Now that post install is running properly, when CI executes the
tests the zabbix-extra-* packages that have a posinst exit the
post install with an error if ubusd is not running.
Drop the error code from killall with no usbd running to fix this.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
There are no longer packages that need this.
This offers advanced SSH functions in Python.
But for OpenWrt's context it may work with just using ssh commands
with the subprocess package.
Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Before the change, the feed Makefile would not rely on the predefined
logic of building Makefile-based packages, rather it defined its own
`Build/Compile` step that explicitly called `make`. This didn't play
along well with the upstream Makefile, which doesn't respect the CFLAGS
that come from the environment.
The effect of that was that, for example, `CONFIG_PKG_ASLR_PIE_ALL=y`
was not respected: a non-PIE binary was generated regardless. Other
compiler flags were ignored too in the same manner.
Also it would try to determine the mdns-repeater version by calling to
git, which wouldn't work, since the build system downloaded a pre-made
source tarball without a bundled `.git` directory. The result is an
empty version string when calling `mdns-repeater -h`.
To address all of the above, switch to the built-in logic for
Makefile-type projects. Append to MAKE_FLAGS to completely override
upstream CFLAGS/LDFLAGS and pass through the package version. Set a
dummy HGVERSION variable value to prevent the upstream Makefile from
calling `git` unnecessarily.
Signed-off-by: Foster Snowhill <gh@pen.gy>
This adds the feature to choose which plugins to include or not.
Additionally, the needed shared plugins are added automatically.
Signed-off-by: Christian Korber <ckorber@tdt.de>
This commit enables to choose whether to compile the plugins into the
binary or not. If not included, it moves the plugins to
`/usr/lib/ModemManager/`.
Signed-off-by: Christian Korber <ckorber@tdt.de>
Using add_interface_triggers and check_interface_up is a
better separation of concerns and the code has little in
common (no duplication of code concerns).
While we are at it make both nut-monitor and nut-server failure of
add_interface_triggers non-fatal with a logged error message.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
Factor out some common validation code into some validator functions,
and add more UCI config validation.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
We claimed we only supported a single upsd section named 'upsd' and a
single uspmon section named 'upsmon', but when determining RUNAS
and STATEPATH we were using overly-complex functions as we were
attempting to support multiple upsd and upsmon sections.
Simplify the logic by adhering to our stated support.
While were are at it we make RUNAS handling more consistent between
nut-server (upsd) and nut-monitor (upsmon).
We no longer try to limp along if RUNAS is not set by find_runas; to do
this properly would require over-complicated logic when setting file
and directory permissions and ownership.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
* Fixes a missing word in a comment
* Fixes inconsistent indentation in sample configurations
* Drops redundant default assignment in find_statepath()
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
Verify that asgiref.__version__ strictly equals the expected version
argument ($2), instead of just checking that it is non-empty.
Also remove single quotes from EOF to enable variable expansion.
Signed-off-by: Wei-Ting Yang <williamatcg@gmail.com>
LN is already defined as 'ln -sf' in rules.mk, so passing -sf again
results in 'ln -sf -sf ...', which fails with stricter ln implementations.
Fixes#29243.
Signed-off-by: Xuzhen LI <xuzhen.li@outlook.com>
Add missing shell interpreter lines to the arp-scan maintainer scripts.
This lets apk execute the generated install and removal hooks directly.
Use IPKG_INSTROOT in the database postrm cleanup and bump PKG_RELEASE.
Signed-off-by: Dharmik Parmar <dharmikparmar2004@yahoo.com>
The devlink plugin allows updating firmware of network devices
implementing the kernel devlink interface. It is gated on libmnl,
which was hard-disabled so far. Wire the libmnl meson option to a
new FWUPD_PLUGIN_DEVLINK config symbol, enabled by default, and add
the corresponding libmnl runtime dependency to fwupd-libs.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Backport upstream e0c4d324a, which lets a LIB_* variable be emptied to skip
that library's check. Drop 001-disable-extra-stressors.patch, which hardcoded
the same choice, and drive it from a Config.in menu instead. Defaults keep the
previously built set of stressors: only crypt, aio, bsd, jpeg, kmod, sctp and
zlib are probed. apparmor, EGL, GBM, GLES2, IPSec_MB and Judy have no OpenWrt
package, so they are never probed. Drop the backport once it is in a release.
Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Drop 002-disable-atomics-mips-and-ppc.patch. It papered over link errors for
64-bit atomics on MIPS and PPC, which happen because stress-ng strips -latomic
whenever $(CC) matches "musl-gcc" -- as our cross toolchains do. Pass
LIB_ATOMIC=-latomic instead; libatomic is already in DEPENDS. Its stress-fractal
hunk was in any case dead since upstream 24c30b7fa narrowed the row counter back
to int32_t, which MIPS32 and PPC32 increment atomically inline.
Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Many of the binaries in the apache source package do not have a
version output, so we need to add a version override so that CI
can pass.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
Useful for provisioning devices like modems that present themselves
via a serial device.
Signed-off-by: Philip A. Prindeville <philipp@redfish-solutions.com>
Match only explicit default-route subnet tokens when normalizing leftsubnets and rightsubnets.
The previous regex treated dots as wildcards.
Values such as 10.250.0.0/16 and 10.0.0.0/8 were rewritten to 0.0.0.0/0.
Fixes: https://github.com/openwrt/openwrt/issues/23795
Signed-off-by: Dharmik Parmar <dharmikparmar2004@yahoo.com>
Allow CI to pass the version check for packages with executables
with not version output.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
* Update version to latest upstream LTS version.
* The Makefile was using an overly clever means to be compact that the
CI webhook, and possibly the CI test container, didn't understand. We
unroll that logic, and have more usual install and conffiles sections.
In the process we fix a missing conffile install for zbbix-proxy*
packages.
* Whitespace fixups
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
Changelog:
Changes from 1.5 to 1.5.1
-------------------------
* Rename public register definitions on x86 and x86_64 to avoid clashing
with glibc definitions when built in freestanding mode.
Patch contributed by Daan De Meyer.
Changes from 1.4 to 1.5
-----------------------
* Workaround various header compatibility issues in glibc and bionic.
* Allow ppc64 target to be built in freestanding mode.
Changes from 1.3.3 to 1.4
-------------------------
* Add hard-float support to most architectures where relevant.
Some patches contributed by Richard Campbell.
Changes from 1.3.2 to 1.3.3
---------------------------
* Use `bl` instruction instead of explicit GOT lookup for `exit`
in 64-bit PowerPC assembly.
* Fix installation path for standalone libucontext headers.
Patch contributed by Marian Buschsieweke.
* Fix missing includes for freestanding builds.
Patch contributed by David Leeds.
Makefile:
Disabled documents.
Added glibc supports, and remove hardcoded musl dependency.
Put test_libucontext_posix only available in musl runtime.
Build system: x86/64
Build-tested: x86/64-glibc
Build-tested: x86/64-musl
Run-tested: x86/64-glibc
Signed-off-by: Aditya Nugraha <vortexilation@gmail.com>
OpenWrt's .config support declaring qemu-* packages as part of the
Rockchip devices OpenWrt images.
Maintainer: @yousong
Compile tested: snapshot
Run tested: NanoPi R4S
Signed-off-by: David Andreoletti <david.andreoletti.dev@gmail.com>