
Ping();
GetMachineId(out s machine_uuid);
};
@org.gtk.GDBus.Annotation("OnInterface")
@org.gtk.GDBus.Annotation("AlsoOnInterface")
interface org.gtk.GDBus.TestInterface {
methods:
@org.gtk.GDBus.Annotation("OnMethod")
HelloWorld(in s greeting,
out s response);
EmitSignal(@org.gtk.GDBus.Annotation.("OnArg")
in d speed_in_mph);
GimmeStdout();
signals:
@org.gtk.GDBus.Annotation("Onsignal")
VelocityChanged(d speed_in_mph,
@org.gtk.GDBus.Annotation.("OnArg_NonFirst")
s speed_as_string);
properties:
@org.gtk.GDBus.Annotation("OnProperty")
@org.gtk.GDBus.Annotation("OnAnnotation_YesThisIsCrazy")
readonly s FluxCapicitorName = 'DeLorean';
readwrite s Title = 'Back To C!';
readonly s ReadingAlwaysThrowsError;
readwrite s WritingAlwaysThrowsError = "There's no home like home";
writeonly s OnlyWritable;
readonly s Foo = 'Tick';
readonly s Bar = 'Tock';
};
};
This makes it possible to use the cached properties mechanism even if
constructing the proxy with the DO_NOT_LOAD_PROPERTIES flag.
This is useful for cases where you obtain the and track object
properties out-of-band. For example, in udisks, the plan is to have
something like this
Manager.GetObjects (out ao paths, out aa{sa{sv}} all_properties);
Manager.ObjectAdded (o path, a{sa{sv}} all_properties);
Manager.ObjectChanged (o path, a{sa{sv}} all_properties);
Manager.ObjectRemoved (o path, a{sa{sv}} all_properties);
E.g. the first GetObjects() call will return *all* data about *all*
exported objects. Further, this way a client will only need to listen
these three signals (three AddMatch) on the Manager object and it will
never need to do GetAll() etc (e.g. can use DO_NOT_LOAD_PROPERTIES).
(Of course this only works if the client is interested in all
objects... while this is true for udisks it is generally not true for
other D-Bus services).
Also use expected_interface to check for programming errors.
These are included wholesale in the docs, and the copyright
headers make them even more overwhelming. Plus, we don't have
copyright headers on examples anywhere else.
This fixes a problem with services that doesn't implement GetAll() for
one reason or another.
$ gdbus introspect --session --dest org.freedesktop.ReserveDevice1.Audio0 --object-path /org/freedesktop/ReserveDevice1/Audio0
node /org/freedesktop/ReserveDevice1/Audio0 {
interface org.freedesktop.ReserveDevice1 {
methods:
RequestRelease(in i priority,
out b result);
properties:
readonly i Priority = 0;
readonly s ApplicationName = 'PulseAudio Sound Server';
readonly s ApplicationDeviceName = 'Internal Audio Analog Stereo';
};
interface org.freedesktop.DBus.Properties {
methods:
Get(in s interface,
in s property,
out v value);
};
interface org.freedesktop.DBus.Introspectable {
methods:
Introspect(out s data);
};
};
Lots of people been suggesting this. We still use MethodInvocation /
method_invocation for handling incoming method calls so use call()
instead of invoke_method() helps to separate the client and server
facilities. Which is a good thing(tm).
gunixcredentialsmessage.h ought to live with other UNIX headers,
and the credentials are moved from dbus-specific to just GIO sources.
Also move gfiledescriptorbased.c to the UNIX sources.
Things compile and the test-suite passes. Still need to hook up
gio.symbols and docs. There are still a bunch of TODOs left in the
sources that needs to be addressed.
Signed-off-by: David Zeuthen <davidz@redhat.com>
- used in some places as a move-along-as-we-go pointer
- used in other places as a pointer to the fixed base of an array
Switch all users to the first style to avoid a crasher.
socket_strerror() was assuming all "strerror" messages are shorter
than 128 bytes, which is certainly true on Linux, but apparently not
on Windows.
https://bugzilla.gnome.org/show_bug.cgi?id=615494
The messages array was not reallocated correctly because it was using
malloc instead of realloc. Also, if the user requested messages but
none were received we would segfault. Rewrite the code to fix this
and, for better readability, use GPtrArray instead of rolling our own.
Also make the docs mention that the user need to free the returned
GSocketControlMessage objects using g_object_unref().
Clarify that *messages may be set to %NULL if there are no messages
(this will save pointless allocs of arrays).
Finally, the Win32 version didn't set messages to the expected value.
https://bugzilla.gnome.org/show_bug.cgi?id=616877
Signed-off-by: David Zeuthen <davidz@redhat.com>
There are apparently two incompatible ways of naming abstract sockets:
pad the sockaddr with 0s and use the entire thing as the name, or else
don't, and just pass a shorter length value to the relevant functions.
We previously only supported the former method. Add support for the
latter.
Also correctly handle "anonymous" unix sockaddrs (eg, the client side
of a connection, or a socketpair() socket), and add unix domain socket
support to the socket-client and socket-server test programs to make
sure this all works.
https://bugzilla.gnome.org/show_bug.cgi?id=615960
Visual Studio doesn't like slash as directory separator, so use
backslash. While at it, sort the list of files put in the project file
just for clarity.
Gschema-compile uses glob which is available on Unix only. Thus can't
run the gschema-compile test except on Unix either.
To avoid an Automake error, comment out the SOURCES and LDADD of
unix-streams which for some reason has been commented out from
TEST_PROGS.
Can't use a Makefile.am target called foo_PROGRAMS for random files
that aren't actually programs, as Automake assumes EXEEXT should be
appended to the file names.
Correspond to GUnixInputStream and GUnixOutputStream. No true async
support though. But that is how the Win32 API is, for files not
explicitly opened for so-called overlapped IO.
The API to create these streams takes Win32 HANDLEs. Not file
descriptors, because file descriptors are specific to the C library
used. The user code and GLib might be using different C libraries.
Also add a test program for the new classes, and a gio-windows-2.0.pc
file.
- if a key becomes mandatory, drop it from the changeset
- if we still get failing writes (ie: because of a race) then drop the
entire changeset and signal that it happened
