| Age | Commit message (Collapse) | Author |
|---|
|
The native test really wants to run kvm (whether it uses kvm or tcg is
another question), but debvm only depends on qemu-kvm | qemu-system-any.
The latter alternative is of little use for the test and makes it fail.
|
|
|
|
While using the --prefix option is great, it's only available in trixie.
Hence, revert back to sed, this time getting the quoting right (by
avoiding backslashes using character sequences) and also apply to all
lines of the file as systemd seems to be adding users to the front.
Reported-by: Vasyl Vavrychuk
Fixes: 984a1155456d ("debvm-create: revert back to passwd for deleting the root password")
Closes: #17
|
|
|
|
Lintian: spelling-error-in-changelog
|
|
|
|
We were truncating the output file after generating it. Instead, split
the logic and create the file prior to running mmdebstrap to make
permission failures fail early.
|
|
Why?
genext2fs is slow. It has O(n^2) behaviour, which becomes annoying with
larger filesystems. It also creates an ext2 image and we have to upgrade
it to ext4 in multiple steps. Even then, the image has an inode size of
128, which is incompatible with 2038 and does not allow storing
sub-second precision time stamps.
How?
We use mkfs.ext4 from e2fsprogs (which is required anyway) instead.
Running this from within a --customize-hook does not work, because the
user namespace used for the hook often has no privileges to write to the
output image. Instead, we now ask mmdebstrap to output to a directory
and create the filesystem separately. We construct a special namespace
that has both access to the tree and to the output image and perform the
mkfs.ext4 there. As a consequence, we no longer support any mode but
unshare and since mkfs.ext4 consumes the size parameter, the size syntax
is reduced a bit.
|
|
Suggested-by: Paul Gevers <elbrus@debian.org>
|
|
|
|
When mmdebstrap fails, it kills its child processes by killing its
process group. This is harmless in many cases. For instance, when run
from an interactive shell, the shell creates a process group. When run
with user namespaces as non-root, not being root limits the possible
damage. But when run as root from autopkgtest, mmdebstrap runs in the
initial process group and takes down the entire lxc container. Using
setsid, we can isolate the container from this killpg call.
|
|
When passing an empty release to ./tests/create-and-run.sh, we now use
the host's apt sources and pinning. This is an additional gitlab-ci test
and replaces the previous autopkgtest involving unstable, which thus
becomes non-flaky.
|
|
Fixes: e2e003c4415f ("tests/create-and-run.sh: choose ports mirror automatically")
|
|
|
|
Fix a syntax error in control.
Fixes: 1c98a5b3b36f ("bin/debvm-run: qemu (>> 1:8.0) provides symlinks for qemu-system-${debarch} as well as qemu-system-any")
|
|
|
|
as well as qemu-system-any
|
|
Closes: #1036932
|
|
|
|
|
|
|
|
|
|
Thanks to Johannes Schauer for trying it on debci and seeing that it
properly tests the environment.
|
|
Since qemu-kvm is virtual, it doesn't have a version and we cannot
depend on it like that. Rather than revert back to qemu-system-misc
(which would defeat the purpose), just drop the dependency as testing
has a fixed version now.
Thanks to Jochen Sprickerhof for explaining the cause of badpkg.
|
|
|
|
Use two tests. One in --mode=root and another skippable one in
--mode=unshare. The latter will likely be skipped on Debian
infrastructure.
|
|
It's only a trigger for s390x yet, due to being a dependency of debvm
otherwise.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
We completely treat all 32bit arm as armhf now. Booting the rpi kernel
installed by debvm-create for armel will not work at all. The only sane
way to boot armel is to enable a multiarch kernel.
Refine the VMARCH guess from /bin/true using the file output for the
kernel image for the common cases of i386+amd64 and armel/armhf+arm64.
|
|
|
|
|
|
|
|
|
|
|
|
This is a workaround for https://bugs.debian.org/1029309
|
|
* ppc64el needs seabios for vgabios-stdvga.bin
* We need ipxe-qemu for -device virtio-net-pci,netdev=net0
|
|
|
|
|
|
|
|
|
|
|
|
systemd-binfmt implements what we need here and that's what most users
will have, so only recommend binfmt-support when systemd is missing.
Reported-by: Jochen Sprickerhof <git@jochen.sprickerhof.de>
|
|
We really are using 1.3.0 features already.
|
|
|
|
|
