Passing the hostname allows netavark to include it in DHCP lease
requests which, in an environment where DDNS is used, can cause
DNS entries to be created automatically.
* The current Hostname() function in container.go was updated to
check the new `container_name_as_hostname` option in the
CONTAINERS table of containers.conf. If set and no hostname
was configured for the container, it causes the hostname to be
set to a version of the container's name with the characters not
valid for a hostname removed. If not set (the default), the original
behavior of setting the hostname to the short container ID is
preserved.
* Because the Hostname() function can return the host's hostname
if the container isn't running in a private UTS namespace, and we'd
NEVER want to send _that_ in a DHCP request for a container, a new
function NetworkHostname() was added which functions like Hostname()
except that it will return an empty string instead of the host's
hostname if the container is not running in a private UTS namespace.
* networking_common.getNetworkOptions() now uses NetworkHostname()
to set the ContainerHostname member of the NetworkOptions structure.
That member was added to the structure in a corresponding commit in
common/libnetwork/types/network.go.
* Added test to containers_conf_test.go
Signed-off-by: George Joseph <g.devel@wxy78.net>
When running this test on a system without unqualifiedsearch registries
it will fail with a different error causing the test to fail. to avoid
that case define our own registries.conf that defines quay.io as
registry. This should make the test pass in the debian env.
Signed-off-by: Paul Holzinger <git@holzinger.dev>
If the source dir is owned by another user then the test the chown will
fail assuming we run the tests rootless. This function is only used by
the quadlet tests and for the purpose all we need is to read the files
so the simple fix is remove the chown as this should make the tests pass
on the special debian gating env.
Fixes#24802
Signed-off-by: Paul Holzinger <git@holzinger.dev>
This test a pretty much useless, it checks that a connection attempt on
the default socket fails. But this is incorrect as the socket is outside
of the test control as such it might be ready to accept connections as
thus the test can fail locally or as reported here in the debian tests.
Given that a simple connection fails does not add any value I opted to
remove it.
Fixes#24803
Signed-off-by: Paul Holzinger <git@holzinger.dev>
Found in debian testing where by default there are no unqualified search
registries installed. As such the test failed as the FIXME said. Now
there is no need for the test to assume anything.
Instead set our own config via CONTAINERS_REGISTRIES_CONF then we can
do exact matches, except that env was not read in the shell completion
code so move some code around to make it read the var in the same way as
podman login/logout.
Signed-off-by: Paul Holzinger <git@holzinger.dev>
The go module was update to 0.6.0 but podman is still installing an
older binary. This is not really a problem, but better to use the
latest version.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
This commit simplifies the systemd parser logic, and it solves an
infinite loop when using a continuation line.
Closes: https://github.com/containers/podman/issues/24810
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
For some reason, after upgrading to WiX v5, the forceReboot action
appeared before the Dism actions in the msi InstallExecuteSequence
table. As a consequence the user was asked to reboot before WSL or
Hyper-V installation and to reboot a second time after their
installation.
To avoid that the ForceReboot action field `Before=StopServices` is
replaced with `After=DismX86`
Signed-off-by: Mario Loriedo <mario.loriedo@gmail.com>
When building an installer from main branch, and using it
to upgrade the latest Podman release on Windows, a reboot
is triggered if WSL is not installed.
This is a regression caused by an update of the condition
to execute `ForceReboot`.
This commit fixes the condition and updates some defaults
to make it even more unlikely that reboot happens withtout
a specific user request for it.
It doesn't fix the v5.3.1 to v5.3.2 upgrade though. v5.3.1
has been released already and this commit doesn't avoid that it
triggers a reboot when updated.
Signed-off-by: Mario Loriedo <mario.loriedo@gmail.com>
Adding a patch to the bundle so that the update from previous version
(v5.3.1) is a minor update, not a major one. A minor update prevents the
full uninstallation of v5.3.1 and an unrequested reboot of the machine.
Signed-off-by: Mario Loriedo <mario.loriedo@gmail.com>
