freedom-maker (0.34)
[PTS] [DDPO]
COMMITS: VCS has seen 9 commits since the v0.34 tag
- Git: https://salsa.debian.org/freedombox-team/freedom-maker.git
-
- Branch: main
- Path: debian/changelog
- Repo size: 884736
- Browser: https://salsa.debian.org/freedombox-team/freedom-maker
- Last scan: 2025-10-16 06:47:02+00
- Next scan: 2025-10-25 05:13:00+00
- Open issues: 9
- CI pipeline status: success
- Debian changelog in Git:
freedom-maker (0.34) unstable; urgency=medium
[ Philippe Coval ]
* builder: Add support for Olimex A20-OLinuXino-MICRO-eMMC
[ Sunil Mohan Adapa ]
* internal: Allow builders to customize the built image
* vagrant: Build vagrant box without virtualbox
* d/control: Add dependency on systemd-resolved
* builders: i368: Drop all i386 targets
* d/tests: Drop dependency on pkg_resources (Closes: #1094668)
* d/gbp.conf: Set git-buildpackage branch to 'main'
[ James Valleroy ]
* d/tests: Add breaks-testbed restriction
* debian: Follows policy 4.7.2
-- James Valleroy <jvalleroy@mailbox.org> Mon, 21 Apr 2025 21:05:28 -0400
- This branch is 9 commits ahead of tag v0.34
- Git log:
commit 0165bf07cdaabe7d851d38d757f9ac7f62d064f3
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Fri Oct 10 22:08:57 2025 -0700
Use btrfs discard feature instead of creating zeros file
- btrfs has a mount option to set discard=sync. When this is set, during file
deletion and umount, the freed up space is 'TRIM'ed for the sake of solid state
storage. This also means that backing files for loop back devices become sparse.
So, instead of storing deleted data from files in unused blocks, the is
discarded leading to a very lean backing file.
- Extra storage was also needed as files removed during image build don't get
immediately freed up. This result in the partition running out of space while
the image is being built. With discard=sync (or perhaps recent versions of
btrfs), deleted files are cleaned up better and don't lead to disk filling up.
Tests:
- Create a 1G file with trancate --size. Create btrfs file system in it with
mkfs.btrfs. Then mount it using 'mount -o loop -o discard=sync file /mnt/'. Then
create a file inside with 'dd if=/dev/urandom of=/mnt/test bs=1M count=850'.
Then delete the file and 'umount /mnt'. The size occupied by the backing file is
very small when checked with 'du -sh' or 'ls -sh'. Without the discard mount
option, async discard is default and it will take time to free up the spaced
that was used by deleted file.
- This seems to work better than the zeros file we were using for cleaning up
the disk. With amd64 trixie image, the size is now 528638600 (with patch)
instead of 841966096 (with zeros approach).
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
commit 3cf3cfb81c26cd063a47c3e86a391eae0e8e5d38
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Thu Oct 9 10:28:19 2025 -0700
virtualbox-amd64: Use qemu-img instead of VBoxManage to create .vdi
- This removes dependency on the virtualbox package (for building
'virtualbox-amd64') which is ever more harder to install on Debian.
- There is a VM based test that still uses virtualbox to run the test but this
does not warrant a Suggests: relationship.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
commit b184426299265b760a31bf06d7115482f9f09443
Author: James Valleroy <jvalleroy@mailbox.org>
Date: Thu Oct 2 19:50:14 2025 -0400
Update stable release to trixie
Tests:
- Build an image for trixie. Backports are enabled, and freedombox
package from backports is installed.
Reviewed-by: Sunil Mohan Adapa <sunil@medhas.org>
commit 8c9d880d92f416485ccd56e9362e18047de72e68
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Mon Aug 25 08:42:34 2025 -0700
amd64, arm64: Fix issue with no passing correct arguments to grub
- When base class name is used instead of super(), the cls value passed is same
as the base and not the derived class value as expected. This results in
--target and --uefi-secure-boot arguments not being sent to grub-install. This
results in build failure on the build machine but not on development machine.
- This must be due a change that happened after testing.
Tests:
- Build amd64 image. Ensure that --target and --uefi-secure-boot arguments are
being sent to grub-install.
- Boot the built image and ensure that mokutil --sb-status shows that secure
boot is enabled. FreedomBox setup completes as expected and first wizard
finishes normally.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
commit 43d753870413cbdbd83a7d9a8304b02ad774c800
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Thu Aug 21 23:00:18 2025 -0700
passwd_in_image: Wait for nbd device to become available
Otherwise, the tool fails when trying to run partprobe. qemu-nbd does not seem
to have an option to wait for the device to become available unlike kpartx.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
Reviewed-by: James Valleroy <jvalleroy@mailbox.org>
commit 8fffb0830badc3dd61bca40485068580264351ae
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Thu Aug 21 22:59:31 2025 -0700
tests: Update virtualbox test to enable EFI booting
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
Reviewed-by: James Valleroy <jvalleroy@mailbox.org>
commit a703f647597f4e64b1f2a523422bb160a7a7512a
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Tue May 27 13:41:41 2025 -0700
amd64: Build UEFI + secure boot images instead of BIOS images
- The new image is no longer compatible with BIOS booting modes. This allows to
focus on some features such as unlocking encrypted disks with TPM. PCs that
support only BIOS booting are rare. Most 64-bit machines support EFI boot mode.
We have already dropped support for 32-bit machines.
- KVM/Qemu/Virt-manager machines need additional 'ovmf' package and
configuration modification to boot in UEFI mode. VirtualBox needs configuration
change to boot in UEFI mode.
- Drop the amd_intel.py abstraction now that support for Intel's 32bit
architecture has been dropped.
Tests:
- On a virt-manager machine, enable UEFI booting (with ovmf package) and boot
the machine.
- On a VirtualBox machine, enable UEFI booting and boot the machine.
- On a physical PC, enable UEFI booting and boot the machine. Both secure boot
enabled and secure boot disabled works.
- After booting 'mokmanager --sb-status' shows that secure boot is enabled if
BIOS/UEFI has secure booting enabled.
- While running the machine, running dpkg-reconfigure linux-image* copies
grub.cfg into /boot/efi/ partition.
- While running the machine, running dpkg-reconfigure grub-efi-signed-amd64*
copies signed binaries from /usr/lib/grub/x86_64-efi-signed/ into /boot/efi/
partition.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
Reviewed-by: James Valleroy <jvalleroy@mailbox.org>
commit 7b1c29be9bb76df5ec78a4485ab951a661746c41
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Tue May 27 13:36:34 2025 -0700
internal, library: Generalize setting flags on partitions
- Allow setting arbitrary flags on partitions. Flags are as defined by parted.
- Set 'esp' flag on ESP partition.
Tests:
- Build amd64 image (with EFI + secure boot changes) and notice that ESP
partition has the correct GUID and boot flag.
- Build raspberry64 image and notice that / partition has boot flag on it.
- Build a20-olinuxino-lime2 image and notice that /boot partition has boot flag
on it.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
Reviewed-by: James Valleroy <jvalleroy@mailbox.org>
commit 9d2e835698880b6b456e16f000a2a644b89105de
Author: Sunil Mohan Adapa <sunil@medhas.org>
Date: Tue May 27 12:50:28 2025 -0700
library: Allow passing UEFI secure boot support to grub-install
- When secure boot support is desired, signed versions of GRUB, shim, etc. are
copied to the ESP partition instead of unsigned versions with this parameter.
Tests:
- Unit tests.
Signed-off-by: Sunil Mohan Adapa <sunil@medhas.org>
Reviewed-by: James Valleroy <jvalleroy@mailbox.org>
