dpkg-deb prints a BARE VALUE when one control field is requested and
"<Field>: <value>" when several are. Every reader was written for one shape and
broke on the other, so no pool could be read and no pin could be matched against
real files:
- build-index.sh: kept the label, so "Package: curl" failed the name check and
the pool came out empty -- the build refused with "no .deb files";
- add-packages.sh (read_deb): rejected every genuine .deb as an invalid name;
(scan_existing): produced empty keys, so a version conflict went undetected
and the same package was copied twice;
- build-offline-repo.sh: compared a labelled value with a "package=version" pin,
which can never match.
A first attempt with `sed 's/^Field: //'` was wrong in the other direction: sed
-v prints nothing when the pattern does not match, which emptied the single-field
reads. The label is now dropped only when it is present
(awk: if the first word ends with a colon), which leaves a bare value and a
version like "1:2.3-4" untouched.
Also fixed in build-index.sh: the foreign-file guard did not list
repo-manifest.yaml, the record this script writes into the bundle itself, so
every second build stopped on its own previous output. Accepting it is what
makes a rebuild possible at all.
Tests
- The dpkg-deb stub now reproduces both shapes, the single- and the multi-field
one. It emitted only one, which is exactly why this class stayed invisible
while the suite was green.
- The rebuild case mirrors the production layout (record inside the output
directory) and is proven by mutation: removing repo-manifest.yaml from the
guard's allow-list turns it red.
The package pool was committed but the installer still stopped at "no local
package repository": it requires the compiled bundle (Packages, Packages.gz,
packages.lock.tsv, SHA256SUMS), and building that had never run.
The bundle
- 1321 packages, 884 MiB, Ubuntu 26.04 amd64, built with apt-ftparchive on the
fleet's own release so the closure matches the devices.
- built/pool/ is gitignored: those .deb files are already tracked once under
provisioning/offline-repo/packages/, and the same ~890 MB does not need a
second path. What the installer reads from the bundle is the index.
- The output directory is chmod 0755: mktemp -d created it 0700 and left it
so, which made a valid bundle unreadable to the installer's probe.
dpkg-deb output was parsed wrong in four places
- `dpkg-deb -f <file> Package Version` prints "<Field>: <value>", and the
callers kept the whole line, so the value carried its label. build-index.sh
therefore accepted no package at all and reported an empty pool;
add-packages.sh rejected every real .deb as an invalid name, missed version
conflicts, and copied a duplicate; build-offline-repo.sh never matched a pin.
- The test suite could not see any of this: its dpkg-deb stub emitted bare
values. The stub now prints the labels exactly as the real tool does, which
is what surfaced the remaining three sites.
Tests
- The secret scan covers the files we author, not apt-ftparchive's index over
third-party packages: polkit's own description reads "...asked for their
password:", which is prose in the packages, not a leak. Any secret reaching a
manifest, README or tool is still caught.
- The "test build must not touch the bundle" assertion now compares the
directory before and after instead of demanding it be empty, so a
maintainer's real build is not reported as test spill.
The field installer needs no network at all, which means the packages have to
travel with the repository. This adds the resolved pool and pins every package
to the exact version the pool carries.
Package pool
- 1312 .deb files (884 MB), resolved on Ubuntu 26.04 in a container so the
closure matches the release the fleet runs: a .deb built against another glibc
cannot be installed on the devices.
- Tracked through Git LFS (*.deb), so history stays navigable while the working
tree keeps the real files the installer reads from disk.
- Two RustDesk dependencies (libxdo3, libva-x11-2) were missing: the vendor .deb
is installed with dpkg, so nothing else pulls them in and `dpkg -i rustdesk`
would have left the package unconfigured on a device.
- Every file validated in the same release: 1312/1312 readable, 0 corrupt, 0
foreign architecture, and the pool satisfies all 24 pins (PLAN READY).
Pins
- All 24 UNPINNED sentinels replaced with the versions read from the packages
themselves (dpkg-deb), with requirements.tsv updated in the same change so the
contract has no drift.
build-index.sh: stop destroying the shipped record
- The repository record next to this script is the reviewed contract template
(state: UNBUILT, packages: []) that documents the schema. Its default record
destination pointed at that same file, so any default-path run overwrote the
template with BUILT rows -- which happened twice, once from a hand run and
once while this change was being developed.
- The record now defaults to <output>/repo-manifest.yaml, and pointing it at the
template is refused before any work starts. The template header documents this.
Tests
- The unpinned-pin and missing-pin cases read the sentinel out of the real
manifests, so they became no-ops the moment the contract was filled instead of
failing. They now set up the condition explicitly, and the fixture manifests
carry the synthetic versions their pool is built from.
- check-provisioning-static required packages/.gitkeep; the pool is what keeps
that directory in git now, so it asserts a populated pool directory instead.
Switches the operator path from bash-only to a Python console, keeps the bash
menu as a fallback, and makes the offline package set travel with the repository
so the person installing a device never downloads anything.
bfos/ (stdlib only, no textual/rich: those would need a download, which defeats
the offline goal):
- simple mode offers three choices (Kur / Durum / Çıkış); the detailed mode with
all 21 fleet operations and the local installer is reachable from the menu or
with --full, so a first-time operator is not buried in options
- gate parity with the bash console: scope read from the playbook file and
refused on mismatch, production needs a typed token plus a separate approval,
release-gated runs need their wave name typed, destructive runs need a second
approval, "all" walks wave by wave and stops at the first failure
- Turkish throughout; audit log at /var/log/blueforce-console.log (600, falls
back to the user state dir) with a single redaction choke point
- bfos/ui_gui.py adds an optional tkinter window over the same operations layer;
headless and missing-tkinter cases explain themselves in Turkish
admin/bf-bootstrap.sh: links the Python console as `bf` (primary) and keeps
`bf-menu` (fallback), reports both, and recommends `sudo bf`.
Offline packages are committable now (this was the blocking contradiction):
- .gitignore ignored `Packages`, `Packages.gz` and `packages.lock.tsv` everywhere,
so a built bundle could never be committed and the installer would always have
been told to download. provisioning/offline-repo/built/ is now allowed through
while scratch output stays untracked
- provisioning/offline-repo/{add-packages.sh,build-index.sh,repo-manifest.yaml}
plus a Turkish step-by-step README for the repository owner
- the over-broad `*token*` ignore rule is narrowed to `*[-_.]token*` so ordinary
filenames are not silently excluded while token files still are
Tests: run-all.sh now runs the Python suites too (previously it collected only
*.sh, so 212 Python assertions were silently skipped) and executes from the repo
root. 11/11 pass, exit 0.
Two test expectations were corrected rather than the code: a read-only
wave-scoped report legitimately asks for no gate token (the bash console never
did either), so that assertion now checks the command carries operation_wave and
operation_confirmed, and a new test asserts release-gated runs still require
their token. The bootstrap completion contract moved from `sudo bf-menu` to
`sudo bf`.
emirhanmamak-blueforce