kea: ship admin-utils.sh and kea-shell Python modules
kea-admin and kea-shell were shipped without their helper script
and Python modules respectively, leaving both non-functional. Ship
the missing files, add +python3-light +python3-urllib +python3-openssl
to kea-shell, and rewrite its meson-baked host-absolute shebang and
sys.path.append to the on-target /usr/bin/python3 + $(PYTHON3_PKG_DIR).
Exercise anyio end-to-end with the structured-concurrency example
from the upstream "Creating and managing tasks" docs: spawn 5 child
tasks via create_task_group(), each appending to a shared list, then
verify all completed.
python-maxminddb: extend test.sh past the import smoke check
Previous test.sh only confirmed the import smoke check. Add a $2
version check, assert MODE_* constants are distinct ints, and exercise
the loader's error paths (missing file, non-MMDB temp file).
python-maxminddb: add setuptools-scm host build dep
maxminddb 3.x switched to PEP 517 with build-system.requires including
setuptools-scm, so the wheel build now fails with "ERROR Missing
dependencies: setuptools-scm". Wire it into PKG_BUILD_DEPENDS and bump.
zipp is a small backport / forward-port of zipfile.Path, originally
useful when zipfile.Path was new in stdlib (Python 3.8). With the
python3 package now tracking 3.14, zipfile.Path covers what
zipp.Path exposes, and no other package in this repository declares
zipp as a runtime or build dependency.
Drop the package to remove the dead weight; users who still need it
can install it via pip on the target.
None of the shipped binaries (mdnsd, dns-sd, mDNSClient/Net/Proxy*,
mDNSResponder) print the package version, so the generic probe always
fails. Add a test-version.sh exiting 0 for the three subpackages.
alsa-lib ships only a shared library and aserver has no --version, so
neither can satisfy the generic version probe. Add a test-version.sh
that exits 0 for both arms and exits 1 for unknown subpackages.
kea: depend on the concrete log4cplus, not the -any virtual
Commit 1ea43c352 ("kea: fix kea-libs dependencies") replaced the
direct +log4cplus dependency on kea-libs with the virtual name
+log4cplus-any. With only one log4cplus implementation in the feed,
nothing actually PROVIDES log4cplus-any other than the auto-generated
"package name + -any" alias on log4cplus itself.
apk treats `-any`-suffixed names as ambiguous virtuals that the
admin must resolve explicitly, even when only one provider exists,
and refuses to install:
ERROR: unable to select packages:
log4cplus-any (virtual):
note: please select one of the 'provided by'
packages explicitly
provided by: log4cplus
required by: kea-libs-3.0.2-r7[log4cplus-any]
That bubbles up to every kea subpackage runtime test in CI
(kea-admin, kea-ctrl, every kea-dhcp* and kea-hook-*) and fails
the multi-arch test build with exit code 3.
Switch back to +log4cplus directly; the +libstdcpp addition from
that earlier commit stays — it was the genuinely useful part of
the change. Bump PKG_RELEASE for the redo.
kea: fix BOOST_STATIC_ASSERT use in src/lib/dns/rdataclass.cc
Snapshot builds for i386_pentium-mmx (and presumably every other
target now that Boost 1.91 has been bumped in the feed) fail with::
src/lib/dns/rdataclass.cc:1401:5: error: 'BOOST_STATIC_ASSERT'
was not declared in this scope; did you mean
'BOOST_HAS_STATIC_ASSERT'?
1401 | BOOST_STATIC_ASSERT(sizeof(numdata_) ==
``BOOST_STATIC_ASSERT`` lives in ``<boost/static_assert.hpp>``. In
earlier Boost releases it was reachable through any number of
transitive includes pulled in by Kea's other DNS headers; in 1.91
those transitive paths have been pruned, so call sites must include
the declaring header explicitly. ``050-boost-static-assert.patch``
already does this for ``src/lib/log/logger_level_impl.cc``;
``src/lib/dns/rdataclass.cc`` uses the same macro at SOA::getMinimum()
to size-check ``numdata_`` and was missed.
Extend the existing patch with the matching include in rdataclass.cc.
Same fix shape as the earlier hunk; no behavioural change.
knot: add functional test.sh covering all CLI subpackages
Cover each user-facing subpackage with a real functional check instead
of relying solely on the CI's generic --version probe:
- knot: knotc conf-check on a minimal YAML server config
- knot-dig: kdig -h (CLI parser smoke check)
- knot-host: khost -h
- knot-nsupdate: feed 'quit' through the REPL
- knot-zonecheck: validate a minimal example.com zone file end to end
- knot-keymgr: initialise a KASP DB in a temp directory
knot-libs, knot-libzscanner, and knot-tests are library/harness
subpackages; the generic ELF/SONAME checks already cover them.
knot: drop kru.inc.c static_assert that requires lock-free 16-bit atomics
The RRL module's kru.inc.c has a `static_assert(ATOMIC_CHAR16_T_LOCK_FREE
== 2, ...)` that breaks the build on any target whose toolchain does
not advertise lock-free 16-bit atomics (e.g. ARMv5 arm926ej-s).
Upstream considers the assertion non-essential and plans to drop it;
they recommended a patch over a Makefile-level RRL disable. Bump
PKG_RELEASE for the patch swap.
Suggested-by: Daniel Salzman <redacted> Signed-off-by: Alexandru Ardelean <redacted>
python3: pin host curses to the SDK's narrow ncurses
Removing the ncursesw path from include avoids seeing the wide-char
version of ncurses, so the host build picks up the narrow static lib
that the OpenWrt SDK stages. To keep that pin from leaking back in
through other curses headers the build host may have under
/usr/include, also force every header check in Python's configure.ac
AC_CHECK_HEADERS(curses.h ncurses.h ncursesw/ncurses.h
ncursesw/curses.h ncursesw/panel.h panel.h) line to "no" via cached
ac_cv_header_* variables.
The narrow ncursesw_* trio was enough on the original reporter's
host; Jan Kardell's review of the first cut hit a build where
`/usr/include/curses.h` and `/usr/include/ncurses.h` from the
distro's libncurses-dev were still being picked up. Adding the
non-wide entries (and panel.h for symmetry, in case a host also
ships /usr/include/panel.h alone) makes the override insensitive to
whichever curses dev packages the host distro happens to provide.
Reported-by: Jan Kardell <redacted> Signed-off-by: Alexandru Ardelean <redacted>
libtheora: disable ARM assembly on pre-ARMv7 targets only
libtheora's configure probes the host assembler for NEON support
rather than the target CPU, so on pre-ARMv7 (arm926ej-s, etc.) it
falsely succeeds and the build emits NEON the CPU can't execute.
Force --disable-asm only when the ARM target's TARGET_CFLAGS does
not advertise armv7 or armv8, so cortex-a* keeps the NEON path.
This is the latest 3.14.x point release with assorted bug fixes
from upstream. PYTHON3_PIP_VERSION is bumped from 26.0.1 to 26.1.1
to match the pip wheel bundled in the 3.14.5 source tarball; the
host build's ensurepip lookup of pip-$(PYTHON3_PIP_VERSION).whl
otherwise fails. The set of OpenWrt-side patches still applies
against the new source; only quilt context-line offsets needed
refreshing for the patches that touch Makefile.pre.in.
Noah Meyerhans [Mon, 25 May 2026 15:09:01 +0000 (11:09 -0400)]
bind: bump to 9.20.23
Resolves several security issues:
- CVE-2026-3592: Limit resolver server list size.
- CVE-2026-3039: Fix GSS-API resource leak.
- CVE-2026-5950: Avoid unbounded recursion loop.
- CVE-2026-5947: Fix crash in resolver when SIG(0)-signed responses are
received under load.
- CVE-2026-3593: Add system test for HTTP/2 SETTINGS frame flood.
- CVE-2026-5946: Disable recursion, UPDATE, and NOTIFY for non-IN views.
Complete list of changes is available upstream at
https://ftp.isc.org/isc/bind9/9.20.23/doc/arm/html/changelog.html
Dirk Brenken [Fri, 29 May 2026 11:57:31 +0000 (13:57 +0200)]
banip: update 1.8.8-5
* bugfix: only load the configuration once per run: a new `ban_confload`
guard short-circuits `f_conf()` on subsequent calls, avoiding
repeated `config_load` invocations
* new: the per-set report now sorts elements by their packet counter in
descending order before truncating to the top 50, so the report
shows the most active elements instead of just the first 50 found
python-cryptography: update to 48.0.0 + backport cross-include fix
Bump from 46.0.7 to the current 48.0.0 release. Notable upstream
changes since 46.0.7:
- 48.0.0 drops Python 3.8 support (requires 3.9+); raises
Py_LIMITED_API floor to 0x030900f0.
- Adds ML-KEM / ML-DSA post-quantum primitives via OpenSSL 3.5.0+
(in addition to existing AWS-LC / BoringSSL paths).
- BACKWARDS INCOMPATIBLE: stricter X.509 CRL signature-algorithm
matching (mismatched inner/outer algs now raise ValueError at parse
time).
- Drops 32-bit Windows wheels and ships macOS only on arm64.
Replace the old downstream cross-compile fix with a backport of the
upstream-merged version from pyca/cryptography PR #14904
(commit 5d072cb2a685, scheduled for the release after 48.0.0).
Daniel Golle [Tue, 26 May 2026 15:56:14 +0000 (16:56 +0100)]
libx264: update to git HEAD of 2025-09-10
Changes since 2024-05-13: c24e06c2 configure: Check for SVE support in MS armasm64 via as_check 3a8b5be2 aarch64: Use elf_aux_info() for CPU feature detection on FreeBSD/OpenBSD 1243d9ff Provide x264_getauxval() wrapper for getauxvaul() and elf_aux_info() 80c1c47c configure: Add DragonFly support 3a21e97b Fix build with Android NDK and API < 24 for 32-bit targets b1d2de88 Use getauxval() on Linux and elf_aux_info() on FreeBSD/OpenBSD on arm/ppc da14df55 Make use of sysconf(3) _SC_NPROCESSORS_ONLN and _SC_NPROCESSORS_CONF 023112c6 aarch64: defines involving bit shifts should be unsigned 938601b9 Use sysctlbyname(3) hw.logicalcpu on macOS a64111b1 Enable use of __sync_fetch_and_add() wherever detected instead of just X86 450946f9 ci: Test compiling for Android 52f7694d Use sched_getaffinity on Android 373697b4 Bump dates to 2025 c80f8a28 msvsdepend: Allow using the script for .S sources too 27d83708 Makefile: Generate dependency information implicitly while compiling a0191bd8 configure: Use as_check for checking for aarch64 features 72ce1cde configure: Use as_check for the main check for whether NEON is supported f87ca183 configure: Check for .arch and .arch_extension for enabling aarch64 extensions 87044b21 aarch64: Use configure detected directives for enabling SVE/SVE2 fc4012fb configure: Check for the dotprod and i8mm aarch64 extensions 0e48d072 aarch64: Add flags for runtime detection of dotprod and i8mm 570f6c70 aarch64: Add runtime detection of extensions on Windows and macOS fe9e4a7f Provide implementations for functions using the instructions SDOT/UDOT in the DotProd Armv8 extension. 32c3b801 lavf: Update the code to work with the latest libavutil API 4360ac37 ci: Fix ffmpeg build 40617ddb ci: Remove vlc-contrib dependency 85b5ccea Update gas-preprocessor.pl to the latest upstream version ff620d0c configure: Use MSYSTEM_CARCH for default arch on msys2 714e07b4 arm: Don't test x264_cpu_fast_neon_mrc_test on Windows 291476d7 windows: Fix named pipes detection b35605ac i8mm & neon hpel_filter optimization 0480cb05 riscv64: add compile support
Tyrel M. McQueen [Wed, 27 May 2026 16:27:31 +0000 (12:27 -0400)]
unbound: update to 1.25.1
From upstream: 1.25.1 consolidates security fixes for issues
reported over a period of time. There are fixes for CVE-2026-33278,
CVE-2026-42944, CVE-2026-42959, CVE-2026-32792, CVE-2026-40622,
CVE-2026-41292, CVE-2026-42534, CVE-2026-42923, CVE-2026-42960,
CVE-2026-44390 and CVE-2026-44608.
Dirk Brenken [Tue, 26 May 2026 15:16:35 +0000 (17:16 +0200)]
adblock: update 4.5.5-6
- f_search: refactored backup and local block-/allowlist scan to run in parallel
- f_search: raise the search timeout in backups to max. 90 seconds
- f_load: skip f_fetch for search action, no remote download is needed for local lookups
- f_dns: skip destructive paths (file wipe, f_dnsup) for search and report actions
- LuCI: adapt search backend changes in the frontend
- LuCI: minor frontend fixes & improvements
Qingfang Deng [Tue, 26 May 2026 03:27:34 +0000 (11:27 +0800)]
ovpn-dco: fix build on kernel 6.18.33
Kernel version 6.18.33 backports kmalloc_obj macros but GFP flags are
required, which causes build failures as ovpn omits GFP flags. Undef
those macros to fix the build.
The three CLI helpers shipped by python3-argcomplete
(activate-global-python-argcomplete, register-python-argcomplete,
python-argcomplete-check-easy-install-script) don't accept a --version
flag and emit no PKG_VERSION string in their usage output. With all
three executables missing the version, the generic CI test stage
fails with "No executables in the package provided version 3.6.3".
Add a test-version.sh that emits a line containing PKG_VERSION so the
framework's "Version check override" passes. The existing test.sh
already exercises the Python module import.
Security fix:
- Fix Cursor.callproc() to escape procedure name, preventing SQL injection
when calling a procedure with a string received from an untrusted source
- NOTICE: Backward compatibility change - procedure names like
"dbname.funcname" are now backtick-quoted: ``CALL \`dbname.funcname\` ``
Other changes:
- CI: use ubuntu-slim, add dependabot for GitHub Actions
- Bump GitHub Actions (checkout v4→v6, setup-python v5→v6, codecov v5→v6)
- Add publish.yml workflow (copied from psf/requests)
- Upgrade dependencies: cryptography>=46.0.7, PyNaCl>=1.6.2
- Drop Python 3.8, require Python 3.9+
New features (pipx 1.12.0):
- Add --fetch-python / PIPX_FETCH_PYTHON env var (always|missing|never)
to control standalone Python interpreter downloads
- Add opt-in "uv" backend: pipx can now use "uv venv" and "uv pip" for
managing virtual environments
* When "uv" is on PATH, defaults to using uv for NEW venvs
* Existing venvs keep their recorded backend (pip or uv)
* Set PIPX_DEFAULT_BACKEND=pip to force pip even with uv available
* pipx install pip always uses the pip backend (uv venvs have no pip)
Deprecations:
- --fetch-missing-python and PIPX_FETCH_MISSING_PYTHON deprecated;
use --fetch-python=missing or PIPX_FETCH_PYTHON=missing instead
Fixes:
- Moved "headers" input type back to Mapping to avoid invariance issues
with MutableMapping and inferred dict types.
Users calling Request.headers.update() may need to narrow typing in code
(Closes #7441).
Security:
- CVE-2026-25645: Fixed extract_zipped_paths to extract contents to
a non-deterministic temp directory, to prevent malicious file replacement.
Does not affect default usage of Requests, only apps calling this utility
directly.
Fixes (click 8.3.3):
- Fix help strings for "help_option_names" that do not contain "-"
- Help string generation now properly handles option names with dashes
The RTKLIB command-line tools (convbin, pos2kml, rnx2rtkp, rtkrcv,
str2str) don't accept a --version flag; passing any unrecognized option
triggers printhelp()/printusage() which only emits a synopsis block. The
generic CI version probe therefore can't find PKG_VERSION and marks the
package as missing a version match.
Add a test-version.sh that exit-0's for all five subpackages to skip the
generic version check, and a test.sh that exercises each binary's
synopsis output as a basic functional smoke test.
The MIPS variants (mips_24kc, mips_4kec, mipsel_24kc, mipsel_74kc) all
fail to compile preceph.c with an internal compiler error:
during RTL pass: reload
src/preceph.c:317:1: internal compiler error:
in lra_update_fp2sp_elimination, at lra-eliminations.cc:1416
This is a GCC LRA pass bug triggered when compiling with -mips16. Set
PKG_BUILD_FLAGS:=no-mips16 to strip the -mips16 / -minterlink-mips16
flags from CFLAGS for this package, matching the approach already used
by stress-ng for the same class of issue.
Bump PKG_RELEASE since only the build flags change.
the original node_exporter exposes a node_os_info metric with a set of
data about the system [1] which is then used by several dashboards.
openwrt.lua already exposes OS info, but using the node_openwrt_info
metric requires changes to existing dashboards, and would require more
complex lookups when there are non-OpenWrt hosts in the overview too.
as we've already called ubus and fetched the data, we can expose it in
two formats easily.
Add a Prometheus collector for ModemManager that exports cellular modem
signal metrics via mmcli. Supports multiple modems (labeled by D-Bus
object path), exports overall signal quality and detailed per-technology
signal parameters (LTE, NR5G, UMTS, GSM, CDMA, ...).
Requires signal refresh to be enabled on the modem:
mmcli -m <id> --signal-setup=<interval_seconds>
Tested on: ath79/generic, GL.inet GL-X300B, OpenWrt 23.05.5
Co-authored-by: Claude <redacted> Signed-off-by: Jean-Laurent Girod <redacted>
erlang: fix PKG_CPE_ID escaping for apk ADB format
apk's ADB binary package format rejects both the backslash-escape and
the percent-encoding variants of the previous CPE id:
cpe:/a:erlang:erlang\/otp ERROR: info field 'tags' has invalid value
cpe:/a:erlang:erlang%2Fotp ERROR: info field 'tags' has invalid value
apk's tag value parser only accepts a restricted alphabet for ADB
package format and neither '\' nor '%' make the cut. The result is
that the package never produces an .apk.
Drop the '/otp' suffix entirely and use cpe:/a:erlang:erlang, which
matches the higher-level Erlang CPE entry. cve scanners that walked
the more specific erlang\/otp entry will fall back to this one.
This effectively reverts the product portion of bfdf01496 ("lang/erlang:
fix PKG_CPE_ID"), which was correct against the NIST 2.3 string but
incompatible with apk's tag parser.
libxerces-c: fix CI build/test failures for apk + samples
Two unrelated issues both fixed here so the package + its samples
sub-package land in CI green:
1. PKG_CPE_ID escaping.
apk's ADB binary package format rejects both the backslash-escape
and the percent-encoding variants of the previous CPE id:
cpe:/a:apache:xerces-c\+\+ ERROR: info field 'tags' has invalid value
cpe:/a:apache:xerces-c%2B%2B ERROR: info field 'tags' has invalid value
apk's tag value parser only accepts a restricted alphabet for ADB
package format and neither '\' nor '%' make the cut. With xerces-c
unable to build, downstream consumers (notably sumo) also fail at
cmake configure time with "Failed to find XercesC".
Drop the '++' suffix entirely and use cpe:/a:apache:xerces-c, which
matches the higher-level Apache Xerces-C CPE entry. cve scanners
that walked the more specific xerces-c++ entry will fall back to
this one.
2. Generic version-check override for libxerces-c-samples.
The samples sub-package ships upstream demo programs
(CreateDOMDocument, DOMCount, DOMPrint, SAX2Count, ...) which do
not accept --version / -v / -V and therefore fail the framework's
"executable prints PKG_VERSION" probe, making the package overall
report "Generic tests failed". Add a minimal test-version.sh that
exits 0 so the version-probe is skipped and the remaining generic
checks (executable, no hardcoded paths, stripped, linked libs)
still run for every binary.
lmdb: add a test-version.sh version-check override
lmdb-test ships /usr/bin/mtest, a stress test that unconditionally
opens ./testdb in the current directory and aborts (SIGABRT) when
that directory is missing. In the CI runtime container that's
always the case, so the framework's --version probe ends up with
"Aborted" output, no PKG_VERSION match, and the package gets
reported as "Generic tests failed - No executables in the package
provided version 0.9.35".
Add libs/lmdb/test-version.sh that handles each sub-package by name:
lmdb (library) and lmdb-test (no usable version probe) pass the
override, lmdb-utils runs 'mdb_dump -V' and matches against
PKG_VERSION, and unknown sub-packages fail loudly to force this
script to be updated. The other generic checks (no hardcoded paths,
stripped, linked libs) still run for every binary.
Bump from 2.0.1 to the latest 2.0 LTS release. This pulls in
upstream commit 53cb738795 ("dnsdist: make code boost-1.91
compatible", Otto Moerbeek, 2026-04-29), which fixes the build
break against Boost 1.91 currently shipped by OpenWrt:
dnsdist-lua.cc:3086:101: error: converting to
'boost::optional<unordered_map<...>>' from initializer list
would use explicit constructor 'constexpr boost::optional<T>::
optional(U&&) [...]'
Xing-Kai Wang [Sun, 10 May 2026 15:52:03 +0000 (23:52 +0800)]
openfortivpn: add realm parameter support
Some FortiGate VPN gateways require a specific authentication realm
when multiple domains or user groups are configured on the same server.
This commit updates the netifd protocol script to parse the 'realm'
option from the UCI configuration and correctly append it to the
openfortivpn command line arguments.
Josef Schlehofer [Sun, 10 May 2026 20:58:19 +0000 (22:58 +0200)]
treewide: fix dangling SONAME symlinks when using ABI_VERSION
A previous commit attempted to introduce proper SONAME symlinks for
packages utilizing ABI_VERSION. However, it incorrectly copied only
the symlink without the underlying physical library file, resulting
in broken packages with dangling symlinks.
Before:
```
tar -Oxzf bin/packages/arm_cortex-a9_vfpv3-d16/packages/oniguruma_6.9.9-r1_arm_cortex-a9_vfpv3-d16.ipk ./data.tar.gz | tar -tzvf -
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./usr/
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./usr/lib/
lrwxrwxrwx 0 0 0 0 Feb 5 2024 ./usr/lib/libonig.so.5 -> libonig.so.5.4.0
```
After:
```
tar -Oxzf bin/packages/arm_cortex-a9_vfpv3-d16/packages/oniguruma_6.9.9-r1_arm_cortex-a9_vfpv3-d16.ipk ./data.tar.gz | tar -tzvf -
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./usr/
drwxr-xr-x 0 0 0 0 Feb 5 2024 ./usr/lib/
lrwxrwxrwx 0 0 0 0 Feb 5 2024 ./usr/lib/libonig.so.5 -> libonig.so.5.4.0
-rwxr-xr-x 0 0 0 526493 Feb 5 2024 ./usr/lib/libonig.so.5.4.0
```
This properly standardizes shared library packaging, prevents shipping
duplicate full-sized files (as seen previously in packages like libre2),
and aligns the packages feed with core OpenWrt practices.
As you could see in the previous commit (see it in the ``Fixes`` tag):
Before:
```
tar -Oxzf bin/packages/arm_cortex-a9_vfpv3-d16/packages/re2_2023.02.01\~b025c6a3-r1_arm_cortex-a9_vfpv3-d16.ipk ./data.tar.gz | tar -tzvf -
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./usr/
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./usr/lib/
-rw-r--r-- 0 0 0 331875 Aug 18 2024 ./usr/lib/libre2.so
-rw-r--r-- 0 0 0 331875 Aug 18 2024 ./usr/lib/libre2.so.10
-rw-r--r-- 0 0 0 331875 Aug 18 2024 ./usr/lib/libre2.so.10.0.0
```
After:
```
tar -Oxzf bin/packages/arm_cortex-a9_vfpv3-d16/packages/re2_2023.02.01\~b025c6a3-r1_arm_cortex-a9_vfpv3-d16.ipk ./data.tar.gz | tar -tzvf -
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./usr/
drwxr-xr-x 0 0 0 0 Aug 18 2024 ./usr/lib/
lrwxrwxrwx 0 0 0 0 Aug 18 2024 ./usr/lib/libre2.so -> libre2.so.10
lrwxrwxrwx 0 0 0 0 Aug 18 2024 ./usr/lib/libre2.so.10 -> libre2.so.10.0.0
-rwxr-xr-x 0 0 0 331875 Aug 18 2024 ./usr/lib/libre2.so.10.0.0
```
Fixes: 537c2a631dd117564720a80f6976280901b010f5 ("treewide: avoid deref symlinks when installing .so") Signed-off-by: Josef Schlehofer <redacted>
Andy Chiang [Mon, 18 May 2026 16:56:53 +0000 (23:56 +0700)]
qemu-ga: skip starting on bare metal device
The same firmware image may be deployed on either bare metal device or
virtualized platforms (e.g., Proxmox VE).
On bare metal device, `qemu-ga` may still be started even though no
virtio-serial channel is available, resulting in repeated attempts to
access /dev/virtio-ports/org.qemu.guest_agent.0.
This causes continuous service respawning by procd and unnecessary log
spam.
This commit adds a pre-check for /dev/virtio-ports to avoid starting
`qemu-ga` when virtio-serial support is not present.
Andy Voigt [Sun, 10 May 2026 19:36:29 +0000 (21:36 +0200)]
libsml: add new package
libSML implements the Smart Message Language protocol used by German smart
meters (FNN specification). It is used by projects like volkszaehler for
reading smart meter data.
Fotios Kitsantas [Tue, 19 May 2026 18:39:50 +0000 (19:39 +0100)]
ddns-scripts: add blazingfast.io Anycast DNS provider
Add DDNS update support for blazingfast.io Anycast DNS via their
REST API. Authentication is performed via JWT token obtained from
the login endpoint. Zone records are fetched to verify the record
type before update, ensuring IPv4 services only target A records
and IPv6 services only target AAAA records.
Service, zone and record IDs are passed via param_opt as
space-separated key=value pairs:
service_id=X zone_id=Y record_id=Z
curl --config file approach is used throughout to avoid eval and
shell injection from user-controlled values. Supports both IPv4
and IPv6. For dual-stack, create two separate DDNS service sections
with their respective record IDs.
Tested on GL.iNet MT5000 (Brume 3) running OpenWrt with
ddns-scripts 2.8.2.
Erik Conijn [Thu, 21 May 2026 14:09:40 +0000 (16:09 +0200)]
openvpn: add missing host routes
Maintainer: Alexandru Ardelean <redacted>
ping @feckert
First of all big thanks to all involved devs, porting this to proto is not a minor task and besides some small quirks it is working well.
(Not all that happy with the use of a default route instead of /1 routes, because you loose internet if the tunnel goes down but that is just me nitpicking)
However I had problems with default routing as the host routes to the server endpoint were missing.
I tracked it down to code in the `openvpn-hotplug` script and made some changes and in my testing it appears to work now.
As a bonus I also added code for a future implementation of the `nohostroute` option.
Problem:
The host routes were created by just using route setup this however does not work.
Solution:
using `proto_add_host_dependency` seems the better solution.
Furthermore the correct guard for IPv6 seems to be `net_gateway_ipv6` instead of `route_ipv6_gateway` however even the correct guard is only working if ipv6 source routing is disabled on wan6, so perhaps we should consider removing the guard entirely.
For now I left it in place with a warning.
I have tested it on X86 running master build from 5 days ago, both for IPv4 and IPv6
Large version jump from 4.8.1 to 4.19.4 (latest upstream LTS).
Build changes:
- Refresh patches/004-fix-su-controoling-term.patch: su.c moved the
ioctl() call from line 1122 to 1169 and changed (char *) 0 to
(char *) NULL; update patch context and re-canonicalise through
quilt (blank context line spacing).
- New CONFIGURE_ARGS:
* --disable-logind: 4.19.4 added an optional libsystemd-based
logind integration which OpenWrt doesn't ship.
* --without-libbsd: shadow's configure now hard-fails on missing
readpassphrase() unless libbsd is found; the in-tree
lib/readpassphrase.c fallback is enabled by --without-libbsd.
* --without-sssd: avoid dragging in an sssd build dep.
* --disable-subordinate-ids: 4.19.4 builds libsubid (subuid/subgid
runtime API) unconditionally when subids are enabled, and its
libtool -export-symbols-regex generates a version script that
binutils 2.40+ rejects against libxcrypt's versioned
crypt_checksalt@@XCRYPT_4.3 symbol. Disabling subordinate-ids
skips libsubid entirely; OpenWrt doesn't ship libsubid.
- Drop newgidmap, newuidmap, lastlog and groups from SHADOW_APPLETS:
newgidmap/newuidmap are only built when subordinate-ids are
enabled, lastlog defaults to disabled in 4.19.4, and the groups
binary was removed from shadow upstream (use coreutils).
Test coverage:
- Replace the per-applet --version check in test.sh with per-applet
functional tests:
pwck -> 'pwck -r' read-only consistency check; accept
non-zero exit since the CI container's /etc/passwd
trips minor warnings.
grpck -> 'grpck -r' read-only consistency check.
chage -> 'chage -l root' lists password aging info.
useradd -> 'useradd -D' dumps defaults without modifying state.
passwd -> 'passwd -S root' prints the password status line.
faillog -> create empty /var/log/faillog then 'faillog -a'
must emit a header line.
login/su -> PAM-interactive; presence covered by generic tests.
Other applets -> verify binary presence (CI's generic tests
already check stripped, no build paths, linked-libs).
- Add test-version.sh as a generic-version-check override: shadow
tools don't honour --version (only --help), so the framework's
probe finds no PKG_VERSION match in any binary and would otherwise
fail Generic tests for every sub-package.
Update from 0.26.5 to 1.0.4 (major version jump 0.x -> 1.x).
Changes:
- Add +libcurl to DEPENDS: libupnpp 1.0.4 introduced a new mandatory
dependency on libcurl (used for HTTP/SOAP communications)
- Build system remains meson (already in use)
- No patches needed
libudev-zero: backport hwdb USB ID lookup from upstream master
Replace the four stub udev_hwdb_*() functions with a working
implementation that looks up vendor and product names from
/usr/share/hwdata/usb.ids, so callers using the standard libudev
hwdb API benefit without needing package-specific patches.
The patch is a clean backport of upstream commit 2bebebc9e0444
("udev: implement hwdb USB ID lookup from usb.ids (#80)") merged
to illiliti/libudev-zero master on 2026-05-19, post-1.0.3. Drop
when the package is bumped to the next libudev-zero release.
Upstream now defaults USB_IDS_PATH to ${SHAREDIR}/hwdata/usb.ids
with SHAREDIR=${PREFIX}/share, so the explicit
USB_IDS_PATH=/usr/share/hwdata/usb.ids in MAKE_FLAGS is no longer
needed and is dropped.
Fixes: https://github.com/openwrt/packages/issues/29386 Signed-off-by: Alexandru Ardelean <redacted>
Update from 017 to 019. Version 019 dropped autoconf in favour of meson,
so switch to include/meson.mk and drop PKG_FIXUP:=autoreconf and the
autoconf CONFIGURE_ARGS.
The binary lsusb no longer reads usb.ids directly; it now queries the
udev hardware database. lsusb.py still searches /usr/share/hwdata/usb.ids
for device name resolution.
On some build systems, the miniperl binary is created without execute
permission (errno 126 when running it as /bin/sh). This breaks building
the mro extension and cascades to all dependent packages.
Fix by ensuring chmod +x on miniperl after the main build step.
This matches how many other build systems handle this same issue.
Daniel Golle [Tue, 19 May 2026 10:25:40 +0000 (11:25 +0100)]
lvm2: update to 2.03.41
Stable bug-fix release in the 2.03.x series. The bundled
device-mapper library bumps from 1.02.209 to 1.02.215; track that
in PKG_VERSION_DM as well so the libdevmapper package shows the
correct upstream version.
$(PKG_CONFIG_PATH) here is the make-time variable, which is never
set for host context (include/host-build.mk only exports it as a
per-rule env variable, not a make variable). It therefore expanded
to the empty string, and the resulting single-entry PKG_CONFIG_PATH
overrode the exported defaults from host-build.mk, hiding uuid.pc
from the host pkg-config wrapper.
Fix: refer to the env variable via shell expansion ($$PKG_CONFIG_PATH)
instead of the make variable. The make recipe expands $$ to $, and
the shell resolves $PKG_CONFIG_PATH from the inherited environment
populated by include/host-build.mk (which sets it to
$(STAGING_DIR_HOST)/lib/pkgconfig:$(HOST_BUILD_PREFIX)/lib/pkgconfig).
Fixes: 5e4f937e ("ola: use protobuf-compat instead of protobuf") Signed-off-by: Alexandru Ardelean <redacted>
rsyslog: make libyaml support optional via RSYSLOG_libyaml
The upstream rsyslog build links against libyaml-0.so.2 whenever
pkg-config detects yaml-0.1, which made libyaml a hard dependency
of the rsyslog binary without any way to opt out from OpenWrt
config. libyaml is only used for YAML (.yaml/.yml) configuration
files and for loading rate-limiting policies from external files;
RainerScript (.conf) installs do not need it.
Introduce a new RSYSLOG_libyaml Config.in switch, default off, and
gate the +libyaml DEPENDS entry on it. Pass --enable-libyaml or
--disable-libyaml to configure based on the switch.
The upstream configure script in 8.2604.0 does not understand a
--disable-libyaml flag (libyaml was unconditionally autodetected).
Backport upstream commit c5c244861 ("configure: make libyaml
default-on explicit") as 001-configure-make-libyaml-default-on-
explicit.patch, which adds the AC_ARG_ENABLE(libyaml) block. The
patch dissolves cleanly once the package is bumped to 8.2606.0+.
PKG_FIXUP:=autoreconf is added so the configure.ac change flows
into the generated configure script during the SDK build.
git.99rst.org / openwrt-packages.git / log