glib/gio/gtestdbus.c

891 lines
24 KiB
C
Raw Normal View History

/* GIO testing utilities
*
* Copyright (C) 2008-2010 Red Hat, Inc.
* Copyright (C) 2012 Collabora Ltd. <http://www.collabora.co.uk/>
*
* SPDX-License-Identifier: LGPL-2.1-or-later
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General
2014-01-23 12:58:29 +01:00
* Public License along with this library; if not, see <http://www.gnu.org/licenses/>.
*
* Authors: David Zeuthen <davidz@redhat.com>
* Xavier Claessens <xavier.claessens@collabora.co.uk>
*/
#include "config.h"
#include <stdlib.h>
#include <stdio.h>
#include <errno.h>
#include <string.h>
#include <gstdio.h>
#ifdef G_OS_UNIX
#include <unistd.h>
#endif
#ifdef G_OS_WIN32
#include <io.h>
#include <fcntl.h>
#include <windows.h>
#endif
#include <glib.h>
#include "gdbusconnection.h"
#include "gdbusprivate.h"
#include "gfile.h"
#include "gioenumtypes.h"
#include "gtestdbus.h"
#include "glibintl.h"
#ifdef G_OS_UNIX
#include "glib-unix.h"
#include "glib-unixprivate.h"
#endif
/* -------------------------------------------------------------------------- */
/* Utility: Wait until object has a single ref */
typedef struct
{
GMainLoop *loop;
gboolean timed_out;
} WeakNotifyData;
static void
on_weak_notify_timeout (gpointer user_data)
{
WeakNotifyData *data = user_data;
data->timed_out = TRUE;
g_main_loop_quit (data->loop);
}
static gboolean
Revert "gtestdbus: Properly close server connections" This reverts commit baf92d09d69de0d9f9b2d0f77fc62c21fdef4da8. Closes #787 According to the original commit, this change was made because otherwise g_test_dbus_down() following a g_test_dbus_stop() hangs until it times out. The timeout being referred to is the 30 seconds which are waited by _g_object_unref_and_wait_weak_notify() for the GWeakNotify to be triggered when the last strong reference to the singleton GDBusConnection object is dropped. But the patch was not correct and the leak should have instead been fixed by having the last strong reference holder drop their reference on the GDBusConnection before calling g_test_dbus_down(). Timing out after 30 seconds is the desired behavior in the case where someone holds a reference to the singleton for that entire period. There are a few problems with this patch. First, as pointed out here[1], calling g_object_run_dispose() in the idle callback means we are causing the GWeakNotify to trigger ~immediately rather than waiting 30 seconds to give another owner a chance to unref. Second, since someone else may still hold a reference on the object being disposed, they may call methods on it after it's been disposed which can seg fault as documented here[2] and as I also saw recently in another project. It's unclear what the original leak being fixed was, but many have been fixed between 2013 and now. I ran all the unit tests under valgrind, and some do fail (some consistently and some intermittently) but none of the failures seem to only happen after this reversion commit. I also couldn't find anywhere in the valgrind output where any GDBusConnection objects are definitely being lost. [1] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214226 [2] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214237
2019-07-02 01:26:57 +02:00
unref_on_idle (gpointer object)
{
g_object_unref (object);
return FALSE;
}
static gboolean
Revert "gtestdbus: Properly close server connections" This reverts commit baf92d09d69de0d9f9b2d0f77fc62c21fdef4da8. Closes #787 According to the original commit, this change was made because otherwise g_test_dbus_down() following a g_test_dbus_stop() hangs until it times out. The timeout being referred to is the 30 seconds which are waited by _g_object_unref_and_wait_weak_notify() for the GWeakNotify to be triggered when the last strong reference to the singleton GDBusConnection object is dropped. But the patch was not correct and the leak should have instead been fixed by having the last strong reference holder drop their reference on the GDBusConnection before calling g_test_dbus_down(). Timing out after 30 seconds is the desired behavior in the case where someone holds a reference to the singleton for that entire period. There are a few problems with this patch. First, as pointed out here[1], calling g_object_run_dispose() in the idle callback means we are causing the GWeakNotify to trigger ~immediately rather than waiting 30 seconds to give another owner a chance to unref. Second, since someone else may still hold a reference on the object being disposed, they may call methods on it after it's been disposed which can seg fault as documented here[2] and as I also saw recently in another project. It's unclear what the original leak being fixed was, but many have been fixed between 2013 and now. I ran all the unit tests under valgrind, and some do fail (some consistently and some intermittently) but none of the failures seem to only happen after this reversion commit. I also couldn't find anywhere in the valgrind output where any GDBusConnection objects are definitely being lost. [1] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214226 [2] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214237
2019-07-02 01:26:57 +02:00
_g_object_unref_and_wait_weak_notify (gpointer object)
{
WeakNotifyData data;
guint timeout_id;
data.loop = g_main_loop_new (NULL, FALSE);
data.timed_out = FALSE;
g_object_weak_ref (object, (GWeakNotify) g_main_loop_quit, data.loop);
/* Drop the strong ref held by the caller in an idle callback. This is to
* make sure the mainloop is already running when weak notify happens (when
* all other strong ref holders have dropped theirs). */
Revert "gtestdbus: Properly close server connections" This reverts commit baf92d09d69de0d9f9b2d0f77fc62c21fdef4da8. Closes #787 According to the original commit, this change was made because otherwise g_test_dbus_down() following a g_test_dbus_stop() hangs until it times out. The timeout being referred to is the 30 seconds which are waited by _g_object_unref_and_wait_weak_notify() for the GWeakNotify to be triggered when the last strong reference to the singleton GDBusConnection object is dropped. But the patch was not correct and the leak should have instead been fixed by having the last strong reference holder drop their reference on the GDBusConnection before calling g_test_dbus_down(). Timing out after 30 seconds is the desired behavior in the case where someone holds a reference to the singleton for that entire period. There are a few problems with this patch. First, as pointed out here[1], calling g_object_run_dispose() in the idle callback means we are causing the GWeakNotify to trigger ~immediately rather than waiting 30 seconds to give another owner a chance to unref. Second, since someone else may still hold a reference on the object being disposed, they may call methods on it after it's been disposed which can seg fault as documented here[2] and as I also saw recently in another project. It's unclear what the original leak being fixed was, but many have been fixed between 2013 and now. I ran all the unit tests under valgrind, and some do fail (some consistently and some intermittently) but none of the failures seem to only happen after this reversion commit. I also couldn't find anywhere in the valgrind output where any GDBusConnection objects are definitely being lost. [1] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214226 [2] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214237
2019-07-02 01:26:57 +02:00
g_idle_add (unref_on_idle, object);
/* Make sure we don't block forever */
timeout_id = g_timeout_add_seconds_once (30, on_weak_notify_timeout, &data);
g_main_loop_run (data.loop);
if (data.timed_out)
{
g_warning ("Weak notify timeout, object ref_count=%d",
G_OBJECT (object)->ref_count);
}
else
{
g_source_remove (timeout_id);
}
g_main_loop_unref (data.loop);
return data.timed_out;
}
/* -------------------------------------------------------------------------- */
/* Utilities to cleanup the mess in the case unit test process crash */
#ifdef G_OS_WIN32
/* This could be interesting to expose in public API */
static void
_g_test_watcher_add_pid (GPid pid)
{
HANDLE job = NULL;
if (g_once_init_enter (&job))
{
JOBOBJECT_EXTENDED_LIMIT_INFORMATION info;
HANDLE tmp = CreateJobObjectW (NULL, NULL);
memset (&info, 0, sizeof (info));
info.BasicLimitInformation.LimitFlags = 0x2000 /* JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE */;
if (!SetInformationJobObject (tmp, JobObjectExtendedLimitInformation, &info, sizeof (info)))
g_warning ("Can't enable JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE: %s", g_win32_error_message (GetLastError()));
g_once_init_leave_pointer (&job, tmp);
}
if (!AssignProcessToJobObject(job, pid))
g_warning ("Can't assign process to job: %s", g_win32_error_message (GetLastError()));
}
static void
_g_test_watcher_remove_pid (GPid pid)
{
/* No need to unassign the process from the job object as the process
will be killed anyway */
}
#else
#define ADD_PID_FORMAT "add pid %d\n"
#define REMOVE_PID_FORMAT "remove pid %d\n"
static void
watch_parent (gint fd)
{
GIOChannel *channel;
GPollFD fds[1];
GArray *pids_to_kill;
channel = g_io_channel_unix_new (fd);
fds[0].fd = fd;
fds[0].events = G_IO_HUP | G_IO_IN;
fds[0].revents = 0;
pids_to_kill = g_array_new (FALSE, FALSE, sizeof (guint));
do
{
gint num_events;
gchar *command = NULL;
guint pid;
guint n;
GError *error = NULL;
num_events = g_poll (fds, 1, -1);
if (num_events == 0)
continue;
gtestdbus: Fix watcher crash on FreeBSD In file gio/gtestdbus.c, function watch_parent, there is a loop which waits for commands sent from the parent process and kills all processes recorded in 'pids_to_kill' array on parent process exit. The detection of parent process exit is done by calling g_poll and checking whether the returned event is G_IO_HUP. However, 'revents' is a bit mask, and we should use a bitwise-AND check instead of the equality check here. It seems to work fine on Linux, but it fails on FreeBSD because the g_poll returns both G_IO_IN and G_IO_HUP on pipe close. This means the watcher process continues waiting for commands after the parent process exit, and g_io_channel_read_line returns G_IO_STATUS_EOF with 'command' set to NULL. Then the watcher process crashes with segfault when calling sscanf because 'command' is NULL. Since the test result is already reported by the parent process as 'OK', this kind of crash is likely to be unnoticed unless someone checks dmesg messages after the test: pid 57611 (defaultvalue), uid 1001: exited on signal 11 pid 57935 (actions), uid 1001: exited on signal 11 pid 57945 (gdbus-bz627724), uid 1001: exited on signal 11 pid 57952 (gdbus-connection), uid 1001: exited on signal 11 pid 57970 (gdbus-connection-lo), uid 1001: exited on signal 11 pid 57976 (gdbus-connection-sl), uid 1001: exited on signal 11 pid 58039 (gdbus-exit-on-close), uid 1001: exited on signal 11 pid 58043 (gdbus-exit-on-close), uid 1001: exited on signal 11 pid 58047 (gdbus-exit-on-close), uid 1001: exited on signal 11 pid 58051 (gdbus-exit-on-close), uid 1001: exited on signal 11 pid 58055 (gdbus-export), uid 1001: exited on signal 11 pid 58059 (gdbus-introspection), uid 1001: exited on signal 11 pid 58065 (gdbus-names), uid 1001: exited on signal 11 pid 58071 (gdbus-proxy), uid 1001: exited on signal 11 pid 58079 (gdbus-proxy-threads), uid 1001: exited on signal 11 pid 58083 (gdbus-proxy-well-kn), uid 1001: exited on signal 11 pid 58091 (gdbus-test-codegen), uid 1001: exited on signal 11 pid 58095 (gdbus-threading), uid 1001: exited on signal 11 pid 58104 (gmenumodel), uid 1001: exited on signal 11 pid 58108 (gnotification), uid 1001: exited on signal 11 pid 58112 (gdbus-test-codegen-), uid 1001: exited on signal 11 pid 58116 (gapplication), uid 1001: exited on signal 11 pid 58132 (dbus-appinfo), uid 1001: exited on signal 11 If the watcher process crashes before killing the dbus-daemon process spawned by the parent process, the dbus-daemon process will keep running after all tests complete. Due to the implementation of 'communicate' function in Python subprocess, it causes meson to crash. 'communicate' assumes the stdout and stderr pipes are closed when the child process exits, but it is not true if processes forked by the child process doesn't exit. It causes Python subprocess 'communicate' function to block on the call to poll until the timeout expires even if the test finishes in a few seconds. Meson assumes the timeout exception always means the test is still running. It calls 'communicate' again and crashes because pipes no longer exist. https://gitlab.gnome.org/Infrastructure/GitLab/issues/286 https://github.com/mesonbuild/meson/issues/3967 https://bugs.python.org/issue30154
2018-08-19 05:30:44 +02:00
if (fds[0].revents & G_IO_HUP)
{
/* Parent quit, cleanup the mess and exit */
for (n = 0; n < pids_to_kill->len; n++)
{
pid = g_array_index (pids_to_kill, guint, n);
g_printerr ("cleaning up pid %d\n", pid);
kill (pid, SIGTERM);
}
g_array_unref (pids_to_kill);
g_io_channel_shutdown (channel, FALSE, &error);
g_assert_no_error (error);
g_io_channel_unref (channel);
exit (0);
}
/* Read the command from the input */
g_io_channel_read_line (channel, &command, NULL, NULL, &error);
g_assert_no_error (error);
/* Check for known commands */
if (sscanf (command, ADD_PID_FORMAT, &pid) == 1)
{
g_array_append_val (pids_to_kill, pid);
}
else if (sscanf (command, REMOVE_PID_FORMAT, &pid) == 1)
{
for (n = 0; n < pids_to_kill->len; n++)
{
if (g_array_index (pids_to_kill, guint, n) == pid)
{
g_array_remove_index (pids_to_kill, n);
pid = 0;
break;
}
}
if (pid != 0)
{
g_warning ("unknown pid %d to remove", pid);
}
}
else
{
g_warning ("unknown command from parent '%s'", command);
}
g_free (command);
}
while (TRUE);
}
static GIOChannel *
watcher_init (void)
{
static gsize started = 0;
static GIOChannel *channel = NULL;
int errsv;
if (g_once_init_enter (&started))
{
gint pipe_fds[2];
/* fork a child to clean up when we are killed */
if (!g_unix_open_pipe_internal (pipe_fds, TRUE, FALSE))
{
errsv = errno;
g_warning ("pipe() failed: %s", g_strerror (errsv));
g_assert_not_reached ();
}
/* flush streams to avoid buffers being duplicated in the child and
* flushed by both the child and parent later
*
* FIXME: This is a workaround for the fact that watch_parent() uses
* non-async-signal-safe API. See
* https://gitlab.gnome.org/GNOME/glib/-/issues/2322#note_1034330
*/
fflush (stdout);
fflush (stderr);
switch (fork ())
{
case -1:
errsv = errno;
g_warning ("fork() failed: %s", g_strerror (errsv));
g_assert_not_reached ();
break;
case 0:
/* child */
close (pipe_fds[1]);
watch_parent (pipe_fds[0]);
break;
default:
/* parent */
close (pipe_fds[0]);
channel = g_io_channel_unix_new (pipe_fds[1]);
}
g_once_init_leave (&started, 1);
}
return channel;
}
static void
watcher_send_command (const gchar *command)
{
GIOChannel *channel;
GError *error = NULL;
GIOStatus status;
channel = watcher_init ();
do
status = g_io_channel_write_chars (channel, command, -1, NULL, &error);
while (status == G_IO_STATUS_AGAIN);
g_assert_no_error (error);
g_io_channel_flush (channel, &error);
g_assert_no_error (error);
}
/* This could be interesting to expose in public API */
static void
_g_test_watcher_add_pid (GPid pid)
{
gchar *command;
command = g_strdup_printf (ADD_PID_FORMAT, (guint) pid);
watcher_send_command (command);
g_free (command);
}
static void
_g_test_watcher_remove_pid (GPid pid)
{
gchar *command;
command = g_strdup_printf (REMOVE_PID_FORMAT, (guint) pid);
watcher_send_command (command);
g_free (command);
}
#endif
/* -------------------------------------------------------------------------- */
/* GTestDBus object implementation */
/**
* GTestDBus:
*
* A helper class for testing code which uses D-Bus without touching the users
* session bus.
*
* Note that `GTestDBus` modifies the users environment, calling
* [`setenv()`](man:setenv(3)). This is not thread-safe, so all `GTestDBus`
* calls should be completed before threads are spawned, or should have
* appropriate locking to ensure no access conflicts to environment variables
* shared between `GTestDBus` and other threads.
*
* ## Creating unit tests using `GTestDBus`
2014-02-01 16:48:36 +01:00
*
* Testing of D-Bus services can be tricky because normally we only ever run
* D-Bus services over an existing instance of the D-Bus daemon thus we
* usually dont activate D-Bus services that are not yet installed into the
* target system. The `GTestDBus` object makes this easier for us by taking care
2014-02-01 16:48:36 +01:00
* of the lower level tasks such as running a private D-Bus daemon and looking
* up uninstalled services in customizable locations, typically in your source
* code tree.
*
* The first thing you will need is a separate service description file for the
* D-Bus daemon. Typically a `services` subdirectory of your `tests` directory
* is a good place to put this file.
2014-02-01 16:48:36 +01:00
*
* The service file should list your service along with an absolute path to the
* uninstalled service executable in your source tree. Using autotools we would
* achieve this by adding a file such as `my-server.service.in` in the services
* directory and have it processed by configure.
*
* ```
* [D-BUS Service]
* Name=org.gtk.GDBus.Examples.ObjectManager
* Exec=@abs_top_builddir@/gio/tests/gdbus-example-objectmanager-server
* ```
*
2014-02-01 16:48:36 +01:00
* You will also need to indicate this service directory in your test
* fixtures, so you will need to pass the path while compiling your
* test cases. Typically this is done with autotools with an added
* preprocessor flag specified to compile your tests such as:
*
* ```
* -DTEST_SERVICES=\""$(abs_top_builddir)/tests/services"\"
* ```
*
* Once you have a service definition file which is local to your source tree,
* you can proceed to set up a GTest fixture using the `GTestDBus` scaffolding.
2014-02-01 16:48:36 +01:00
*
2014-02-02 02:41:47 +01:00
* An example of a test fixture for D-Bus services can be found
* here:
* [gdbus-test-fixture.c](https://gitlab.gnome.org/GNOME/glib/-/blob/HEAD/gio/tests/gdbus-test-fixture.c)
2014-02-02 02:41:47 +01:00
*
2014-02-01 16:48:36 +01:00
* Note that these examples only deal with isolating the D-Bus aspect of your
* service. To successfully run isolated unit tests on your service you may need
* some additional modifications to your test case fixture. For example; if your
* service uses [class@Gio.Settings] and installs a schema then it is important
* that your test service not load the schema in the ordinary installed location
* (chances are that your service and schema files are not yet installed, or
* worse; there is an older version of the schema file sitting in the install
* location).
2014-02-01 16:48:36 +01:00
*
* Most of the time we can work around these obstacles using the
* environment. Since the environment is inherited by the D-Bus daemon
* created by `GTestDBus` and then in turn inherited by any services the
* D-Bus daemon activates, using the setup routine for your fixture is
* a practical place to help sandbox your runtime environment. For the
* rather typical GSettings case we can work around this by setting
* `GSETTINGS_SCHEMA_DIR` to the in tree directory holding your schemas
* in the above `fixture_setup()` routine.
*
* The GSettings schemas need to be locally pre-compiled for this to work. This
* can be achieved by compiling the schemas locally as a step before running
* test cases, an autotools setup might do the following in the directory
* holding schemas:
2014-02-01 16:48:36 +01:00
*
* ```
* all-am:
* $(GLIB_COMPILE_SCHEMAS) .
*
* CLEANFILES += gschemas.compiled
* ```
*
* Since: 2.34
*/
typedef struct _GTestDBusClass GTestDBusClass;
typedef struct _GTestDBusPrivate GTestDBusPrivate;
struct _GTestDBus {
GObject parent;
GTestDBusPrivate *priv;
};
struct _GTestDBusClass {
GObjectClass parent_class;
};
struct _GTestDBusPrivate
{
GTestDBusFlags flags;
GPtrArray *service_dirs;
GPid bus_pid;
gchar *bus_address;
gboolean up;
};
enum
{
PROP_0,
PROP_FLAGS,
};
G_DEFINE_TYPE_WITH_PRIVATE (GTestDBus, g_test_dbus, G_TYPE_OBJECT)
static void
g_test_dbus_init (GTestDBus *self)
{
self->priv = g_test_dbus_get_instance_private (self);
self->priv->service_dirs = g_ptr_array_new_with_free_func (g_free);
}
static void
g_test_dbus_dispose (GObject *object)
{
GTestDBus *self = (GTestDBus *) object;
if (self->priv->up)
g_test_dbus_down (self);
G_OBJECT_CLASS (g_test_dbus_parent_class)->dispose (object);
}
static void
g_test_dbus_finalize (GObject *object)
{
GTestDBus *self = (GTestDBus *) object;
g_ptr_array_unref (self->priv->service_dirs);
g_free (self->priv->bus_address);
G_OBJECT_CLASS (g_test_dbus_parent_class)->finalize (object);
}
static void
g_test_dbus_get_property (GObject *object,
guint property_id,
GValue *value,
GParamSpec *pspec)
{
GTestDBus *self = (GTestDBus *) object;
switch (property_id)
{
case PROP_FLAGS:
g_value_set_flags (value, g_test_dbus_get_flags (self));
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, property_id, pspec);
break;
}
}
static void
g_test_dbus_set_property (GObject *object,
guint property_id,
const GValue *value,
GParamSpec *pspec)
{
GTestDBus *self = (GTestDBus *) object;
switch (property_id)
{
case PROP_FLAGS:
self->priv->flags = g_value_get_flags (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, property_id, pspec);
break;
}
}
static void
g_test_dbus_class_init (GTestDBusClass *klass)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
object_class->dispose = g_test_dbus_dispose;
object_class->finalize = g_test_dbus_finalize;
object_class->get_property = g_test_dbus_get_property;
object_class->set_property = g_test_dbus_set_property;
/**
* GTestDBus:flags:
*
* #GTestDBusFlags specifying the behaviour of the D-Bus session.
*
* Since: 2.34
*/
g_object_class_install_property (object_class, PROP_FLAGS,
g_param_spec_flags ("flags", NULL, NULL,
G_TYPE_TEST_DBUS_FLAGS, G_TEST_DBUS_NONE,
G_PARAM_READWRITE | G_PARAM_CONSTRUCT_ONLY |
G_PARAM_STATIC_STRINGS));
}
static gchar *
write_config_file (GTestDBus *self)
{
GString *contents;
gint fd;
guint i;
GError *error = NULL;
gchar *path = NULL;
fd = g_file_open_tmp ("g-test-dbus-XXXXXX", &path, &error);
g_assert_no_error (error);
contents = g_string_new (NULL);
g_string_append (contents,
"<busconfig>\n"
" <type>session</type>\n"
#ifdef G_OS_WIN32
" <listen>nonce-tcp:</listen>\n"
#else
" <listen>unix:tmpdir=/tmp</listen>\n"
#endif
);
for (i = 0; i < self->priv->service_dirs->len; i++)
{
const gchar *dir_path = g_ptr_array_index (self->priv->service_dirs, i);
g_string_append_printf (contents,
" <servicedir>%s</servicedir>\n", dir_path);
}
g_string_append (contents,
" <policy context=\"default\">\n"
" <!-- Allow everything to be sent -->\n"
" <allow send_destination=\"*\" eavesdrop=\"true\"/>\n"
" <!-- Allow everything to be received -->\n"
" <allow eavesdrop=\"true\"/>\n"
" <!-- Allow anyone to own anything -->\n"
" <allow own=\"*\"/>\n"
" </policy>\n"
"</busconfig>\n");
close (fd);
g_file_set_contents_full (path, contents->str, contents->len,
G_FILE_SET_CONTENTS_NONE,
0600, &error);
g_assert_no_error (error);
g_string_free (contents, TRUE);
return path;
}
static gboolean
make_pipe (gint pipe_fds[2],
GError **error)
{
#if defined(G_OS_UNIX)
return g_unix_open_pipe (pipe_fds, O_CLOEXEC, error);
#elif defined(G_OS_WIN32)
if (_pipe (pipe_fds, 4096, _O_BINARY) < 0)
{
int errsv = errno;
g_set_error (error, G_SPAWN_ERROR, G_SPAWN_ERROR_FAILED,
_("Failed to create pipe for communicating with child process (%s)"),
g_strerror (errsv));
return FALSE;
}
return TRUE;
#else
g_set_error (error, G_SPAWN_ERROR, G_SPAWN_ERROR_FAILED,
_("Pipes are not supported in this platform"));
return FALSE;
#endif
}
static void
start_daemon (GTestDBus *self)
{
const gchar *argv[] = {"dbus-daemon", "--print-address", "--config-file=foo", NULL};
gint pipe_fds[2] = {-1, -1};
gchar *config_path;
gchar *config_arg;
gchar *print_address;
GIOChannel *channel;
gsize termpos;
GError *error = NULL;
if (g_getenv ("G_TEST_DBUS_DAEMON") != NULL)
2012-04-30 12:31:20 +02:00
argv[0] = (gchar *)g_getenv ("G_TEST_DBUS_DAEMON");
make_pipe (pipe_fds, &error);
g_assert_no_error (error);
print_address = g_strdup_printf ("--print-address=%d", pipe_fds[1]);
argv[1] = print_address;
g_assert_no_error (error);
/* Write config file and set its path in argv */
config_path = write_config_file (self);
config_arg = g_strdup_printf ("--config-file=%s", config_path);
argv[2] = config_arg;
/* Spawn dbus-daemon */
g_spawn_async_with_pipes_and_fds (NULL,
argv,
NULL,
/* We Need this to get the pid returned on win32 */
G_SPAWN_DO_NOT_REAP_CHILD |
G_SPAWN_SEARCH_PATH |
/* dbus-daemon will not abuse our descriptors, and
* passing this means we can use posix_spawn() for speed */
G_SPAWN_LEAVE_DESCRIPTORS_OPEN,
NULL, NULL,
-1, -1, -1,
&pipe_fds[1], &pipe_fds[1], 1,
&self->priv->bus_pid,
NULL, NULL, NULL,
&error);
g_assert_no_error (error);
_g_test_watcher_add_pid (self->priv->bus_pid);
/* Read bus address from pipe */
channel = g_io_channel_unix_new (pipe_fds[0]);
pipe_fds[0] = -1;
g_io_channel_set_close_on_unref (channel, TRUE);
g_io_channel_read_line (channel, &self->priv->bus_address, NULL,
&termpos, &error);
g_assert_no_error (error);
self->priv->bus_address[termpos] = '\0';
close (pipe_fds[1]);
pipe_fds[1] = -1;
/* start dbus-monitor */
if (g_getenv ("G_DBUS_MONITOR") != NULL)
{
gchar *command;
command = g_strdup_printf ("dbus-monitor --address %s",
self->priv->bus_address);
g_spawn_command_line_async (command, NULL);
g_free (command);
g_usleep (500 * 1000);
}
/* Cleanup */
g_io_channel_shutdown (channel, FALSE, &error);
g_assert_no_error (error);
g_io_channel_unref (channel);
/* Don't use g_file_delete since it calls into gvfs */
if (g_unlink (config_path) != 0)
g_assert_not_reached ();
g_free (print_address);
g_free (config_path);
g_free (config_arg);
}
static void
stop_daemon (GTestDBus *self)
{
#ifdef G_OS_WIN32
if (!TerminateProcess (self->priv->bus_pid, 0))
g_warning ("Can't terminate process: %s", g_win32_error_message (GetLastError()));
#else
kill (self->priv->bus_pid, SIGTERM);
#endif
_g_test_watcher_remove_pid (self->priv->bus_pid);
g_spawn_close_pid (self->priv->bus_pid);
self->priv->bus_pid = 0;
g_free (self->priv->bus_address);
self->priv->bus_address = NULL;
}
/**
* g_test_dbus_new:
* @flags: a #GTestDBusFlags
*
* Create a new #GTestDBus object.
*
* Returns: (transfer full): a new #GTestDBus.
*/
GTestDBus *
g_test_dbus_new (GTestDBusFlags flags)
{
return g_object_new (G_TYPE_TEST_DBUS,
"flags", flags,
NULL);
}
/**
* g_test_dbus_get_flags:
* @self: a #GTestDBus
*
* Get the flags of the #GTestDBus object.
2012-09-24 02:14:27 +02:00
*
* Returns: the value of #GTestDBus:flags property
*/
GTestDBusFlags
g_test_dbus_get_flags (GTestDBus *self)
{
g_return_val_if_fail (G_IS_TEST_DBUS (self), G_TEST_DBUS_NONE);
return self->priv->flags;
}
/**
* g_test_dbus_get_bus_address:
* @self: a #GTestDBus
*
* Get the address on which dbus-daemon is running. If g_test_dbus_up() has not
* been called yet, %NULL is returned. This can be used with
* g_dbus_connection_new_for_address().
*
* Returns: (nullable): the address of the bus, or %NULL.
*/
const gchar *
g_test_dbus_get_bus_address (GTestDBus *self)
{
g_return_val_if_fail (G_IS_TEST_DBUS (self), NULL);
return self->priv->bus_address;
}
/**
* g_test_dbus_add_service_dir:
* @self: a #GTestDBus
* @path: path to a directory containing .service files
*
* Add a path where dbus-daemon will look up .service files. This can't be
* called after g_test_dbus_up().
*/
void
g_test_dbus_add_service_dir (GTestDBus *self,
const gchar *path)
{
g_return_if_fail (G_IS_TEST_DBUS (self));
g_return_if_fail (self->priv->bus_address == NULL);
g_ptr_array_add (self->priv->service_dirs, g_strdup (path));
}
/**
* g_test_dbus_up:
* @self: a #GTestDBus
*
* Start a dbus-daemon instance and set DBUS_SESSION_BUS_ADDRESS. After this
* call, it is safe for unit tests to start sending messages on the session bus.
*
* If this function is called from setup callback of g_test_add(),
* g_test_dbus_down() must be called in its teardown callback.
*
* If this function is called from unit test's main(), then g_test_dbus_down()
* must be called after g_test_run().
*/
void
g_test_dbus_up (GTestDBus *self)
{
g_return_if_fail (G_IS_TEST_DBUS (self));
g_return_if_fail (self->priv->bus_address == NULL);
g_return_if_fail (!self->priv->up);
start_daemon (self);
g_test_dbus_unset ();
g_setenv ("DBUS_SESSION_BUS_ADDRESS", self->priv->bus_address, TRUE);
self->priv->up = TRUE;
}
/**
* g_test_dbus_stop:
* @self: a #GTestDBus
*
* Stop the session bus started by g_test_dbus_up().
*
* Unlike g_test_dbus_down(), this won't verify the #GDBusConnection
* singleton returned by g_bus_get() or g_bus_get_sync() is destroyed. Unit
* tests wanting to verify behaviour after the session bus has been stopped
* can use this function but should still call g_test_dbus_down() when done.
*/
void
g_test_dbus_stop (GTestDBus *self)
{
g_return_if_fail (G_IS_TEST_DBUS (self));
g_return_if_fail (self->priv->bus_address != NULL);
stop_daemon (self);
}
/**
* g_test_dbus_down:
* @self: a #GTestDBus
*
* Stop the session bus started by g_test_dbus_up().
*
* This will wait for the singleton returned by g_bus_get() or g_bus_get_sync()
* to be destroyed. This is done to ensure that the next unit test won't get a
* leaked singleton from this test.
*/
void
g_test_dbus_down (GTestDBus *self)
{
GDBusConnection *connection;
g_return_if_fail (G_IS_TEST_DBUS (self));
g_return_if_fail (self->priv->up);
connection = _g_bus_get_singleton_if_exists (G_BUS_TYPE_SESSION);
if (connection != NULL)
g_dbus_connection_set_exit_on_close (connection, FALSE);
if (self->priv->bus_address != NULL)
stop_daemon (self);
if (connection != NULL)
Revert "gtestdbus: Properly close server connections" This reverts commit baf92d09d69de0d9f9b2d0f77fc62c21fdef4da8. Closes #787 According to the original commit, this change was made because otherwise g_test_dbus_down() following a g_test_dbus_stop() hangs until it times out. The timeout being referred to is the 30 seconds which are waited by _g_object_unref_and_wait_weak_notify() for the GWeakNotify to be triggered when the last strong reference to the singleton GDBusConnection object is dropped. But the patch was not correct and the leak should have instead been fixed by having the last strong reference holder drop their reference on the GDBusConnection before calling g_test_dbus_down(). Timing out after 30 seconds is the desired behavior in the case where someone holds a reference to the singleton for that entire period. There are a few problems with this patch. First, as pointed out here[1], calling g_object_run_dispose() in the idle callback means we are causing the GWeakNotify to trigger ~immediately rather than waiting 30 seconds to give another owner a chance to unref. Second, since someone else may still hold a reference on the object being disposed, they may call methods on it after it's been disposed which can seg fault as documented here[2] and as I also saw recently in another project. It's unclear what the original leak being fixed was, but many have been fixed between 2013 and now. I ran all the unit tests under valgrind, and some do fail (some consistently and some intermittently) but none of the failures seem to only happen after this reversion commit. I also couldn't find anywhere in the valgrind output where any GDBusConnection objects are definitely being lost. [1] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214226 [2] https://gitlab.gnome.org/GNOME/glib/issues/787#note_214237
2019-07-02 01:26:57 +02:00
_g_object_unref_and_wait_weak_notify (connection);
g_test_dbus_unset ();
_g_bus_forget_singleton (G_BUS_TYPE_SESSION);
self->priv->up = FALSE;
}
/**
* g_test_dbus_unset:
*
* Unset DISPLAY and DBUS_SESSION_BUS_ADDRESS env variables to ensure the test
* won't use user's session bus.
*
* This is useful for unit tests that want to verify behaviour when no session
* bus is running. It is not necessary to call this if unit test already calls
* g_test_dbus_up() before acquiring the session bus.
*/
void
g_test_dbus_unset (void)
{
g_unsetenv ("DISPLAY");
g_unsetenv ("DBUS_SESSION_BUS_ADDRESS");
g_unsetenv ("DBUS_STARTER_ADDRESS");
g_unsetenv ("DBUS_STARTER_BUS_TYPE");
/* avoid using XDG_RUNTIME_DIR/bus */
g_unsetenv ("XDG_RUNTIME_DIR");
}