Turris OS building scripts

apt install ca-certificates git build-essential zlib1g-dev gawk libssl-dev \
 subversion unzip libncurses-dev wget python3 file rsync gcc-multilib \
 g++-multilib libc6-dev-i386 protobuf-c-compiler which libyaml-dev

Sometimes you want to just get OpenWrt build system. That is possible with the following command:

Where -t means -target, so the only valid values are turris1x, omnia, mox.

This gives you fully patched OpenWrt build system. The "patched" here is important. You should not use scripts/feeds script for fetching feeds as that can remove local patches. Instead to refetch feeds you can use ./compile_pkgs repatch_feeds.

To build a package you have to first prepare build directory. That can be achieved by running in target directory.

Where -t means -target, so the only valid values are turris1x, omnia, mox.

Once, you have compiled required tools, you can build packages using this command:

The name is package name, which you want to compile.

You can also clean single package just by replacing compile with clean.

The resulting package is placed in directory bin/packages/ARCH/REPO where the ARCH is target architecture specific string and the REPO is the name of package’s source repository.

Occasionally, you can meet cases, where you need to run make package/name/compile with parameters -j1 V=s or -j1 V=sc to be able to see, what went wrong.

The lists are rules for Turris packages manager called Updater-ng. They are processed using M4 macro language. To generate the lists you have to run:

The output is placed to directory generated_lists (feel free to use some other path).

Generating medkits is using Turris package manager Updater-ng and using for signing.

Medkits are always generated against official Turris repositories. There is no easy way to generate medkits against locally compiled packages and generated lists. This also means that you don’t have to generate lists nor compile packages to generate medkits.

Requirements to be able to compile Updater-ng and usign.

If you installed all the requirements for Updater-ng, you can now use following command:

Where -t means --target, so the only valid values are turris1x, omnia, mox.

Default config for OpenWrt is put in configs directory either in directory common or in a directory named after the target board. Files in those directories get merged (common ones first) and will form the final .config file. Some parts of compile_pkgs might modify it further.

There is a NEWS file in this directory that contains the version number and after it lines with new features in the release. This file is used to generate turris-version package. compile_pkgs script reads it, finds the newest version and from notes underneath it creates the package with release notes.

When new version is released to Snails (see workflow for explanation) then new commit with hashes has to be created and pushed. This commit have to be tagged with appropriate version tag in format vVERSION where VERSION is released version. To make this all simple and to correctly set always all configurations we have script helpers/new_release.sh. Run this script after release from turris-build project root directory and it is going to automatically detect, commit and tag new version. You should review commit and tag it created for you and then push it using git push --tags.

Script helpers/new_release.sh can generate various errors and warnings. It is advised to run it even before release in verify mode to review possible problems with release.

According to workflow new releases are forked from parent branch. During this process care should be taken to tweak defaults to appropriate values. Following list should be taken as a checklist for new release branch.

1. Set PUBLISH_BRANCH in defaults.sh. Master branch should always be set to hbd and release branches should be set to hbs.

2. Set branches in feeds.conf. You should append string like this: ;openwrt-18.06. This has to be done for all OpenWRT feeds as well for OpenWRT it self. URL used by compile_pkgs is specified as first line in feeds.conf and with exception of first column it has same format as feeds. Note that that line is intentionally commented out as that is not feed.

If you need to patch OpenWRT or any of the feeds, put a git formatted patch into the respective subdirectory of patches directory. In patches directory, there is openwrt directory and directory for each feed we are patching. Inside those directories, patches are also divided into subdirectories like to-upstream, wip or hack to somehow distinguish between the quality of patches.

Rebasing patches

find . -name '*.rej' -delete                # Clean all existing reject files
git am --reject .../patches/.../xyz.patch   # Try hard to apply patch
find . -name '*.rej'                        # List all failed chunks