If the user defines a $max_timeout of 30, the service will wait 30 seconds
before it considers lxc-stop complete even though lxc-stop might actually
finish much sooner. This introduces an unneeded delay.
This commit changes the behavior to check once per second to see when lxc-stop
actually stops doing so up to $max_timeout. It also slightly simplifies the
code with logic to append the -t $max_timeout to the script.
Signed-off-by: John Audia <graysky@archlinux.us>
A new server was added which runs within the OpenWrt cloud, it's much
faster and should be used instead. For development the server at
https://asu.aparcar.org stays available.
Signed-off-by: Paul Spooren <mail@aparcar.org>
I am unaware of any kernel currently provided that retains cgroup v1 support.
This patch removes these lines in /usr/share/lxc/config/common.conf to allow
for error free usage.
Providing common.conf as-is will result in failure to start. One solution is to
comment out the legacy lines.[1] This requires users to either provide a custom
version of this file on their builds or to manually edit it with each update.
Since many do not build their own, the first option is not available to them.
Manually editing the file with each update will cause a failure to start
containers set to auto-start upon rebooting into the update.
1. https://forum.openwrt.org/t/openwrt-arm64-quick-lxc-howto-guide-lms-in-debian-system-in-lxc-container/99835
Signed-off-by: John Audia <graysky@archlinux.us>
Earlier versions of podman did not make use of TMPDIR when running "podman
run ...". Podman's default, /var/tmp, presents a problem to rootless
use since OpenWrt's /var/tmp does not permit writes by non-root users.
Podman 3.3.1 makes full use of TMPDIR.
This is part of an attempt to get rootless podman to work on OpenWrt.
See https://github.com/openwrt/packages/issues/15096.
See also the upstream issue at
https://github.com/containers/podman/issues/10698.
Signed-off-by: W. Michael Petullo <mike@flyn.org>
One of big change is that QEMU has switched to Meson build system.
That result in few changes to build scripts to fix python interpreter
usage.
Second change that it's not possible to select binaries to build, so
now we have to build all targets at once (that require --enable-tools).
Options --disable-sheepdog and --disable-vxhs was removed from qemu
6.1.0 and 6.0.0 accordingly.
Signed-off-by: Vladimir Ermakov <vooon341@gmail.com>
They don't use python3-websocket-client since 2014, but they still list it in
dependencies: https://github.com/docker/compose/issues/8455
The dependency is still pulled in through python3-docker
* switch to AUTORELEASE
Signed-off-by: Michal Vasilek <michal.vasilek@nic.cz>
Use /srv/prometheus instead of /data, because user `prometheus` doens't
have permissions to create `/data/` in `/`.
Instead this commit puts prometheus data into `/srv/prometheus` by
default, which is a cleaner path, and it'll create tsdb path & then assign
required permissions on each prometheus service start.
This way, also, the cases when users re-configure tsdb to point to
external USB - it'll also be created and assigned required permissions for
prometheus.
Signed-off-by: Alex Simkin <sashasimkin@gmail.com>
Some versions of killall do support the `killall -SIGNAL` syntax and
have only `-s SIGNAL` which should be supported everywhere.
I see the problem with *killall (PSmisc) 23.3* on latest TurrisOS 5.2
Signed-off-by: Jan Baier <jan.baier@amagical.net>
Some versions of killall do support the `killall -SIGNAL` syntax and
have only `-s SIGNAL` which should be supported everywhere.
I see the problem with *killall (PSmisc) 23.3* on latest TurrisOS 5.2
Signed-off-by: Jan Baier <jan.baier@amagical.net>
Use POSIX shebang for egrep/fgrep instead of the variable/Bash.
The issue affects both master and openwrt-21.02.
Signed-off-by: Vladislav Grigoryev <vg.aetera@gmail.com>
Change the sensors plugin's dependency from libsensors
to lm-sensors, so that also the /usr/sbin/sensors executable
gets installed if the plugin is selected.
Although that is not needed for collectd itself,
the LuCI statistics app expects to find the sensors executable
if the plugin has been installed.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
The "-s -w" flags in GO_PKG_LDFLAGS tells the Go compiler to strip the
binaries it produces. Since the default Go package build process will
strip binaries when CONFIG_USE_STRIP or CONFIG_USE_SSTRIP are selected,
these flags are unnecessary.
When CONFIG_NO_STRIP is selected, these flags override the user's
intention of building unstripped packages.
This removes these flags for all relevant packages.
Signed-off-by: Jeffery To <jeffery.to@gmail.com>
Upstream announced new release which includes fixes for 21 CVEs (see
[1]).
Upstream changed the symlink "mkfs.ntfs" to respect "--sbin-dir". So
this commit adjusts this likewise in the Makefile.
[1] https://www.openwall.com/lists/oss-security/2021/08/30/1
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
As /var/state is used for UCI state, move persistent config for uxc
from /var/state/uxc to /var/run/uxc.
uxc itself has already been changed to look for it there.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Remove getline patch. It seems to be for uClibc, which is no longer in
the tree.
Remove commands patch. Issue was fixed upstream. Same with the tests
patch.
Remove gpg patch. It's an upstream backport.
Refreshed others.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
A stray ')' made auc skip the check if all mandatory fields of a
branch are present. Fix paranthesis in that if-clause to actually
check for all fields.
Don't even fetch package lists if checking only for release.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
* use regular verrevcmp for branch names, fixes update indication
* break out usage() from main() function
* require '-f' in case installed package is missing in remote lists
* don't use ENETUNREACH for unlrelated error
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Remove dependency on usbids as lsusb binary does not use usb.ids file
anymore, instead it uses udev hardware database.
Signed-off-by: Kuan-Yi Li <kyli@abysm.org>
Contains fix for time jumping back 1024 weeks after 23-October-2021.
Changes:
- Between current version and 3.23 the ABI version changed from 27 to
29. Add ABI_VERSION variable to package as the jumps occur often.
- 0001-no-distutils-import.patch: removed because current patch doesn't
work with new release and host python3-distutils is a build
requirement anyway.
- 0002-ncurses6_detection.patch: fixed the patch. Currently the first
"elif" checks for "ncurses6-config". But this is only provided by
ncurses/host package, so the CFLAGS then contain includes that point
into $(STAGING_DIR_HOSTPKG). The ncurses target package only installs
"ncursesw6-config" into the staging area. So this is the only tool of
interest. Also, the hunk moved from SConstruct to SConscript.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
The server API was unified and therefore the servers response is now
`detail` rather than `message`, all found under the new endpoint
/api/v1/build.
Signed-off-by: Paul Spooren <mail@aparcar.org>
