Leave a hint about core-dumps.
Also, now we have `contrib/fedora/rpm/configure-for-system.sh` script,
which can configure the build in a way similar to what we get
when doing an RPM build.
That means, inside "contrib/scripts/nm-in-container.sh" we
can just type `make install`, and it will replace the pre-installed
NetworkManager.
The main advantage is that it becomes convenient to run NetworkManager
as a systemd service. Previously, the suggested was to to install
NetworkManager inside another prefix, and run it in the terminal.
Running NetworkManager as systemd service is also necessary for NM-ci,
which restarts the NetworkManager service, and you couldn't run a test,
if you just started NetworkManager in a terminal.
Previously, you had to build a complete RPM, which takes a lot of time.
Yes, it's a large dependency. But on your outer host you
probably configured NetworkManager with QT enabled (for the
example scripts). We want to compile the same work tree inside
the container. So install qt-devel.
This will use the same option as when we do an RPM build.
The purpose is that you could type `make install` with such
a build, and it would replace the files that you'd get by installing
the NetworkManager RPMs.
Of course, you would not want to do that on your work station, but it
will be useful in a container, where we don't mind messing up the
installation.
autotools accepts both --with-$OPTION/--without-$OPTION and
--with-$OPTION=yes|no. Consistently use the latter.
The advantage is that whether it's enabled becomes an argument, so in a
script you could do
"--with-$OPTION=$VALUE"
Same for enable/disable option.
The test bcond is used to make test suite failure terminate the %check
phase. That should generally be done for distro builds.
What the distro builds shouldn't do is terminate the build on compiler
warnings, because they're fairly likely to appear on toolchain updates
without indicating actual bugs.
However, currently that's precisely what do we do now.
Let's use -Werror on debug builds instead. They are intentionally made
more prone to fail (e.g. trip more runtime assertions) because failures
can be more rapidly addressed and don't affect distro builds.
Resolve the defaults in build.sh instead of RPM macros. This looks less
terrible maintaining the same defaults as well as options to override it
upstream.
Moving it to the block that downstreams (Fedora, Red Hat) keep
customized makes it possible for them to also maintain customized
defaults here.
In particular, the downstreams should be able to enable bcond_test
at least for their production release (otherwise there's little point in
actually running tests at package build time).
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/1286
Otherwise, the path "src/c-stdaux/src/docs/conf.py" matches for
formatting. But this file is imported via git-subtree, we don't
want to format it.
Filter out such paths.
Recent gettext version can extract and merge back strings from and to
various file formats, no need for intltool anymore.
https://wiki.gnome.org/Initiatives/GnomeGoals/GettextMigrationhttps://gitlab.freedesktop.org/NetworkManager/NetworkManager/issues/133https://github.com/NetworkManager/NetworkManager/pull/303https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/96
Clarification about the use of AM_GNU_GETTEXT_REQUIRE_VERSION:
In configure.ac, specify the minimum gettext version we require, rather
than the exact one. This fixes a situation where the autoconf macros
used for gettext will be the latest available on the system (for
example, 0.20); but the copied-in Makefile.in.in will be for the exact
version specified in configure.ac (in this case, 0.19).
In that situation, the gettext build rules will error out at `make` time
with the message:
*** error: gettext infrastructure mismatch: using a Makefile.in.in
from gettext version 0.19 but the autoconf macros are from gettext
version 0.20
Avoid that by specifying a minimum version dependency rather than an
exact one. This should not cause problems as we haven’t committed any
generated or external gettext files into git, so each developer will end
up regenerating the build system for their system’s version of gettext,
as expected.
See the subsection of
https://www.gnu.org/software/gettext/manual/html_node/Version-Control-Issues.html
for more information.
Note that autoreconf currently doesn’t recognise
AM_GNU_GETTEXT_REQUIRE_VERSION, so we must continue also using
AM_GNU_GETTEXT_VERSION. autopoint will ignore the latter if the former
is present. See
https://lists.gnu.org/archive/html/autoconf-patches/2015-10/msg00000.html.
[lkundrak@v3.sk: Fixed the meson build, adjusted autogen.sh:
droped "|| exit 1", dropped call to aclocal,
dropped --copy from gtkdocize.]
Apt is run for each package separately and errors are ignored. This is
not great -- it's slow and ignores errors. Therefore we sometimes end
up without packages we need.
Let's tolerate errors only for packages that we are know can fail to
install safely.
It's just convenient to have some tools around, not only
for testing, but also for (some limited) development.
In particular, because we bind-mount .vimrc inside the container, and
if I use vim, black/clang-format is just one key binding away.
You can of course just clone NetworkManager repository and start hacking
as you like. However, there are a few things like git-notest which are
interesting to setup.
Add a script to do this.
The script is supposed to be idempotent and do nothing, unless
necessary. By default it also only prints what it would do.
NetworkManager does not support by default legacy ifcfg configuration
files anymore, this support is now provided in a separate package
(https://fedoramagazine.org/converting-networkmanager-from-ifcfg-to-keyfiles/).
ifcfg directory (/etc/sysconfig/network-scripts/) should always be present,
regardless of NetworkManager support for network scripts. This change makes the
directory always present, not only when the recently splitted ifcfg subpackage
is installed, and also make it persistent after the package removal.
Fixes: 50a6627fd7 ('rpm: split ifcfg-rh settings plugin into a separate package')
It doesn't work anymore:
$ git clone git://github.com/thom311/libnl.git
Cloning into 'libnl'...
fatal: remote error:
The unauthenticated git protocol on port 9418 is no longer supported.
Please see https://github.blog/2021-09-01-improving-git-protocol-security-github/ for more information.
On recent Fedora and RHEL we no longer have differing "rpm_version"
and "real_version". So usually "rpm_version" is just the same as
"real_version".
Update the template spec file to reflect that. For the "build_clean.sh"
script, we anyway always set them both to "__VERSION__".
With "rc1" mode, we install more than one tarballs (the one for 1.37.90
and 1.39.0). If we reach this point, we already pushed the git tags.
There is no way back.
Ignore errors at first and try to release all tarballs. Only signal the error
at the end.
"find-backports" script parses the commit messages to figure out which
patches to backport. We use "refs/notes/bugs" notes to extend the
meta data after the commit was merged. If you don't setup the
notes, the output is likely incomplete or wrong.
Yes, this is annoying. It requires you to setup the notes as described
in "CONTRIBUTING.md". Also because the "release.sh" script runs "find-backports",
so that means you cannot do releases without setting up the notes
(unless you manually disable running "find-backports"). But you really shouldn't
make a release based on incomplete information.
"Ignore-Backport:" is already in use. For the find-backports script it
has the same meaning as a "cherry picked from" line, that means, we
assume that the referenced patch was backported already and the fix
applied.
This is of course useful to make the script shut up about backports that
we don't want to do. However, it requires us to tag the old branch
with this, so that the script thinks that the patch is already there.
Imaging we have a wrong commit on "next" branch with a Fixes line. We
don't want to backport it, so we would have to tag the "old" branch with
"Ignore-Backport:". That is cumbersome.
Instead, now also support that if a commit contains a "Fixes:" line any
an "Ignore-Fixes:" for the same fixed commit, then this let's the
"Fixes:" line be ignored.
This is more for completeness, to go along "nm-code-format.sh"
script.
Usually it's very simple to run black directly (you may still do that).
However, black by default only reformats files with ".py" extension.
So to get all our python files, you'd need to know and explicitly
select them... or use this script.
Also, `black .` scans the entire source tree, and is rather slow.
This script knows which files to select and is thus faster.
By having a ".md" extension, gitlab renders a nice page instead of
showing as plain text.
Currently our README is pretty bad. Partly, because it doesn't get
shown nicely.
Rename. The file effectively was already markdown. The old file is
gone.
For this we also need to change the automake flavor to "foreign"
(See [1]).
[1] https://autotools.info/automake/options.html#automake.options.flavors
This is needed to get "/usr/share/gettext/its/polkit.its",
otherwise msgfmt will fail on Debian:
/usr/bin/msgfmt: cannot locate ITS rules for data/org.freedesktop.NetworkManager.policy.in
When we have a GError* variable on the stack, we usually want to pass
it on to function that can fail. In that case, the variable MUST be
initialized to NULL. This is an easy mistake to make.
Note that this check still can have lots of false positives, for
example, if you have a struct with an GError field. In that case, you
would need to ensure that the entire struct is initialized. Ignore the
warning then.
Also, the check misses if you declare multiple variables on one line.
But that is already discouraged by our style.
NetworkManager-wait-online is a constant source of confusion,
as it seems to delay the boot (when it's often just the messenger
or either a network problem, a NetworkManager misconfiguration
or a misconfiguration of other systemd services).
Try to clear that up with a manual page.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/1130
Recent meson versions treat unknown options as error and break now the
build. Good from them. This was an oversight.
Fixes: bbb1f5df2f ('libnm: always build libnm with JSON validation')
"-" is not allowed as D-Bus path and interface name, and discouraged as
bus name. This cause nm-priv-helper to crash, because GDBus asserts the
the object path is valid.
Replace the '-' with '_'. This way, it's consistent with
"nm_dispatcher".
Fixes: d68ab6b8f0 ('nm-sudo: rename to nm-priv-helper')
Previously (on RHEL<=8 and Fedora<=35), NetworkManager package contains
the compat scripts nm-ifup/nm-ifdown.
If initscripts package (not network-scripts!) is installed, then a RPM
trigger links them as alternatives for the ifup/ifdown commands.
One problem is that `dnf provides /usr/sbin/ifup` lists the
NetworkManager package. Which is technically true, but on RHEL9 where
initscripts is not installed by default, `dnf install NetworkManager`
does not actually create those scripts.
Solve that by moving those scripts to a new subpackage
NetworkManager-initscripts-updown. The %post script now always creates the
alternatives links, regardless whether initscripts package is installed.
Note that on RHEL8, NetworkManager package not only Obsoletes: but also
Suggests: the new package.
The name "initscripts-updown" is chosen because in the future we might
have additonal initscripts/ifcfg related subpackages to contain the
ifcfg-rh plugin (NetworkManager-initscripts-ifcfg) or ifcfg-rh migration
tools (NetworkManager-initscripts-tools).
https://bugzilla.redhat.com/show_bug.cgi?id=2022418https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/1061
There are very few places left where we would accept tabs in a source
file. Warn about that, even if it might cause some false positives.
I think this line was commented out due to a mistake.
The "Ignore-Backport" tag can be used to mark a commit that should not
be backported. Similar to the "cherry picked from" line, which indicates
that the patch was backported.
Anyway, this didn't work correctly, because we first pre-filter the
commits we search (as a performance optimization) by using `git-log` to
get a subset of the commits we want to investigate.
So if you had a commit with an "Ignore-Backport" tag, but without "cherry
picked from" line, then it wasn't found.
Fix that.
Older branches, like "nm-1-32" will always be formatted with a
different, older clang-format version. Luckily we also have on "nm-1-32"
branch the "nm-code-format-container.sh" script, so we can still
reformat the sources using the container.
However, as the name of the container was always "nm-code-format",
we would have to re-generate the container when we switch between
branches. As the container really only depends on the Fedora version
(as the clang-format version is tied to the corresponding Fedora
version), let's include the Fedora version in the name of the container.
Completely rework IP configuration in the daemon. Use NML3Cfg as layer 3
manager for the IP configuration of an interface. Use NML3ConfigData as
pieces of configuration that the various components collect and
configure. NMDevice is managing most of the IP configuration at a higher
level, that is, it starts DHCP and other IP methods. Rework the state
handling there.
This is a huge rework of how NetworkManager daemon handles IP
configuration. Some fallout is to be expected.
It appears the patch deletes many lines of code. That is not accurate, because
you also have to count the files `src/core/nm-l3*`, which were unused previously.
Co-authored-by: Beniamino Galvani <bgalvani@redhat.com>
We have multiple ways to define properties (like, GVariant based
nm_setting_option_*() or GObject based properties). For the latter,
they nowadays should all be implemented via _nm_setting_property_define_direct_*()
API.
There is a mix of new /usr/lib/systemd/libsystemd-shared-239.so
(systemd-libs rpm) and old /usr/bin/udevadm (systemd-udev rpm) on
the system at the point NetworkManager's post scriptlet is run,
what causes warning messages when updating NetworkManager's version.
This commit fixes this.
https://bugzilla.redhat.com/show_bug.cgi?id=2012123
Make checkpatch.pl identify subtree merges in "git am"-formatted
patches and reconstruct the full path names based in the subtree root.
This fixes some spurious warnings for parts of the tree that use
different coding style from what we usually do.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/989
It doesn't actually work inside the root-less container...
Well, it works as far as starting to activate, before it
fails. That is still somewhat useful. So have it there...
The majority of times when I call this script, I want it to do the reformatting,
not the check-only mode. This is also because we use git, so I start with a
clean working directory and run the reformatting code. In the best case, there
is nothing to reformat, and all is good. I seldom want to only check.
Change the default of the script.
"nm-code-format.sh" is going to change the default behavior from "-n" to
"-i", that is, from check-only to reformat. Explicitly pass "-n" where
we want it.
There was always the idea that you could pass paths and filenames
to "nm-code-format.sh" to format only a subset. However, the script
also needs to honor files that should be excluded and don't need
formatting.
Previously, what was implemented via `git ls-files -- ':(exclude)...'`
command, but git-ls-files has a bug ([1]) and might not list all files.
Refactor and do the filtering ourselves.
[1] https://www.spinics.net/lists/git/msg397982.html
The only reason is consistency. The majority of times we
do use g_snprintf(). As there are no strong reasons to
prefer one over the other, prefer the one that use use
most of the time.
NetworkManager runs as root and has lots of capabilities.
We want to reduce the attach surface by dropping capabilities,
but there is a genuine need to do certain things.
For example, we currently require dac_override capability, to open
the unix socket of ovsdb. Most users wouldn't use OVS, so we should
find a way to not require that dac_override capability. The solution
is to have a separate, D-Bus activate service (nm-sudo), which
has the capability to open and provide the file descriptor.
For authentication, we only rely on D-Bus. We watch the name owner
of NetworkManager, and only accept requests from that service. We trust
D-Bus to get it right a request from that name owner is really coming
from NetworkManager. If we couldn't trust that, how could PolicyKit
or any authentication via D-Bus work? For testing, the user can set
NM_SUDO_NO_AUTH_FOR_TESTING=1.
https://bugzilla.redhat.com/show_bug.cgi?id=1921826
g_idle_add() is discouraged, and the checkpatch.pl script warns
about it.
Sometimes there is a legitimate use of it, when you want to always
schedule an idle action (without intent to cancel or track it). That
makes more sense for g_idle_add() than it does for g_timeout_add(),
because a timeout really should be tracked and cancelled if necessary.
Add a wrapper to rename the legitimate uses. This way, we can avoid the
checkpatch.pl warnings, and can grep for the remaining illegitimate uses.
The numeric source IDs exist from a time before 2000, when there
was only one "GMainContext" singleton instance. Nowadays, the source
ID is only relative to one GMainContext, and you'd have to track
that association yourself. Als, g_source_remove() requires an additional
hash lookup, when you could simply track the GSource instance from the
start.
This API should not be used anymore. Operate on GSouce instances
direclty and use API like
nm_clear_g_source_inst()
nm_g_idle_add_source()
nm_g_idle_souce_new()
nm_g_source_attach()
g_source_attach
g_source_destroy
g_source_unref
etc.
Note that if you don't care about to ever remove a source again, like
scheduling an idle action that should not be cancelled, then
g_idle_add(callback, user_data);
is fine. It is only problematic to do something with those numeric IDs.
checkpatch.pl would also flag those uses, but these are just warnings
and in the few cases where such a warning is emitted wrongly, it's find
to ignore them.
It's an example for how to use libnm and asynchronous API.
But it's also a script I will use to test activating many
profiles in parallel.
Also add a test script that creates many veth interfaces and connection
profiles. So now you can do:
sudo NUM_DEVS=100 contrib/scripts/test-create-many-device-setup.sh setup
./examples/python/gi/nm-up-many.py c-a{1..100}
and cleanup with
nmcli connection down c-a{1..100}
sudo contrib/scripts/test-create-many-device-setup.sh cleanup
Of course, be careful to do this on your production machine.
Changing "NetworkManager.conf" is problematic, because the package management
system will detect if the user modified the file and leave .rpmnew files (or
similar).
Still, we only recently modified the file already to mention Libera.Chat.
So now is the time for more rewording.
Ups, we actually still require libuuid. Actually, we only need to
to build the example script `examples/C/glib/add-connection-gdbus.c`.
The proper solution would be to make this an optional dependency.
So far this was not yet done. Also, libuuid is really an ubiquitous
dependency on Linux, so it's not really a problem to have this build
dependency, even if it's just to build the examples.
This reverts commit c0a3947ff9.
These subpackages existed before commit 886366d0fd ('contrib/rpm:
update spec file after renaming NM plugins') (2014, before 0.9.9.95).
rpm warns about unversioned obsoletes like:
It's not recommended to have unversioned Obsoletes: Obsoletes: NetworkManager-atm
It's not recommended to have unversioned Obsoletes: Obsoletes: NetworkManager-bt
These packages are so long gone by now, let's just drop the Obsoletes.
"dhcdbd" is gone since 2007. Drop it. Also, rpm doesn't really like
unversioned obsoletes and warns:
It's not recommended to have unversioned Obsoletes: Obsoletes: dhcdbd
We really only require "iptables" as build dependency to autodetect the
path where iptables is installed. On Fedora/RHEL, this is always /usr/sbin,
so we can just as well hard code this.
Alternatively, if the autodetection is really necessary, we would also require
a build dependency on /usr/sbin/nft. That seems a waste.
"/etc/NetworkManager/VPN" was historically the place for .name files for
VPN plugins. In the meantime, those should be under "/usr/lib/NetworkManager/VPN".
Still, NetworkManager honors (and possibly watches) the directory in
/etc. Mark the directory as %ghost.
The exact effect of this is not clear to me. It seems however right to
do, and works for my testing.
Since commit a447942fc0 ('contrib/rpm: rename package
"NetworkManager-config-routing-rules" to
"NetworkManager-dispatcher-routing-rules"'), the config-routing-rules
subpackage is gone.
This way to specify the version number with a variable parameter, causes
repeated messages in rpmdiff:
INFO NetworkManager-dispatcher-routing-rules changed from Obsoletes: NetworkManager-config-routing-rules < 1:1.32.0-0.2.el8 to Obsoletes: NetworkManager-config-routing-rules < 1:1.32.0-0.3.el8 on noarch
Avoid this by hard coding the obsoleted version.
This "Conflicts" is since commit b85b8ed6fa ('contrib/rpm: let
NetworkManager-libnm and NetworkManager-glib of differing version
conflict'). This was probably fine back then, but NetworkManager-glib is
long gone.
Also, not hard coding the version number leads to rpmdiff messages like:
NEEDS INSPECTION NetworkManager-libnm changed from Conflicts: NetworkManager-glib < 1:1.32.0-0.2.el8 to Conflicts: NetworkManager-glib < 1:1.32.0-0.3.el8 on all architectures
As NetworkManager-glib is long gone, hard code the version with which
we conflict.
On Fedora 33, we get it automatically because "clang" package
has an indirect (weak) dependency for clang-tools-extra. On
Fedora 34, that is no loger the case.
We need to explicitly install it.
When supported by the D-Bus daemon, it's better to have service files
in /usr rather than in /etc. Change the path for RHEL 8.
See also commit ef8c292881 ('contrib/rpm: install D-Bus service
files to /usr if we can').
We should write our CONTRIBUTING files in markdown syntax, because
it's nice to read a plain text and gets nicely rendered.
However, if the file doesn't have a ".md" extension, gitlab's
web interface shows it as plain text file.
Rename the file.
This possibly breaks links like [1], but referring to a branch name
(and not a commit ID or a tag) is anyway fragile. Hence, I don't try
to fix that by adding a symlink or similar, because I think that just
makes it more confusing.
[1] https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/blob/master/CONTRIBUTING
"--with test" does two things:
(1) it enables "-Werror" compiler option. We always enable all
compiler warnings we care about, but this option makes all
warnings fatal.
Compiler warnings depend on compiler version and build options.
It's hard to build without any compiler warnings, in particular
for *future* compiler versions which we don't know yet. It
is desirable that a SRPM from yesterday can also be build
tomorrow.
(2) it fails build if any unit tests fail. We always run all
unit tests, but "--with test" makes it fatal. Again, we
have many unit tests that interact with the system (that is,
make system calls, like creating IP addresses or write files).
It is surprisingly hard to get them pass 100% on all the systems
we care. For example, on copr a test setup randomly fails during
ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
nm_utils_ifname_cpy(ifr.ifr_name, TEST_IFNAME);
r = ioctl(fd, TUNSETIFF, &ifr);
It's not clear why, nor is it at all clear that there is a bug
in NetworkManager. Making tests fatal basically means that a build
on copr infrastructure fails with a probability from a few percent.
Enough to be seriously annoying.
Note that on copr we actually build "--with test", because we want to catch these
issues. Likewise for our CI builds we explicitly specify "--with test".
In general, we build with various build configurations (compiler warnings)
and run unit tests on a source package many times. Starting on the
developer machine (`make check`), gitlab-ci, copr builds,
NetworkManager-ci. If you build an SRPM with such sources, a failure
of the unit tests is much more likely a glitch than an actual issue.
This is about changing the default if you build a Fedora/RHEL package.
That is with the Fedora/RHEL packages that are build in koji/brew.
Well, at least usually. In practice, we don't build frequently on non
x64_86 archs, so what I said there is less true. But the package build
is not there to replace CI/testing. The package build is there to get
a (mostly) working binary.
Note that RHEL packages anyway go through rpmdiff too, and rpmdiff
parses the build log and complain if `make check` fails.
This reverts commit e68e5c0a4c.
Currently "src/" mostly contains the source code of the daemon.
I say mostly, because that is not true, there are also the device,
settings, wwan, ppp plugins, the initrd generator, the pppd and dhcp
helper, and probably more.
Also we have source code under libnm-core/, libnm/, clients/, and
shared/ directories. That is all confusing.
We should have one "src" directory, that contains subdirectories. Those
subdirectories should contain individual parts (libraries or
applications), that possibly have dependencies on other subdirectories.
There should be a flat hierarchy of directories under src/, which
contains individual modules.
As the name "src/" is already taken, that prevents any sensible
restructuring of the code.
As a first step, move "src/" to "src/core/". This gives space to
reorganize the code better by moving individual components into "src/".
For inspiration, look at systemd's "src/" directory.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/743
