cargo test used to be unreadable: subprocesses inherited the terminal, so
dpkg-buildpackage, apt and configure output interleaved with the harness
summary, and env_logger lines from parallel tests crossed each other.
New test_support module, compiled into test binaries only (inert stubs
otherwise) and initialized before main via .init_array:
- all log output goes to target/pkh-test-logs/<test>.log, one file per
test thread, so concurrent tests never interleave
- context-launched commands are captured line by line into the same file
(driver-level wrapper); test-code spawns use run_logged()
- a panic hook records failures and an atexit callback prints a matrix
(test name, panic location, message, log path) after the libtest
summary; tests panicking on purpose can opt out with a guard
Also fixes two test bugs found on the way:
- diff_checkbuilddeps_matrix compared dpkg-checkbuilddeps diagnostics
against English messages without pinning the locale
- run_source_build in differential tests now captures output like the
live-UI path does
Busy mirrors and CDNs routinely break bulk fetches: pooled keep-alive
connections get closed remotely ('error sending request'), downloads are
cut short (surfacing as bogus checksum mismatches), and index generations
momentarily drift from the Release file fetched moments before.
- shared client: short idle-pool timeout and TCP keepalive, and a
bounded-retry GET helper now used for index, Release, keyring and
Launchpad fetches (previously reqwest::get, which has no timeouts)
- downloads: retry the whole download, and check the content length so
truncation is reported as such instead of a checksum mismatch
- sources index: on a checksum mismatch against the Release file, retry
pinned to the exact listed generation via Debian's by-hash mechanism;
body-read errors are retried and reported per component instead of
aborting the whole lookup
Upload built source packages over SFTP with host-key verification
(Launchpad fingerprints pinned in host_keys.yml, ask-to-accept
otherwise), Launchpad account discovery (git config lp.user), and
pre-flight checks the upload queue itself never does: changes file
discovery/validation, PPA existence via the Launchpad API, target
series validity, and debian/control Section validity (sections
bundled in distro_info.yml). Upload log prevents duplicate uploads
unless --force.
The binary build exported DEB_BUILD_OPTIONS='parallel=<context nproc>
nocheck' (or the -j override) but the generated .buildinfo recomputed
the environment from host state: host core count, no nocheck, and
vendor profiles that ignored DEB_BUILD_PROFILES (a cross build recorded
no 'cross' profile). generate_binary_metadata now records the exact env
map that was exported to the build steps, and the recorded profiles
come from the exported DEB_BUILD_PROFILES when set.
Also unifies vendor parsing on one helper (the context-side copy lacked
the Origin: fallback of the source-build path).
Log retention only matched 'deb-*' logs, so source-build ('build-*')
and placeholder ('pkh-*') logs accumulated forever, and the 'keep the
newest' sort was lexicographic on names that sort by package/version
first, so arbitrary logs were kept. All three log shapes are matched
now and retention orders by the timestamp embedded in the name (mtime
fallback). Stale-lockfile pruning no longer deletes lockfiles younger
than 24h: a fresh <tarball>.lock is the mutual-exclusion signal of a
concurrent download and deleting it could corrupt the shared tarball
cache.
write_paragraph emitted a bare-space continuation line for empty lines
inside a value, which parse_paragraphs treated as a paragraph separator
and silently dropped the rest of the value; blank lines are now encoded
as ' .' like dpkg does and decoded back on read. Tab-indented
continuation lines now strip exactly one tab instead of keeping it.
Clearsigned .dsc content no longer leaks armor metadata into parsed
fields: the Hash:/Comment: header and the signature trailer are
stripped before parse_paragraphs at both .dsc parse sites.
The source-build pipeline and the binNMU metadata path had drifted into
two inline parsers with different acceptance rules: binary.rs filled
names from any line with a third column but partials only from
exactly-three-column lines, so a 4+ column Checksums line made
&partials[name] panic by map index. Both paths now share one parser
that accepts the modern 3-column and the legacy 5-column Files layout,
rejects anything else with an error naming the field and line, and all
remaining lookups go through .get() with a clear error instead of
indexing. Legacy 5-column Files md5s were previously attributed to the
section token instead of the file name.
- distro_info: malformed CSV rows are skipped with a warning, dates
that fail to parse become None, and all plain HTTP requests go
through a shared reqwest client with connect/total timeouts
- package_info: the Sources stanza iterator is iterative (a crafted
index with many blank stanzas overflowed the stack), stanzas missing
a Version are skipped, and failed series/pocket probes are summarized
in the final 'not found' error instead of being silently dropped
- pull: no double unwrap on the remote-derived artifact filename, an
empty series list is an error, and streaming downloads get a
per-request timeout
- deb/cross: dpkg-architecture output parsing skips unexpected lines
and its exit status is checked, as is dpkg --add-architecture
- changelog: version increments parse as u64 with checked arithmetic
(1.0-20250123123456 used to panic on the u32 parse)
set_current held the config RwLock for writing across make_context,
which for a context with a parent re-entered the same lock through
Context::new's global-manager lookup, deadlocking 'pkh context use'.
Context building is now lock-free by construction: make_context
resolves parent chains against a snapshot map (also rejecting parent
cycles), and neither set_current nor remove_context holds a guard
while building a Context.
A corrupt contexts.json no longer aborts every command at manager
init: load falls back to the default local-only config, backs the
corrupt file up to contexts.json.bak so a later save cannot silently
destroy it, and a dangling current/parent context falls back to local
with an error log instead of panicking.
build_binary_package installed its ephemeral chroot context into the
process-global manager and read it back with context::current(),
ignoring its ctx parameter: two concurrent builds would re-point each
other's global and each drop would clean up whichever chroot was
current at the time. The guard now keeps the Arc of the context it
created (parented directly on the base context, not on a config-name
lookup), exposes it via context(), and Drop cleans up exactly that
context and restores the exact handle that was current at creation,
so overlapping builds no longer cross-destroy each other.
The SIGINT handler libc::_exit(130)s, skipping EphemeralContextGuard's
drop and leaking the freshly bootstrapped chroot with its bind-mounted
/proc and overlay mounts. Resources now register a self-contained
cleanup hook in a process-global registry that the handler drains right
before exiting: the hook unmounts everything under the chroot path
(children first, lazy fallback) and removes the tree, using only stored
paths and direct umount/rm subprocesses so it cannot deadlock on a lock
the interrupted thread may hold; sudo -n keeps it from ever hanging on
a password prompt. Drop deregisters the hook first, so the normal
cleanup path is unchanged.
Also fixes the hex grouping of the CRC-24 polynomial in apt::release.
Debian Policy 7 defines the deprecated single-character spellings as
'earlier/later or equal' (i.e. <= and >=), and dpkg still accepts them
that way; the parser mapped them to the strict << and >> instead, so
'foo (< 1.0)' was wrongly reported unmet against installed 1.0.
Without an explicit --pocket, find_package stops at the first pocket
containing the package, but the search order listed '-proposed' first
and never included '-security': unreleased proposed packages won by
default and security-only updates were unreachable. Search the main
archive first, then updates, security, and proposed last.
A --ppa value that was not exactly 'user/name' (full URL, extra
segment, empty halves) made base_url None and pulled the package from
the main archive without any warning. Error out naming the expected
format instead, and document the format in --help.
The Sources index was downloaded with no authentication: per-artifact
checksums were verified, but against hashes taken from an index a MITM
could substitute along with the artifacts. Fetch each suite's InRelease
(or Release + Release.gpg), verify the signature with gpgv against the
archive keyring (or the PPA signing key) the same way apt does, and
checksum-check every Sources index against it before parsing.
Distro archives and PPAs verify strictly: an invalid or unverifiable
signature, or a missing gpgv binary, is a hard error. Flat repositories
keep working without a Release file or without a verifiable one (warned
as unauthenticated), but tampering evidence is a hard error there too.
Also switches all archive, PPA and keyring base URLs to https, and
reads suite components from the verified Release instead of fetching
them separately over an unauthenticated channel.
All three non-local drivers assembled remote/chroot command strings by
raw concatenation: ssh pushed args verbatim (TODO: escape), schroot
interpolated env values raw so DEB_BUILD_OPTIONS='parallel=4 nocheck'
made sh treat 'nocheck' as the command, and unshare wrapped args in
unescaped double quotes letting quotes break out and $/backticks
expand. Add a shared POSIX shell_quote helper and use it for every
component interpolated into a shell string, including ssh copy_path
(which used Rust's {:?}, not shell quoting) and schroot write_file
(which also switches echo -ne to printf %s so backslash sequences in
content are no longer interpreted).
Saving the modified source entries with save_legacy rewrote every entry
into /etc/apt/sources.list in legacy format, destroying deb-src entries
and signed-by/trusted options, duplicating every distro entry that came
from a deb822 file (which stayed in place), and hardcoding the Ubuntu
keyring on cross builds. Entries now remember the file and format they
were loaded from and are written back there; new entries (PPAs, ports)
go to a pkh-owned /etc/apt/sources.list.d/pkh-added.list, and a one-time
<path>.pkh-backup copy is made before overwriting an existing file.
Also fixes: 'Types: deb deb-src' stanzas are split instead of being
treated as binary-only, commented-out legacy entries are kept disabled
instead of deleted, debian.sources is actually read on Debian (the old
else-if never fired), and the double blank lines save_legacy emitted.
The shared world-writable /tmp/pkh-keyrings directory, combined with the
skip-if-exists logic, let any local user pre-plant keyrings that pkh
then trusts into the chroot's trusted.gpg.d. Use a per-uid 0700
directory instead, refuse to reuse a pre-existing directory that is not
owned by the current user or is group/other-writable, and drop the now
unnecessary world-accessibility chmods (mmdebstrap in unshare mode runs
with the same real uid).
Entry::unpack performs no path sanitization, so a malicious or malformed
tarball (PPA, flat repository) could write files outside the package
directory via '..' components or absolute entry paths. Refuse such
entries with an error naming the offending path.
Without --host-architecture, the second build-dep pass re-resolves the
whole Build-Depends field for the native architecture: apt swaps
host-arch -dev packages for native ones (e.g. libcurl4-gnutls-dev,
whose arch-differing curl-config makes dpkg refuse the co-install) and
breaks the cross build environment.
Per dpkg-checkbuilddeps, both Build-Depends and Build-Depends-Indep
resolve for the host architecture in cross mode, so pass
--host-architecture to the second pass as well. Skip the pass entirely
when the source declares no Build-Depends-Indep.
Add an end-to-end regression test building a package that declares
libdb-dev in both fields and links a host-arch binary against it: the
test only passes if the arch-indep pass did not swap the arm64 -dev
packages for native ones.
Add a --repository flag taking the full suite URL of an external flat
repository (e.g. https://pkg.noctalia.dev/deb/resolute/), i.e. one with
no dists/ hierarchy, like apt's exact-path suites ('Suites: resolute/').
The suite name is read from the root Release file (Codename/Suite), the
sources index is fetched from the repository root as Sources.xz/gz/plain,
and package files are resolved against the URL root, ignoring the stanza
Directory field like apt does. As with PPAs, the stanza Vcs-Git is never
used for external repositories, so the source always comes from the
repository itself.
Also make the sources index parser detect compression by magic bytes
(gz/xz/plain) instead of assuming gzip, and fix extraction of archives
with './'-prefixed entries, which previously aborted and are now
extracted in place instead of being relocated.
build_binary_package_impl copied the whole parent directory into the
build root (ensure_available) and then retrieved every *.deb / *.changes
/ *.buildinfo it found there. That surfaced stale files already sitting
next to the package tree in the "Built in Ns:" summary.
Now local::build returns the exact set of artifacts produced by this
build — the binary packages registered in debian/files plus the
generated .buildinfo/.changes from generate_binary_metadata — and
build_binary_package_impl retrieves that list instead of globbing the
build root. Only files genuinely produced by the current build are
printed.
By default the number of parallel jobs is detected with nproc inside
the build context. Add a -j/--jobs option so an explicit count can be
honored instead, threading it through build_binary_package and
local::build into DEB_BUILD_OPTIONS=parallel=N.
Copying a git checkout verbatim into /tmp/pkh-build-*/ ships '.git',
flipping autotools' building-from-VCS detection (GNU hello's
BUILD_FROM_GIT): the shipped man page gets cleaned and regenerated
via help2man, which is correctly not in Build-Depends. Skip VCS
dirs (.git/.hg/.svn/.bzr/CVS) in local/unshare/ssh copies, and
prune them after an unshare overlayfs mount.
Route 'pkh build' through the DebUi capture machinery 'pkh deb'
already uses instead of letting dpkg-source inherit the terminal:
- pin LANG=C and LC_ALL=C so dpkg-source emits deterministic English
diagnostics regardless of the session locale;
- new DpkgSourceClassifier rewrites info:/warning:/error: lines into
colored pane entries, telling benign tar warnings from failures;
- DebUi generalizes for reuse (arbitrary phase labels, build-specific
log naming); run_source_build() drives phases and pipes subprocess
output through the sink when a UI is present;
- glyph-free house-style summaries: 'Built in Ns:' plus artifact
paths relative to cwd; failures print captured errors + log path;
- drop/capitalize pipeline chatter, add 'pkh build --verbose' to
bypass the view like 'pkh deb --verbose'.
Extend the metadata writers to binary-only uploads and wire them into
the 'pkh deb' flow:
- build/binary.rs generates <pkg>_<ver>_<arch>.buildinfo/.changes
through any Context: debian/files consumption, encounter-order
Architecture accumulation (sorted in .buildinfo like dpkg-genbuildinfo),
sorted Binary lists, dpkg-formatted Description lines with udeb
suffixes, Installed-Build-Depends closure over the context status DB,
and binNMU handling (Source: pkg (prev), Binary-Only-Changes, previous
.dsc redistribution);
- artifact digests are computed inside the context via coreutils
(md5sum/sha1sum/sha256sum/stat) so chrooted/remote trees work;
- deb/local.rs runs the generation after 'rules binary', exports
SOURCE_DATE_EPOCH from the changelog (reproducibility), and resolves
vendor/profiles inside the context; deb/mod.rs retrieves the new
artifacts alongside the debs;
- reusable helpers added: FilesList::parse/render,
parse_changelog_entry_from_str, parse_previous_version_from_str,
installed_build_depends_from_content.
Differential gate: same tree built with real 'dpkg-buildpackage -b' and
with the pkh flow; .changes/.buildinfo compared field-by-field modulo
machine-dependent fields, artifact checksums included.
Replace dpkg-checkbuilddeps with a native implementation:
- full dependency grammar: comma clauses, | alternatives, << <= = >= >>
relations, :arch qualifiers (any/native/specific), [arch lists] and
<profile restriction> formulas per alternative;
- restriction reduction against active build profiles and the host arch
at parse time (reduce_restrictions semantics);
- evaluation against a parsed dpkg status database with Multi-Arch
semantics (foreign/allowed) and versioned Provides rules (unversioned
provides never satisfy versioned deps; versioned ones must satisfy the
relation);
- clause simplification with implication-based deduplication, rendering
dpkg-compatible 'unmet build dependencies/conflicts' diagnostics.
check_build_depends() consumes debian/control + CheckOpts (-A/-B/-I
equivalents). run_source_build performs the check when forced (-D
parity); source-only builds skip it entirely like dpkg-buildpackage,
and unsatisfied deps propagate as UnmetBuildDependencies -> exit 3.
Unit tests port the Dpkg_Deps.t reduction matrices; differential gate
runs 24 scenarios (alternatives, versions, arch/profile restrictions,
Multi-Arch, Provides, conflicts, -A/-B flags) against real
dpkg-checkbuilddeps comparing exit status and diagnostics.
vhaudiquet