Building Cerowrt on your own Linux machine

NOTE: These instructions are totally out of date and have been replaced with a new script.

Prerequisites

You need some packages that you may download the sources. Then you need some build dependencies to be satisfied. Then you need to checkout the sources.

sudo apt-get install git-all subversion build-essential flex libncurses5-dev libz-dev gawk gettext bison texinfo
mkdir -p ~/src/prep ~/src/cerowrt
cd ~/src/prep
git clone git://github.com/dtaht/cerofiles-3.3.git

There will be a build_cero.sh script in that checkout. There is also a cero_config file here. Copy cero_config to ~/.cero and then edit it to suit your preferences.

cp cerofiles-3.3/cero_config ~/.cero
vim ~/.cero # use the editor you are comfortable with

This example assumes the use of folder ~/src/cerowrt as main directory so you should change the CERO_DIR variable to

CERO_DIR=~/src/cerowrt

then run

~/src/prep/cerofiles-3.3/build_cero.sh init
cd ~/src/cerowrt
# in there will be your checked out dir, usually wndr3700v2, cd into that
make # But please see note below for an extra step to save on download hassles

The first time you run this it will take a VERY long time to build the toolchains, etc. I have generally found -jX to fail on large numbers of processors, so just build it in series and go to a long dinner or lunch. Or…

make -jX
make V=s

Subsequent runs will be much faster, and -jX usually works. When it doesn’t…

make package/the_breaking_package/{clean,compile,install} V=99 

usually will work or show the problem.

NOTE With kernel.org down last year it was impossible to get a build to summon and compile all the needed packages. There is a temporary mirror of the needed packages available via rsync. It’s still faster to just snag all the tarballs this way, so… inside the build dir above:

mkdir dl; cd dl; rsync -av huchra.bufferbloat.net::dl .

(this is a temporary expedient and not actually necessary in normal use)

Check your build

If the wndr3700’s bin/ar71xx/ has images that are less than 5MB, you didn’t get a good .config file, and/or some major packages are missing, and/or the phase of the moon was wrong.

Your final images should be about the same size as on huchra.

Notes:

Given that this builds most of cerowrt against the main git head trees, your build will differ somewhat from what we are making available as release candidates at present. After we get stable, we will tag and freeze the relevant trees.

We note that due to this difference, there may be slight (or even major) differences in the default .config file. It helps to repeatedly cp env/.config onto .config and run make defconfig, and install packages, etc, until they more closely match.

We will improve this script as time goes on, and if you run into trouble, please contact the #bufferbloat irc channel and/or the mailing lists.

Keeping updated

The procedure above makes it possible to track all or parts of openwrt and cerowrt, by keeping multiple repositories around.
You can,for example, merely track ceropackages, by doing a git pull in the ceropackages directory and a ./scripts/feeds update in your build directory. Or setup something to track the mainline repos (this is what I do), do integration, and builds, and not bother the ongoing work in your release process. Doing continuous integration is difficult, but necessary if you wish to advance the state of the art.

changing your cerowrt ip addresses changing your dns domain