Files
firmware/docs/building.md
T
Oleksandr Shyshatskyi 12bbd1e7dc Retire the probe profile: there is one package now
The probe profile was a bring-up aid. It set BUILD_APPLY=0, reinstalled
the stock component byte-for-byte and wrote a report to the stick, so you
could prove the update chain worked on your own machine -- the filename
glob, the DES3 key, the md5 gate, run.sh execution -- before flashing
anything that changed a file. That rehearsal has served its purpose.

With it gone there was exactly one profile left, so the whole PROFILE
indirection goes too: profiles/ is deleted and its six BUILD_* flags are
plain defaults in bin/common.sh, which also means config.env now
overrides them instead of being silently overridden. make default folds
into make build; all-profiles and RECOVERY_PROFILE are gone with it.

What is NOT gone is the undo button: flashing the stock FlashForge
package still restores every file the mod touches, and case-recovery.sh
still proves it. What is gone is the dry run.

Note the loss: PROG_MB/DATA_MB used to come out of the probe report.
The docs now point at df -h over ssh, which you only have after a
successful flash.
2026-08-23 22:15:26 +03:00

7.7 KiB

Building your own packages

You need a stock FlashForge update package for your model — this builds from it, it does not replace it. Everything else is Docker.

cp config.env.example config.env     # what to build, and what ships
$EDITOR config.env                   # point it at your stock package
make build                           # the firmware

The package lands in work/out/. make release builds both models into dist/.

Requirements

Docker. That is the whole list — the build image carries openssl, tar, xz, unzip, python3, binutils, squashfs-tools and qemu-user-static, and every target runs inside it. make shell drops you into it. LOCAL=1 make <target> runs on the host instead.

Only the test targets get the docker socket mounted through, so they can start the replica as a sibling container. A build cannot reach the docker daemon.

The pipeline

fetch-assets.sh  download Mainsail + HelixScreen into vendor/ (see below)
unpack.sh   decrypt the stock .tgz, open the software component
patch.sh    apply the mods to work/software/ and build work/modpayload/
pack.sh     regenerate md5sum.list, tar, encrypt → work/out/<Model>-anvil-<date>.tgz
verify.sh   simulate every check the printer performs, against the built file

make build is all four. Each is idempotent and safe to re-run.

Packages carry only the software component by default. The stock installer skips any component that is absent, so the kernel, the rootfs image and the MCU/board firmware are left untouched — MCU flashing is the riskiest thing in a package and there is no reason to run it for a userspace mod. FULL=1 make <target> carries all four.

One build

There is one package: the forked Klipper with toolchanger support, Mainsail/Moonraker, ssh, and HelixScreen as the UI. bin/common.sh defaults the BUILD_* flags to exactly that, and config.env can override any of them if you want a piece left out.

There used to be a profiles/ directory choosing between builds — five of them at one point (ssh, web, full, helix) climbing one feature at a time, then two (probe, default). The intermediate rungs each cost their own flash-and-verify cycle while the recovery story stayed the same at every one of them (flash the stock package), so they bought caution nobody was spending; the last of them, probe, changed nothing on the printer and only wrote a diagnostic report to the USB stick, which was a bring-up aid rather than something to keep shipping. With one build left, the selection layer went with them.

make build                        # the firmware
make release                      # both models, into dist/
MODEL=Creator5 make build         # just the non-Pro

Two kinds of flag

This distinction is the one to keep straight:

  • BUILD_* decides what goes into a package. Read at build time only, defaulted in bin/common.sh, never present on the printer. (BUILD_KLIPPER, BUILD_TOOLCHANGE, BUILD_MAINSAIL, BUILD_HELIX.)
  • MOD_* are runtime switches. They are written into /usr/data/anvil/anvil.conf, which the printer re-reads at every boot, so they can be changed over ssh afterwards and survive a mod update. (MOD_WEB, MOD_SSH, MOD_WIFI.)

Third-party pieces are downloaded, not vendored

Mainsail and the HelixScreen build for this printer are large binaries, so the repo does not carry them — and never has, so there is nothing in the git history either. versions.env pins each one by version and sha256:

MAINSAIL_VERSION="v2.18.2"
HELIX_VERSION="v0.99.115-creator5"

bin/fetch-assets.sh downloads them into vendor/ (gitignored) and refuses anything whose sha256 does not match the pin. A cached file with the right hash is never re-downloaded, so it is a no-op on every build after the first. bin/build.sh calls it for you; make vendor pre-fetches everything.

To bump a version: edit versions.env, run make vendor, and paste the sha256 it prints back into the file.

Setting MAINSAIL_ZIP or HELIX_TGZ in config.env overrides the download and builds against your own local copy. An explicit path is used as-is and is never checksummed.

If the build asks for Mainsail or HelixScreen and the file is not there, it fails. It used to skip silently, which shipped a package with an empty web root and no way to notice.

The root password

ROOT_PW_HASH in config.env is written straight into the shipped shadow file at build time. Leave it empty and the installer picks a random password on the printer instead, and writes it to anvil-password.txt on the USB stick it was flashed from.

The reason it works that way: a package is one file that many people flash, so any password baked into it would be the same on every machine. Generating it on the device is what makes it per-printer, and the stick is the one channel back to the person standing at the machine.

If the stick cannot be written, no password is set — a password nobody can read is no better than no access, and leaving a guessable one behind would be worse than saying so.

bin/patch.sh decides which of the two applies and sets MOD_PW_AUTO in the injected install block; payload/run-append.sh does the on-device half with the printer's own mkpasswd.

Two config files

config.env The build config: where your stock package and source trees are, which model, the root password hash. Some of it ships.
test.env The replica config: the factory image, the partition sizes, an optional prebuilt printer image. None of it ships, ever.

Both are gitignored; copy the .example of each. CONFIG_ENV=<path> and TEST_ENV=<path> override the locations.

Layout

The repo has two lanes, and the directory a file sits in says which one it belongs to. Nothing from the test lane can reach a printer.

Ships — everything that ends up inside the .tgz and runs on the machine:

payload/        POSIX sh, busybox ash -- runs ON the printer
  firmwareExe     the wrapper that replaces the stock binary
  init.d/         S60web, S70klipper, S80ui
  anvil.conf      runtime switches, preserved across mod updates
  run-pre.sh      backups, injected at the TOP of the stock run.sh
  run-append.sh   payload install, injected before its exit
assets/         nginx.conf, moonraker.conf

Builds it — host-side, never installed:

bin/            fetch-assets -> unpack -> patch -> pack, plus verify
versions.env    pinned Mainsail / HelixScreen versions + sha256
vendor/         where fetch-assets.sh caches them (gitignored)
config.env      your paths, the root password hash, the model
docker/         Dockerfile.build -- the container every target runs in

Tests it — never ships, and never touched by a build:

test/           the suite, the synthetic fixture, the brick lint
  printer/        the replica: binfmt, mount layout, its two Dockerfiles,
                  and the cases that run inside it on the printer's binaries
  extract-rootfs.sh       pulls the real rootfs out of the stock package
  build-printer-image.sh  bakes a prebuilt replica image
test.env        replica settings only -- factory image, partition sizes
docs/           the documentation

Two things keep the boundary from eroding: only payload/ and assets/ are ever copied into a package by patch.sh, and make verify fails if a built package contains any file byte-identical to one in bin/, test/ or docker/.

Rebuilding chelper

klippy/chelper/c_helper.so must be MIPS32r2 / nan2008 / o32 or klippy dies on import — patch.sh refuses to build a package with anything else, and make test-abi checks every shipped binary. Debian's cross-compiler cannot produce one; use the Ingenic/K1 toolchain. See the Klipper fork's own notes.