Using strtol() correctly proves to be hard.
Usually, we want to also check that the end pointer is points to the end
of the string. Othewise, we silently accept trailing garbage.
When the logging level is DEBUG or TRACE, we keep all the sysctl
values we read in a cache to log how they change. Currently there is
no limit on the size of this cache and it can take a large amount of
memory.
Implement a LRU cache where the oldest entries are deleted to make
space for new ones.
https://github.com/NetworkManager/NetworkManager/pull/294
nm_ip_route_get_prefix() and plen are guint type, hence the following
is not correct:
plen = nm_ip_route_get_prefix (route1);
r = plen - nm_ip_route_get_prefix (route2);
if (r)
return r > 0 ? 1 : -1;
Use the macro, it gets subtle cases like this right.
Fixes: b32bb36c61
This enables -Werror for meson builds on gitlab-ci and semaphore.
Not on Travis, the compiler there is too old, giving too many bogus
warnings.
This reverts commit 928d68d04a ("m4:
disable -Wmissing-braces for newer clang").
The right way is IN6_ADDR_INIT_ANY.
While at it, don't initialize multiple variables in the same line.
../src/devices/nm-device-ip-tunnel.c:153:29: error: suggest braces around initialization of subobject [-Werror,-Wmissing-braces]
struct in6_addr local6 = { 0 }, remote6 = { 0 };
^
{}
NMIPAddr contains an unnamed union. We have to either explicitly
initialize one field, or omit it.
../shared/nm-utils/nm-shared-utils.c:38:36: error: suggest braces around initialization of subobject [-Werror,-Wmissing-braces]
const NMIPAddr nm_ip_addr_zero = { 0 };
^
{}
The cast is bogus and leads to a compiler warning:
[424/583] Compiling C object src/devices/wifi/914a32e@@nm-device-plugin-wifi@sha/nm-device-iwd.c.o.
In file included from ../shared/nm-default.h:293,
from ../src/devices/wifi/nm-device-iwd.c:21:
../src/devices/wifi/nm-device-iwd.c: In function ‘nm_device_iwd_set_dbus_object’:
../src/devices/wifi/nm-device-iwd.c:2404:28: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
if (!nm_g_object_ref_set ((GObject **) &priv->dbus_obj, (GObject *) object))
../shared/nm-utils/nm-macros-internal.h:1048:13: note: in definition of macro ‘nm_g_object_ref_set’
typeof (*(pp)) *const _pp = (pp); \
^~
The test should check the behavior with "const typeof(a)" in a macro,
where "a" itself is const. For that we don't need a double const
declaration of v2.
Also, that fixes an actual compiler warning:
../src/tests/test-general.c: In function ‘test_duplicate_decl_specifier’:
../src/tests/test-general.c:1669:8: warning: duplicate ‘const’ declaration specifier [-Wduplicate-decl-specifier]
const const int v2 = 3;
^~~~~
It's not useful for us.
In file included from ../src/systemd/src/libsystemd/sd-event/sd-event.c:14:
../src/systemd/src/libsystemd/sd-event/event-source.h:195:36: error: field 'buffer' with variable sized type 'union inotify_event_buffer' not at the end of a struct or class is a GNU extension [-Werror,-Wgnu-variable-sized-type-not-at-end]
union inotify_event_buffer buffer;
^
Also, let one docker image do multiple builds. We fetch a fedora docker
image, and then install 250 MB of packages. That alone takes a lot of
time and resources. Instead of running a large number of docker images
that only do one build, let one image do several builds.
Also, install ccache. Hopefully this way we can benefit from
building the same sources multiple times.
Also note that building docs does not work currently with clang,
due to g-ir-scanner. See commit 05568860cce5332977d92b85f7c25b8ed646cd58.
g-ir-scanner does not support building with clang, due to [1], [2], [3].
It triggers
checking if /usr/bin/g-ir-scanner works... no (compiler failure -- check config.log)
configure: error: introspection enabled but can't be used
with
clang-7: error: unknown argument: '-fstack-clash-protection'
See also commit 99b92fd992, which adds this configure
check.
Honor the environment variable WITH_DOCS to allow the caller to overwrite
the automatic detection that the script does.
[1] https://bugzilla.gnome.org/show_bug.cgi?id=757934
[2] https://gitlab.gnome.org/GNOME/gobject-introspection/issues/150
[3] c14d037228
To run the tests with python3, we need python3-gobject.
Note that "contrib/fedora/REQUIRED_PACKAGES" is called by
"contrib/scripts/nm-ci-run.sh" script to install the packages
in Fedora.
Currently, default-routes cannot be added like regular static-routes
as ipv4.routes setting.
Instead, one has to configure "ipv4.gateway" and "ipv4.never-default".
That of course should be fixed, for example to configure a default-route
in different routing tables.
As it is, both nmcli's parse function and libnm's
NMSettingIPConfig:verify() functions reject default-routes.
But nmcli goes way beyond that, it also rejects all networks with
"0.0.0.0"/"::" even if their prefix length is not zero. Such routes are
not default-routes, and nmcli has no business rejecting them. The
correct way for checking for a default-route is to check the prefix-length
for zero.
Drop the wrong validation in nmcli.
Note, it may still not be the best idea to add catch-all routes like
"0.0.0.0/1" and "128.0.0.0/1". It just defeats what counts as a default-route.
NM has other means (like configuring the route-metric) to handle routing
in face of multiple interfaces. But sure, whatever works for you.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/issues/114https://gitlab.freedesktop.org/NetworkManager/NetworkManager/merge_requests/75
Otherwise the following fails:
$ ./contrib/fedora/rpm/build_clean.sh -g -s x.1
...
error: parse error in expression
error: /data/src/_NetworkManager/contrib/fedora/rpm/NetworkManager.20190207-165257.XOkW4i/SPECS/NetworkManager.spec:35: bad %if condition
ERROR: rpmbuild FAILED
Even with the fix, not all characters are allowed:
$ ./contrib/fedora/rpm/build_clean.sh -g -s x-1
...
error: line 112: Illegal char '-' (0x2d) in: Release: 22165.x-1.25b13e2053.fc29
ERROR: rpmbuild FAILED
We need to build libnm-core first. Especially, because libnm
sources require the "libnm-core/nm-core-enum-types.h" header
to be generated first.
Add a missing dependency.
"libnm/fake-typelib/meson.build" modifies the variable "sources",
which is defined by the outer "libnm/meson.build" file.
That is confusing. If a variable is not only used within one "meson.build"
file alone, it should have a unique name. Rename the variable to
"libnm_utils_sources".
Also avoid local variable "deps" which is only used at one place.
Just calling nm_connection_verify() is not correct. We need
nm_connection_normalize() because otherwise we miss out on places
where we have common normalization steps implemented to fix a
connection. This is also what server-side is done.
Revert the patch, as it breaks CI tests.
I wonder also whether this is the right place. There are already
several places in "clients/cli/connections.c" that call verify()
and normalize(). These places should be unified so that there is
one place where we complete the connection. And it probably should be
done as a separate step before the add_new_connection()/update_connection()
calls.
This reverts commit ca58bcca0c.
For static functions inside a module, the compiler determines on its own
whether to inline the function.
Also, "inline" was used at some places that don't immediatly look like
candidates for inlining. It was most likely a copy&paste error.
