Python is just too slow on some machines. Needed around twice the previous
limit on BCM2835 with Pidora 20, let's add some safety margin too.
(cherry picked from commit 81cc4d27b1)
If the VPN plugin terminated and the user started it again, then the
quit timer will still be running and it sometimes happens that the
VPN plugin will quit while the UI is asking the user for secrets.
That's not very nice, so don't do that.
Reproducer: while connect to the VPN, suspend your laptop. Then
resume it, and immediately re-start the VPN connection. Watch the
secrets dialog disappear within a very short time.
https://bugzilla.gnome.org/show_bug.cgi?id=752237
NMRemoteSettings queues a bunch of async NMRemoteConnection creations,
but wasn't doing anything to ensure that it was still alive when they
completed. Pointed out by Cosimo Cecchi.
https://bugzilla.gnome.org/show_bug.cgi?id=742138
Error: CHECKED_RETURN (CWE-252): [#def12]
NetworkManager-0.9.11.0/libnm-core/tests/test-general.c:348: check_return: Calling "nm_setting_verify" without checking return value (as is done elsewhere 37 out of 45 times).
...
(cherry picked from commit 6603e7ffde)
Libraries need to include <gi18n-lib.h>, not <gi18n.h>, so that _()
will get defined to "dgettext (GETTEXT_DOMAIN, string)" rather than
"gettext (string)" (which will use the program's default domain, which
works fine for programs in the NetworkManager tree, but not for
external users). Likewise, we need to call bindtextdomain() so that
gettext can find the translations if the library is installed in a
different prefix from the program using it (and
bind_textdomain_codeset(), so it will know the translations are in
UTF-8 even if the locale isn't).
(The fact that no one noticed this was broken before is because the
libraries didn't really start returning useful translated strings much
until 0.9.10, and none of the out-of-tree clients have been updated to
actually show those strings to users yet.)
config.h should be included from every .c file, and it should be
included before any other include. Fix that.
(As a side effect of how I did this, this also changes us to
consistently use "config.h" rather than <config.h>. To the extent that
it matters [which is not much], quotes are more correct anyway, since
we're talking about a file in our own build tree, not a system
include.)
They require a tty or X11 displays, thus are not suitable for headless runs
(such as in mock). Furthermore, they die with the tty or X11 session, which
is somehow late -- a lot of them may accumulate. Let's kill them right away.
When using a private bus connection, the service is never marked
as running when settings are initialized asynchronously. Successfully
opening a socket in NM's runtime directory should already imply
a running service, so just mark it as such (as we already do in
the synchronous path).
nm_client_activate_connection() and
nm_client_add_and_activate_connection() return when the activation has
*started*, not when it *finishes*. Clarify this a bit more in the
libnm docs, and copy that clarification to libnm-glib as well.
https://bugzilla.gnome.org/show_bug.cgi?id=736233
Calling nm_secret_agent_register() on an already registered
instance leads to a warning. Ensure to call register during
auto-registration only when needed.
https://bugzilla.gnome.org/show_bug.cgi?id=727923
Signed-off-by: Thomas Haller <thaller@redhat.com>
_nm_client_is_object_path() had been copied from
g_variant_is_object_path() with a note that it should be replaced when
we depended on GLib > 2.24, but it got missed when we actually updated
the dependency. Fix it now. (Also fix a nearby typo.)
NMObjectCache was assuming there would never be more than one object
with the same path, but since NMClient is an NMObject, it was getting
cached too, so if you created two clients and then unreffed one of
them, it's possible the wrong one could get left in the cache, causing
a crash the next time the other one called nm_object_cache_clear().
Fix this by only adding NMObjects to the cache in the codepaths where
we also check to see if the object was already in the cache.
(This also means we can remove the "except" argument to
nm_object_cache_clear(), since the NMClient won't be cached any more.)
test-nm-client.c and test-remote-settings-client.c were using their
own assertion macros so they could kill the test service on assertion
failure. Except that some new code didn't get the memo and used the
g_assert* macros. Not to mention that sometimes the tests would crash
outside of an assertion macro.
We can make test-networkmanager-service.py notice that its parent has
crashed by opening a pipe between them and taking advantage of the
fact that the pipe will be automatically closed if the parent crashes.
So then test-networkmanager-service.py just has to watch for that, and
exit if the pipe closes.
Then that lets us drop the test_assert* macros and just use g_assert*
instead.
Use "TESTS = tests-nm-client test-remote-settings-client" rather than
overriding "check-local".
Add a script "libnm-test-launch.sh" that will handle redirecting the
test via dbus-launch if needed.
Rather than passing the path to the test service on the command line,
compile it into the test programs.
(Among other things, this makes it easier to run the test directly
from the command line.)
- register a weak references and ensure that the connection
is removed when expected.
- disconnect the vis_new_connection_cb() handler
Signed-off-by: Thomas Haller <thaller@redhat.com>
test_make_invisible() forgot to disconnect handler invis_removed_cb().
Later, during test_remove_connection(), the connection will be eventually
removed and the callback will corrupt the stack by writing to the '&done'
user data.
Signed-off-by: Thomas Haller <thaller@redhat.com>
We cannot rely on connectivity data freeing to disconnect the signal,
because the simple async result might be kept alive by external
code (for example the language runtime for JS), but we must not
call dbus_g_proxy_cancel_call() if that cancellable is cancelled or
reused afterwards.
https://bugzilla.gnome.org/show_bug.cgi?id=733915
Move libnm-glib's test-fake-nm.py and test-remote-settings-service.py
to tools/, merge them together into a single program, and fix a few
bugs (notably some missing signal emissions in the Settings service).
Although they are currently only used by libnm-glib's tests, they are
generic enough that they could be used by other code in the future
(and in particular, they will be used by libnm's tests as well).
Since previous commit dafe51e881,
the gtk-doc is no longer ignored. However, the angle brackets
add invalid XML characters to ./docs/libnm-glib/xml/nm-client.xml
This breaks the build.
make[3]: Entering directory `./NetworkManager/docs/libnm-glib'
DOC Scanning header files
DOC Introspecting gobjects
DOC Building HTML
../xml/nm-client.xml:1110: parser error : Opening and ending tag mismatch: NMActiveConnection line 1109 and para
</para></refsect2>
^
../xml/nm-client.xml:1110: parser error : Opening and ending tag mismatch: para line 1107 and refsect2
</para></refsect2>
^
../xml/nm-client.xml:1211: parser error : Opening and ending tag mismatch: refsect2 line 1104 and refsect1
</refsect1>
^
../xml/nm-client.xml:1262: parser error : Opening and ending tag mismatch: refsect1 line 1095 and refentry
</refentry>
^
../xml/nm-client.xml:1263: parser error : Premature end of data in tag refentry line 7
^
../libnm-glib-docs.xml:67: element include: XInclude error : could not load ../xml/nm-client.xml, and no fallback was found
make[3]: *** [html-build.stamp] Error 6
Signed-off-by: Thomas Haller <thaller@redhat.com>
Some libnm-glib object properties were only documented in the
GParamSpec strings, not via gtk-doc comments, so they became
undocumented when the paramspec strings went away. Fix that.
(Also fix incorrect gtk-doc syntax with several NMClient properties.)
test-remote-setting-client uses a macro:
#define test_assert(condition) \
do { \
if (!G_LIKELY (condition)) \
cleanup (); \
g_assert (condition); \
} while (0)
where cleanup() kills the fake remote-settings service and unrefs
settings. However, in many cases, "condition" would involve a test
against a connection that was owned by settings, so if the check
failed, the connection would end up getting freed by cleanup(), and so
then the second invocation of condition would result in the program
aborting on a failed check somewhere else (eg, "invalid unclassed
pointer in cast to 'NMConnection'") rather than displaying the failed
assertion that had gotten us to that point.
Fix this by not unreffing settings from cleanup(); in the normal exit
case we can just have main() unref it, and in the assertion-failed
case, we don't need to free things anyway.
nm-version.h was getting disted, making srcdir!=builddir work for
tarball builds, but not for git builds.
Also, remove "-I${top_builddir}/include" from all Makefile.ams, since
there's nothing generated in include/ any more.
