It's nicer but also doesn't annoy gcc 8: "error: ‘strncpy’ specified bound
depends on the length of the source argument [-Werror=stringop-overflow=]"
(cherry picked from commit 85c0dc4a92)
This is basically the case in the COPR build system where this
(mount -o bind,ro /proc/sys /proc/sys) is the case for reasons unknown.
(cherry picked from commit 984e9d5655)
Substrings matching the heading of valid values were allowed if not
ambiguous (e.g.: "et" for "eth"). Moreover, upper case variants were
accepted too.
Do a plain string comparison check against the valid values.
Improve also the error message: give a list of valid tx-hashes.
(cherry picked from commit fd5b3f802e)
Will be used by CI trigger to name packages that are build during testing
of a github pull request with the corresponding pull request ID.
"build_clean.sh" now supports a command line option -s|--snapshot. But the
same paramter can also be set via $NM_BUILD_SNAPSHOT environment
variable. Using the environment variable is useful to support older versions
and new versions of "build_clean.sh", so that the script can just ignore the
snapshot setting if it doesn't understand it yet.
(cherry picked from commit 86a18b2df2)
Previously we always updated resolv.conf on quit. When we are using
systemd-resolved the update is not necessary because the resolver on
127.0.0.53 would still be reachable after NM quits. Also, when NM
manages resolv.conf directly there is no need to update the file
again. Let's rewrite resolv.conf only when using dnsmasq.
https://bugzilla.redhat.com/show_bug.cgi?id=1541031
(cherry picked from commit 37eed6984b)
The code was passing the gpointer alias of the GValue, rather than the
GValue* itself. This doesn’t matter normally, but broke an experimental
patch in GLib to remove a cast from G_VALUE_TYPE.
We’ve reverted the patch in GLib (see
https://bugzilla.gnome.org/show_bug.cgi?id=793186), but this should be
fixed in NetworkManager anyway.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
https://bugzilla.gnome.org/show_bug.cgi?id=793302
(cherry picked from commit daadb8fbe9)
The RPM build should be not use silent output -- it better be possible
to check whether the correct compiler flags were present at the build
time.
(cherry picked from commit c4c8ffd113)
If IPV6CP terminates before IPCP, pppd enters the RUNNING phase and we
start IP configuration without having an IP interface set, which
triggers assertions. Instead, reimplement stage3_ip6_config_start to
be a no-op. Note that IPv6 configuration on PPP devices has never been
supported by NM.
This is a simpler version of upstream commit dd98ada33f ("ppp:
introduce SetIfindex pppd plugin D-Bus method") that doesn't require
changing the internal plugin API.
https://bugzilla.redhat.com/show_bug.cgi?id=1515829
Setting the state of NMActiveConnection results in invoking callbacks
in NMManager. Hence, it might be far-reaching. Clear
priv->queued_act_request before invoking the callbacks.
(cherry picked from commit ecf3677e57)
Don't just directly switch to DISCONNECTED state. If we are ACTIVATING
or ACTIVATED, first transition to DISCONNECTING state.
(cherry picked from commit 6d623825f6)
There is a small change in behavior:
Previously, the DEACTIVATING/DEACTIVATED states were set if and only if
the previous state was less or equal then ACTIVATED. For example,
if the state was already DEACTIVATING, it would have done nothing.
Now, nm_active_connection_set_state_fail() transitions the states
depending on the previous state. E.g. it would only set DEACTIVATING
state, if the previous state was ACTIVATING/ACTIVATED. On the other hand,
it would always progress the state to DEACTIVATED.
The new behavior makes more sense to me, although I doubt that there is
a visible difference.
(cherry picked from commit c5a97ad265)
unmanaged_to_disconnected() is supposed to mark the device as managed.
However, it may easily be unable to do so, for example if the device
is unmanaged by NM_UNMANAGED_USER_SETTINGS.
Shortly before actually enqueuing the activation request, check and
error out. Otherwise, we might hit an assertion later in
_device_activate().
(cherry picked from commit c6d0fbe7b0)
Note how recheck_assume_connection() called:
nm_exported_object_export (NM_EXPORTED_OBJECT (active));
active_connection_add (self, active);
nm_device_queue_activation (device, NM_ACT_REQUEST (active));
That differs from the order during _internal_activate_generic(), where
we would end up with:
nm_exported_object_export (NM_EXPORTED_OBJECT (active));
nm_device_queue_activation (device, NM_ACT_REQUEST (active));
active_connection_add (self, active);
It makes more sense to me to *first* add the connection, and only then
starting the activation with nm_device_queue_activation().
Also, let active_connection_add() always export the new active
connection object, if it is not already exported. All callers of
active_connection_add() ensured that the new object is already
exported.
(cherry picked from commit 6b08d2dda2)
Also, drop two redundant g_assert(). If we proceed, we will very soon afterwards
hit a SEGFAULT or a g_return_val_if_fail(), which is just as good.
(cherry picked from commit 0df3837656)
This fixes the GCC 8 build. It disables the warning conditionally so that we
get the warning back if glib gets fixed.
(cherry picked from commit 631982a796)
NM_DEVICE_OVS_INTERFACE_GET_PRIVATE() is implemented via the _NM_GET_PRIVATE()
macro. This macro uses C11's _Generic() to provide additional compiler checks
when casting from an incompatible pointer type.
As such,
NMDevice *device = ...;
NMDeviceOvsInterfacePrivate *priv;
priv = NM_DEVICE_OVS_INTERFACE_GET_PRIVATE (device);
causes a compilation error:
error: ‘_Generic’ selector of type ‘NMDevice * {aka struct _NMDevice *}’ is not compatible with any association
One workaround would be to cast the pointer first:
priv = NM_DEVICE_OVS_INTERFACE_GET_PRIVATE ((NMDeviceOvsInterface *) device);
A better fix is to mark NMDevice as a compatible pointer in _NM_GET_PRIVATE(),
which this patch does.
Previously, this went unnoticed, because due to bug "a43bf3388 build: fix configure
check for CC support of _Generic() and __auto_type", we failed to detect support
for _Generic() when compiling with -Werror. That essentially disables this check,
and NM_DEVICE_OVS_INTERFACE_GET_PRIVATE() would do a direct cast.
A workaround for this build failure might be to build with -Werror, which accidentally
results in not using _Generic().
https://bugzilla.gnome.org/show_bug.cgi?id=793183
Fixes: 8ad310f8e3
(cherry picked from commit 782578122c)
autotools' AC_LANG_PROGRAM() generates a main() function which triggers
a compiler warning (for which we fail with -WError).
conftest.c:92:1: error: function declaration isn't a prototype [-Werror=strict-prototypes]
main ()
^~~~
cc1: all warnings being treated as errors
Fixes: 557d83bf2d
(cherry picked from commit a43bf33888)
OvsInterface can postpone the stage3_ip[46]_config until the link
actually appears. It ought to restart the stage only when the link
appears, not upon further changes to it (which would trip an assertion
when starting the DHCP client while one already exists).
https://bugzilla.redhat.com/show_bug.cgi?id=1540063
(cherry picked from commit 8ad310f8e3)
We already define _GNU_SOURCE in "config.h", depending
on configure checks.
Also, we always should first include "config.h" (which means
to first include "nm-default.h").
Also, we don't need the entire <link.h>, <dlfcn.h> suffices.
(cherry picked from commit 84576ce861)
Hook the signal handlers right before the main loop. Prior to that
the default handlers are good enough and our one crashes (due to
loop being instantialized).
Also, set the return value properly to indicate a termination by a
signal.
(cherry picked from commit edf6f826b5)
On "nmcli agent all", when the polkit agent fails (while the NM agent
succeeds), the failure is not communicated until the client exits.
(cherry picked from commit 7e8a84ae10)
When pushing a warning disable with clang, always disable
-Wunknown-warning-option first -- it might be that clang wouldn't warn
of what we're trying to disable because it doesn't recognize it in the
first place. That is entierely okay.
With clang-5.0.0:
CC libnm/tests/libnm_tests_test_secret_agent-test-secret-agent.o
In file included from libnm/tests/test-secret-agent.c:29:
In file included from ./shared/nm-test-libnm-utils.h:23:
./shared/nm-utils/nm-test-utils.h:432:3: error: unknown warning group '-Wunused-but-set-variable', ignored [-Werror,-Wunknown-warning-option]
NM_PRAGMA_WARNING_DISABLE("-Wunused-but-set-variable")
^
./shared/nm-utils/nm-macros-internal.h:223:9: note: expanded from macro 'NM_PRAGMA_WARNING_DISABLE'
_Pragma(_NM_PRAGMA_WARNING_DO(warning))
^
<scratch space>:204:25: note: expanded from here
GCC diagnostic ignored "-Wunused-but-set-variable"
^
1 error generated.
(cherry picked from commit fc4552d391)
We already avoid committing the IP configuration for external devices
(see commit 60334a2893). However, we still start DHCP/IPv6-autoconf
and, especially, we change sysctl values of the device.
To be sure that no action is taken on the device, return early from
the IP configuration phase, as in the method=disabled/ignore case.
https://bugzilla.redhat.com/show_bug.cgi?id=1530288
(cherry picked from commit 22f32a16f5)
