This typo caused that non htt drivers were always installed.

Karel Koci authored 4 years ago and Karel Koci committed 4 years ago

This disables root account.

Placing of this patch expects it to be caried for a long time so it was
moved to second possition while current second patch was moved to thid.
That makes it of same number as existing patch but that patch is going
to be dropped in future so no harm to have them with same number for
now.

And I just forgot to rename it so it exactly happened that. That means
update of NOR in all routers tracking HBK.
With subsequent build it is going to break so we have to do this rename
anyway.

This fixes fourth LAN port on running Turris Shields.

This effectivelly fixes contract variable on Turris Mox (Shield)
routers with rescue update.

Fields should be compared case insensitive as they can be in any such
format. Using awk instead to compare and correctly parse HTTP head
format.

There is known problem with this code and that is it can't parse
date-time formats with colons. The reason is that printing original
string ($0) in awk sneaks in to it some unwanted characters such as \r
and probably others that are not considered as whitespaces by date
command. This way we take only text.

Michal Hrusecky authored 4 years ago and Karel Koci committed 4 years ago

Settting timezone doesn't work without those.

The condition is that we run on current root (that we even have easy
access to snapshots), that we have snapshots (schnapps installed)
and that version of Turris OS is less than 5.1.1.
We use system version to compare because we can't detect that empty file
any other way.

There might be some settings later on, it is kinda fair to give people
an option to opt out from Sentinel and also dashboard shows warning
about datacolect if the page is not present

Netmetr is nice to have and dev-detect just generates notifications, so
harmless as well.

Standard for_l10n function provided by root updater script is creating
requests with default priority 50. This effectively increases priority
of all packages using it to include their languages.

The solution here is to decrease priority for language packages to
something very small (10 was chosen).

Also because we had to have copy of for_l10n function in bootstrap it
was instead moved from there to utils and provided as our own version
instead of updater's one.
Original function can potentially be dropped later on as Turris lists
are no longer using it thanks to this change.

Atlas and netdata services are new services that can be installed trough
package lists. They should be also automatically activated.

sentinel-dynfw-ipset is not longer present in system as it was drop in
favor of different approach.

socat is not longer used as service since uCollect times.

Michal Hrusecky authored 4 years ago and Karel Koci committed 4 years ago

Add few important packages to be always build on Turris 1.x like
turris1x-support, rainbow and atsha.

Michal Hrusecky authored 4 years ago and Karel Koci committed 4 years ago

Making sure some important packages for Omnia build are always select
and build. Like omnia-support, rainbow and atsha.

Michal Hrusecky authored 4 years ago and Karel Koci committed 4 years ago

Add mox-otp, mox-support and mox-dtb as packages that has to be included
in every build to make sure they are present.

Variable was renamed but not in this specific instace.

This fixes bug that caused normal medkits to not be generated as
contract with no name was not located.

The reason is that Lua's string when they are empty are still true
(contrary to other languages such as Python). This adds explicit checks
if they are non-empty as that is more common and more expected.

This also fixes potential problem when some environment variable is not
defined. This does no happen with our script but can happen if that
script is not used.

This allows contract specific medkits to be generated using this script.

Yeh it is that stupid of a syntax.

There were two issues in this code.

First issue is that options might not be specified and they were not
then initial match returned nil that caused gmatch fail. We should have
check if we have some list options and run loop only if there were some
specified.

Second problem was simply in initial regexp. It was eating up closing
parentheses for options so no options were ever found.

Michal Hrusecky authored 4 years ago and Karel Koci committed 4 years ago

https://github.com/openwrt/openwrt/pull/3243