target_config.lua and target_config_check.lua don't pass a table of
callbacks anymore, so target_config_lib.lua can by simplified by moving
all the code that was in the returned function to the toplevel.
So far, we were using a sort operation on the generated .config to
implement precedence of =y packages over =m, and =m over unset.
Unfortunately, this sort not only used for packages, but for all config
lines. This made it impossible to override settings from targets/generic
in a target config when the new setting was sorted before the generic
setting.
To fix this, track configurations by their keys, so we can properly
override config keys that were set before. Value-based precedence is
only preserved for package configuration.
The config() and try_config() calls always take key and value as
separate arguments now. Strings are quoted automatically; the values
true, nil and false map to y, m and unset for tristate options. config()
can take an optional third argument to override the error message to
display when the setting fails to apply.
All existing target configs generate the same .config with the old and the
new code. The new code is also a bit faster on targets with many devices.
In OpenWrt 19.07, the block size was increased to 1024KiB for "tiny"
devices by default to save flash. Unfortunately, this also significantly
increases the cache memory required by squashfs.
In my test, the increased block size reduced the image size by ~64KiB,
but increased the RAM usage by ~2.6MiB. As most tiny devices have only
32MiB of RAM, this is not a reasonable tradeoff.
The ar71xx-generic target already defines an even lower block size of
64KiB.
By passing the package name through merge_packages, it becomes possible to
override the package choice in GLUON_SITE_PACKAGES again, for example:
GLUON_SITE_PACKAGES += -hostapd-mini hostapd
A bug in batman-adv can lead to a large amount of management traffic being
exchanged between nodes when the multicast optimizations are enabled,
effectively making the mesh unusable. It's safer to disable the feature
for now, until we have a real fix.
sha512sum doesn't add much code that is not also used by sha256sum, but the
change of the configuration hides the segfault issue described in:
https://bugs.lede-project.org/index.php?do=details&task_id=822
While the issue only seemed to affect dhcpv6.script, it would clutter /tmp
with coredumps, eventually leading to OOM.
