So the philosophy here is that there will be one main fluent-bit
instance, which will read from /dev/kmsg and /dev/log and write to
/var/log/messages.
This is done to reduce the chance of introducing errors in config
and making sure that var/log/messages is always populated because
logread depends on it.
For example, an unavailable URL for a udp based syslog output plugin
can cause fluent-bit to exit entirely.
The other fluent-bit instance uses a different "user" config file.
Which will handle log_file, log_remote and included configs. It has
a single hard coded input tail plugin which reads from
/var/log/messages.
Any filter that is applied will act on the main config and thus,
only filtered logs will be available to both fluent-bit instances.
for example, if proto is udp in syslog output and url is not
resolved then fluent-bit stops, we want to make sure that
eventually when the url can be resolved then logging resumes
Change to zstd compressed instead of gz-compressed tarballs. zstd
compresses 7-8 times faster and decompresses 2-3 times faster.
This change also make sharing tarballs between different branches
possible, because OpenWrt 25.12 uses zstd as default.
Signed-off-by: Andreas Gnau <andreas.gnau@iopsys.eu>
according to libco documentation: "Heapsize is the amount of
memory allocated for the cothread stack, specified in bytes. This
is unfortunately impossible to make fully portable. It is
recommended to specify sizes using `n * sizeof(void*)'. It is
better to err on the side of caution and allocate more memory than
will be needed to ensure compatibility with other platforms,
within reason. A typical heapsize for a 32-bit architecture is
~1MB.", so we follow this advice in hopes of reducing crashes.
Always migrate configuration. Do not set executable bit on config
files and uci-defaults scripts. Remove unused /lib/functions.sh
include. Remove redundant "default y" on choice.
If it fails during MP the verfication of buttons will not work
as expected. The problem is that the file is created before
data is commited and when power is cut prematurely it won't
re-create the file, maybe because the uci-default script has
been removed already as well.
