- with DEBUG=yes, allows us to see what is leftover in directory during builds
- with new cleaning in separate commit, there shouldn't be any leftovers _at all_
- with this, every u-boot build will be a full rebuild.
- still, building u-boot is pretty fast
- and it is better to be a bit slower and actually produce correct results
- incl some indicative fdt/extlinux/gpt info
- include full config and defconfig for each target
- mark the old .config file as legacy (it was always the last-target's config)
- fix: savedefconfig after build, not before (as target_map might change it)
- fix: some very old u-boots (2011, odroidc1) do not have `savedefconfig` at all, so make it optional
- fix: same for `.config` -- very old u-boots worked different (`boards.cfg`?)
- refactor `prepare_host_binfmt_qemu()` out of `prepare_host_noninteractive()` and into `rootfs/qemu-static.sh`
- further split into more functions and return early to avoid deep nesting
- implement force import and load of qemu-arm for non-armhf capable arm64 hosts (incl magic numbers)
- enhance `deploy_qemu_binary_to_chroot()` & `undeploy_qemu_binary_from_chroot()`;
- add 2nd param "caller" for better logging/tracking
- does sanity-check and preserve existing binary if it exists
- explicitly deploy/undeploy for the 3 cases:
- image: moved undeploy from `post_debootstrap_tweaks()` into image build proper for consistency
- rootfs: was leaving trash behind (since post_debootstrap_tweaks never ran for rootfs), now properly undeploys
- initrd: was already fine, just added caller info
- added `arch-test` host dependency
- ensure `arch-test ${ARCH}` works during prepare-host
- > tl,dr: "can build 32-bit armv7 armhf using Apple silicon; can use rootfs cache cross-arch reliably"
- this affects `rewrite-kernel-config` and `kernel-config` CLIs
- this does not actually change the .config's we have, that will happen later / gradually as configs are rewritten
- no more `.defconfig` files are produced (they were in gitignore)
In order for the cleanup to be correct for tools, we need to pass
the VMLINUX_BTF variable, which contains the real path to the vmlinux
file we just compiled.
The vmlinux file itself is not involved in cleaning,
but the Makefile checks for its presence and cleaning
is aborted if it is not found.
armbian-config uses debconf-apt-progress which is part of debconf.
libtext-iconv-perl is needed to prevent the following error, when
uninstalling a package through debconf-apt-progress:
Can't locate object method "new" via package "Text::Iconv" (perhaps you forgot to load "Text::Iconv"?) at /usr/share/perl5/Debconf/Encoding.pm line 65, <GEN2> line 2.
dpkg: error processing package nfs-common (--configure):
- we've some smelly stuff in write_uboot_platform for some families that we'd rather catch early
- implement small syntax fixes in setup_write_uboot_platform
* Add a list of BASH shell aliases and merge them with ohmyZSH aliases
* Add command armbian-upgrade to the BSP which executes update + upgrade
* Display armbian-upgrade only if there are upgrades
* Generate empty folder where will store data. This doesn't exists at early start
* Move armbian-install to /usr/bin and autoexecute sudo
- https://tracker.debian.org/news/1579223/software-properties-removed-from-testing/
- that provides add-apt-repository, but we don't use that on Debian, hopefully
- move also cpufrequtils hack (which requires release) to specific function after userspace checks and when DISTRIBUTION is set
- `software-properties-gtk` if of course also gone, so remove that for desktop's sake
The kernel option `CONFIG_MODULE_COMPRESS_NONE` was superseded by `CONFIG_MODULE_COMPRESS` in 6.12.
Adapt the kernel compilation script accordingly to force the correct kernel option.
Source: c7ff693fa2
- Fixes: 71296430fa -- which only set `SKIP_ARMBIAN_REPO=yes` when `rootfs` cli shortcut used
- GHA artifact pipelines are generic across all artifacts, and don't use the cli shortcuts
- instead, they use the `artifact` shortcut, with the `WHAT=rootfs` parameter
- so in GHA pipelines `SKIP_ARMBIAN_REPO=yes` was not taking effect
- rootfs artifact itself now enforces `SKIP_ARMBIAN_REPO` to be `yes` and makes it readonly
- also add `SKIP_ARMBIAN_REPO` to the general change-tracking for more clarity about configs and hooks etc
- add change tracking calls before/after generic artifact config adapter for even more clarity
Framework stores our base-files in rootfs cache with version that comes from trunk and when we make stable images for current or previous release, build fails with error E: Packages were downgraded and -y was used without --allow-downgrades.
* tools: shellfmt: Beautify and add board configs to formatting list
* Format the whole code using `lib/tools/shellfmt.sh`
---------
Co-authored-by: Igor <igor@armbian.com>
This temporary workaround was only needed for tinkerboard and xt-q8l-v10 version 2022.04 U-Boot. The version has since bumped to 2024.07 and builds fine without this.
- both artifacts use functions in the bsp/utils-bsp.sh file, which wasn't being hashed
- bsp-cli hashed non-existent 'packages/bsp-cli' directory, remove
- bsp-cli: be verbose when rsync'ing `packages/bsp/common`
- force sync in a few strategic spots
- utils-bsp.sh: be verbose when copying files (incl `config/optional` stuff)
- this avoids trouble building certain older versions with binman
- as it tries to `os.path.join(os.getenv('HOME'), 'bin')` and gets a `None` and dies
- naming names: `tinkerboard` & `xt-q8l-v10` (BOARDFAMILY=rockchip), which _actually use_ binman & 22.04 combo
- otherwise:
```
fsck.ext4 -a -C0 /dev/mmcblk0p1
/dev/mmcblk0p1 has unsupported feature(s): FEATURE_C12
e2fsck: Get a newer version of e2fsck!
```
- `orphan_file` can only be disabled on recent-enough e2fsprogs, thus check version and compare
Instead of having to add every new release to the list that does not support python2, inverse this if statement and only check if the build host is Debian bullseye or ubuntu jammy. Every release newer than those do not have python2. Older build hosts are unsupported.
Different build hosts have vastly different versions of setuptools and pyelftools depending on the host OS, e.g. Ubuntu 22.04 has setuptools v59 while the latest version at the time of this commit is setuptools v71.
Using Pip instead of APT to download these packages assures that all build hosts use the same version, removing some points of failures and inconsistencies.
Validates the dts/dtb file for the selected board and outputs the validation logs to the user.
This can be used when adding a new board, developing or improving a dts file. Should lead to higher quality device trees and patches overall, if used.
Will show warnings/errors if patches patch in some functionalities to a devicetree file without patching in the dt-bindings .yaml at the same time.
- `NETWORKING_STACK` can be set to `none` (in config phase, pre-extensions) to not-add any networking extensions
- keep defaulting to systemd-networkd if BUILD_MINIMAL and NetworkManager otherwise
- fix typo in extension name
- add `NETWORKING_STACK` to change-tracking
- Use resolved no matter what manages the network (networkd or NetworkManager)
- Use resolved.conf.d/ directory to set DNS as recommended by resolved itself
- In armbian-firstrun, remove config specific to mvebu64|mt7623 since this is now done by default
- Rename extensions with "net-*" prefix
- Put the extensions into their own folder
- Split off time sync packages into their own extensions to be able to be used separately
- Put their config files into directories instead of using inline `cat <<- EOF >`
- Move some other NetworkManager related stuff into the extension
- Remove unneeded steps
- Install iproute2 by default on all images (for the `ip` command)
- use Chrony with Network Manager
- use timesync with systemd-networkd
- use NetPlan with Network manager only
- move command-not-found to CLI image only
- improve firstlogin ip detection
Shellcheck errors/warnings were:
In lib/functions/rootfs/distro-agnostic.sh line 155:
display_alert "Adding to extlinux.conf" "fdtoverlays=${DEFAULT_OVERLAYS[@]}" "debug"
^--------------------^ SC2145 (error): Argument mixes string and array. Use * or separate argument.
In lib/functions/rootfs/distro-agnostic.sh line 156:
echo " fdtoverlays ${DEFAULT_OVERLAYS[@]}" >> "$SDCARD/boot/extlinux/extlinux.conf"
^--------------------^ SC2145 (error): Argument mixes string and array. Use * or separate argument.
In lib/functions/rootfs/distro-agnostic.sh line 193:
if [[ -n $DEFAULT_OVERLAYS && -f "${SDCARD}"/boot/armbianEnv.txt ]]; then
^---------------^ SC2128 (warning): Expanding an array without an index only gives the first element.
doas is not compatible with sudo flags. The codebase was checked for
sudo-specific uses of this function, but none were found, all cases were
in the form of `sudo <command>`. Replacing it with `doas <command>`
yields the same result.
doas is not compatible with sudo flags. The codebase was checked for
sudo-specific uses of this function, but none were found, all cases were
in the form of `sudo <command>`. Replacing it with `doas <command>`
yields the same result.
- in preparation for tightening the shellcheck severity level
- it needs to be able to follow all sources; dynamic ones are ignored, static ones need root-relative prefix
* reload-or-restart ssh prevents botching if sshd is started by a systemd socket
* restart ssh prevents botching if sshd is started by a systemd socket
* move sshd activation from ssh.service to ssh.socket:
- more realiable, avoids possible race condition on first boot
- supplementary to PR#6586 - and commits ffee50a8a6 and 6725032191
* Add display_alert explaining the change in SSH activation
- we had `_DEBUG=n` forced, which conflicted with `_EXPERT=y`;
- some important SBC features (like _GPIO_SYSFS) depend on _EXPERT=y
- we've plans to enable BTF/CO-RE kernels soon, so removing the non-debug enforcement makes some sense
- also, .config's will be free to determine debug config, nothing's changing in those here
- remove the `_EMBEDDED` special case handling, we can do that as well now
- add a separate hook to renable `CONFIG_GPIO_SYSFS=y`; it was a victim of `EXPERT=n` in some kernels
- this does not include rewrites of all the .configs -- those should be done in a separate batch -- either way effect is the same, hooks will always override .config's
- during rewrites, a large number of new options will show up, since `_EXPERT=y` is used as dependency for many of Kernel's experimental-ish features
- armhf was disabled for some releases and now we get spurious failures building certain artifacts (firmware)
Signed-off-by: Ricardo Pardini <ricardo@pardini.net>
Validates the dts/dtb file for the selected board and outputs the validation logs to the user.
This can be used when adding a new board, developing or improving a dts file. Should lead to higher quality device trees and patches overall, if used.
- useful for `collabora` and other experimental kernels, we don't want to have to add it to each individual board's KERNEL_TARGES one by one
- but we don't want to allow typos in BRANCH to emit very strange unrelated errors
* main-config: fix: avoid errors when BRANCH contains a dash; convert to underscore
* rockchip64_common: shellfmt, no changes
* rockchip64_common: move SERIALCON defaulting logic to a (verbose) hook for flexibility
* config: allow to build BRANCHes not listed in KERNEL_TARGET as long as the config is valid
- useful for `collabora` and other experimental kernels, we don't want to have to add it to each individual board's KERNEL_TARGES one by one
- but we don't want to allow typos in BRANCH to emit very strange unrelated errors
* extensions: mesa-oibaf extension by @monkaBlyat - mainline mesa PPA for Ubuntu
- does nothing on Debian
* extensions: (v3) amazingfated-rk35xx, now `rk-multimedia-amazingfate` - panfork-free
- simply skips if not on Jammy
- deploys Chromium + Widevine if desktop
* rockchip-rk3588: introduce `vendor-boogie-panthor` experimental BRANCH/kernel
- original: https://github.com/hbiyik/linux-rockchip.git (branch `rk-6.1-rkr1-panthor-v6`)
- I picked the commits on top of clean armbian/linux-rockchip `6.1-rkr1` as of 2024-04-01
- At https://github.com/rpardini/armbian-linux-rockchip-rk3588/tree/armbian-rk-6.1-rkr1-plus-boogie-panthor-v6
- Diff: https://github.com/armbian/linux-rockchip/compare/rk-6.1-rkr1...rpardini:armbian-linux-rockchip-rk3588:armbian-rk-6.1-rkr1-plus-boogie-panthor-v6
- rockchip-rk3588: introduce `boogie-bsp` BRANCH
- rockchip-rk3588: copy linux-rk35xx-vendor.config into linux-rk35xx-boogie-bsp.config
- rockchip-rk3588: update linux-rk35xx-boogie-bsp.config, no changes
- rockchip-rk3588: linux-rk35xx-boogie-bsp.config: `CONFIG_DRM_PANTHOR=m`
- rockchip-rk3588: linux-rk35xx-boogie-bsp.config: convert to defconfig
- rockchip-rk3588: rename to `BRANCH=vendor-boogie-panthor` for "clarity" (lol)
- rockchip-rk3588: vendor-boogie-panthor: force SERIALCON, full firmware (for blob needed for panthor) & mesa-oibaf extension
- rockchip-rk3588: vendor-boogie-panthor: enable amazingfated-rk35xx extension sans-panfork
Since `armbian_kernel_config` is "NOT for user consumption. Do NOT use this hook, this is internal to Armbian."
`custom_kernel_config` is able to overwrite previously set values.
Disable CONFIG_EXPERT as well since DEBUG_KERNEL=y is a dependency for EXPERT=y, meaning DEBUG_KERNEL would be re-enabled automatically if EXPERT is enabled
Manually forcing kernel options with 'call_extensions_kernel_config()' can introduce missing dependencies or misconfigurations. This commit introduces a function which checks the config, re-establishes its sanity if necessary and outputs any changes to the user.
This also makes `./compile.sh kernel-config` (exiting without changes) and `./compile.sh rewrite-kernel-config` ouput the same config file. Previously, this was not the case.
* Re-add functionality to create linux-libc-dev when building kernel packages.
* linux-libc-dev: append -${BRANCH}-${LINUXFAMILY} to package name and update control file
- This way we can safely include BOOT_SOC in /etc/armbian-release as done in the previous commit
- this will cause a rebuild of _all_ bsp-cli packages, even those for boards that don't define BOOT_SOC
- hopefully one day we will have BOOT_SOC ("What is the SoC?") defined for all boards
- for now, it will be empty for those that don't define it, which should be harmless
- RTW8723DS, RTW8723DU is unsupported/deprecated/obsolete by the author since kernel 6.7 in favor of RTW88, so deprecate it for kernel >=6.8
- RTL88x2cs specifically says in its repo: "USE RTW88 NOT THIS DRIVER", so deprecate in favor of RTW88, except for meson64 family
- move random RTL88x2cs hook specific for meson64 family in drivers_network.sh to their family config
Also add commit dates to make life prettier and easier
The following drivers have been updated with fixes for 6.8
- driver_rtl8811CU_rtl8821C
- driver_rtl88x2bu
- driver_rtl8811_rtl8812_rtl8814_rtl8821
The following drivers have been updated without specific 6.8 patches:
- driver_rtl8189ES (patches for 6.7, deleted two upstreamed patches)
- driver_rtl8189FS (patches for 6.7, deleted four upstreamed patches)
- it prepares git, applies patches to git, and re-writes them back from git
- same as kernel, it does git archeology for mbox-less patches, etc
- `uboot-patches-to-git` alias is also added, but my guess is rewrite is more useful
- refactor a common config function for both kernel and uboot
- example invocation: `./compile.sh BOARD=xxxx BRANCH=edge rewrite-uboot-patches`
- example: `./compile.sh BOARD=xxxxx BRANCH=edge kernel-dtb`
- outputs preprocessed DTS source for board in question to `output/`
- also outputs same preprocessed DTS source, ran through `dtc` with input and output DTS formats for "normalized" comparisions
- `inventory-boards-csv.py`: moar Python atrocities done to data; idea here is to output one line per board, with <<varies>> for fields that, well, vary (mainly across BRANCH'es)
- `inventory-boards` sets `TARGETS_FILE` to something that doesn't exist, so the `default-targets.yaml` is used (so same list for everyone, save for userpatched-boards)
- `DEBOOTSTRAP_DIR` must be set during the 1st stage, but NOT the second stage
- latest devel versions might be missing scripts for old releases; just symlink to the default one if so
- this hopefully should _end_ the debootstrap madness we've been handling, at least until we get around to mmdebstrap (Perl) one day
This fixes the readability of colorized output of compile.sh when
running on a terminal with a light background. It uses the COLORFGBG
environment variable similarly to how the ip(8) command does.
Signed-off-by: Darsey Litzenberger <dlitz@dlitz.net>
The command added to check whether the directory is git or not was
having syntax error. Fixed the same. Also it seems adding safe-directory
with local doesn't really work which actually makes sense. hence
reverting to using global instead. Its still safer than before as
we are not adding * and are adding the actual path of the directory
* using the configured volume group name
* added LVM support
* ensuring /boot never on LVM volume, created hook to setup root device
* preparing root device via extension, not assuming any particular partition for root
* using tab spacing
* using global parameter to require a boot partition
* using boot require, moving cryptroot code to extension
* adds crypt image suffix
---------
Co-authored-by: rafael <rvalle@privaz.io>
* meson-s4t7: bump u-boot to khadas-vims-u-boot-2019.01-v1.6-release
* Use khadas default bootargs as much as possible
* Add new hook to allow copying code into kernel
* meson-s4t7: legacy: Switch to 5.15 kernel
* meson-s4t7: add kernel-config for 5.15 kernel
* device tree overlays for 5.15 kernel for vim1s and vim4
* restructure packaging of bsp files for vim1s/vim4
* silence vblank warning on boot
* Remove display workaround as it doesn't work with 5.15 kernel
* Remove 5.4 kernel patches
- common.conf delegating to new `mainline_kernel_decide_version` hooks
- special case handling for v6.7-rc7
- some examples
- late-stage "use rolling branch version"
* Set dpkg vendor to Armbian in all images
We already set to Armbian, but we didn't set default link. This fixes it for both, Debian and Ubuntu.
* Adjust post install scripts to set correct link on upgrade
* Disable flash kernel and rearrage functions inorder they are executed
* Add board config for raspberry pi 5
* Change config.txt and cmdline.txt to match official rpi version
* Install Rpi related packages from Armbian repository
* Remove ddk branch aka RASPI (Ubuntu) kernel
* We still have unsupported releases, but for different reason
* Lower kernel output at booting to meet our other builds. Default is too mouch for end user
* Add more info where we import packages
* Add testing targets
---------
Co-authored-by: Igor <igor@armbian.com>
Co-authored-by: Igor Pecovnik <igor.pecovnik@gmail.com>
