A very simple build script for bare metal arm toolchain. NO LINUX!
Pull request Compare This branch is 14 commits ahead, 1 commit behind esden:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.




This is fork is still maintained, primarily for the use with the MC HCK, <http://mchck.org>.  You should nevertheless be able to use it for other bare ARM platforms.

To compile the ARM toolchain for barebone ARM devices:
* git clone https://github.com/mchck/summon-arm-toolchain.git
* cd *summon-arm-toolchain*
* ./summon-arm-toolchain
* Profit

Command line options

You can suffix the script call with the following variable parameters:


By default the target is arm-none-eabi, you may want to use a different target, for example arm-elf. Use that option on your own risk it may brake things.


By default the installation prefix is "$(HOME)/sat" you can change it to "/usr" then the binaries will be installed into "/usr/bin" and the toolchain specific
files in "/usr/arm-none-eabi" assuming you did not change the TARGET variable.


When compiling on Mac OS X the build script has to know where MacPorts or Fink
is installed. The default value is "/opt/local".


By default this variable is empty. If you need root rights for the install
step you may set this variable to "sudo".

  $ ./summon-arm-toolchain SUDO=sudo

This will prefix all make install steps with the sudo command asking for
your root password.


By default set to 0. To decrease console output (may increase compile speed
in some cases) you can set this variable to 1.


By default set to 1. To disable the usage of the Linaro GCC and use of the
vanilla GCC instead set to 0.


By default set to 1. To disable compilation of OpenOCD JTAG programming
software set to 0.


By default set to 0. To enable compilation of the non-free libstm32 library
set to 1.


By default set to 1. To disable compilation of the open source libopencm3
library set to 0.


By default set to 0. To enable compilation of binutils and GCC to generate code
for the Cortex-M3 ARM architecture by default set to 1. Tests have shown that
using these options is not very reliable and is discouraged. It is more
reliable to add the necessary options to your build environment instead.


Overrides the autodetection of CPU cores on the host machine. This option
is translated into the -j$CPUS+1 option to the make command when running
the script.


  $ ./summon-arm-toolchain LIBSTM32_EN=1 CPUS=5

This will run the script with libstm32 enabled and with 5 CPUs on your host
machine resulting in calling all make commands with -j6.

Currently tested and known to work target platforms:

* STM32F10x (Olimex STM32-H103 eval board, Open-BLDC v0.1, v0.2, v0.3, v1.0)

Currently tested and known to work host platforms:

* Linux 32bit and 64bit (Debian unstable)
* Mac OS X Snow Leopard with MacPorts

Notes for Mac OS X users:

For Mac OS X there are a few dependencies which must be installed. The wget
and libftdi packages are required and gmp, mpfr, mpc and libiconv are needed
by GCC-4.5.1. These can be installed using MacPorts, DarwinPorts or fink.

  $ port install gmp mpfr libmpc wget libftdi git-core py27-yaml python_select

For XML support in gdb you may want to install expat too. And add the
--with-expat parameter to the GDB target.

Notes for Linux users:
You need to install several packages. On Debian just run:

  $ apt-get install flex bison libgmp3-dev libmpfr-dev libncurses5-dev \
    libmpc-dev autoconf texinfo build-essential libftdi-dev zlib1g-dev \
    git zlib1g-dev python-yaml

You may want to try running the following command instead too:

  $ apt-get build-dep gcc-4.5 git zlib1g-dev python-yaml

For XML support in gdb you may want to install libexpat1 and libexpat1-dev too.

To build a static 32-bit version on a 64-bit system, use

  $ env CC="cc -m32" CXX="c++ -m32" linux32 ./summon-arm-toolchain STATIC=1

Usage notes:

We support multilib now in SAT thanks to Eric Parsonage's and Bernard
Davison's amazing work. You want to use the following GCC flag combinations to
generate full fledged floating point supporting code for some selected
architectures (default is to use software floating point).

* ARM7TDMI-S targets: -mthumb -mcpu=arm7tdmi-s
* Cortex-M0 targets: -mthumb -mcpu=cortex-m0
* Cortex-M3 targets: -mthumb -mcpu=cortex-m3
* Cortex-M4 targets: -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16

If you need support for some other ARM MCU and know the parameters for it,
just drop us a line and we will add that combination to the multilibs.

You can list available combinations by running:

  $ arm-none-eabi-gcc -print-multi-lib

Known Bugs:
- linaro gcc compilation exits without an apparent error
  Sometimes the compilation of GCC exits without any error, RERUNNING the sat
  script manages to compile a working gcc though.

Stable tarball downloads and GIT cloning

The script has a section containing version numbers of the respective packages.
Some packages have beside the stable release version number also a variable
with the suffix _GIT. If that variable is set to either a branch name or a
revision SHA the script will try to clone the repository and checkout the
respective branch or SHA instead of downloading the stable version.

If you provide a GIT branch name the script will contact the GIT server and
query the associated SHA. If this SHA is already present in the sources
directory it will not attempt to reclone the repository.

How to submit improvements and patches

As more and more people start to submit patches and improvements to
Summon-Arm-Toolchain (SAR) this section seems to become necessary.

First of all any way of submission is appreciated, if you just want to dump
your version of the script to us feel free to do so. Still if you want your
improvements and fixes to go upstream quicker there is a good way to do that.

1) Create an account on GitHub (or some other git hosting service, if you
   really have to).
2) Fork the main SAR repository.
3) Clone the forked repository to your disk.
4) Change the script, try to make small changes adding one feature or bugfix at
   a time (that makes review much easier for us).
5) Push your changes to GitHub, or the other service you chose.
6) Test your changes, by compiling the toolchain (you probably want to do that
   with different options).
7) Make sure that everything still works.
8) Test a little bit more.
9) Click on "pull request" on GitHub or drop us a line so that we can pull your

I know that sounds like a lot of work, but if you don't we have to do it and
that means that your awesome improvement or bugfix will take longer to be
integrated into the official script. And as you, we want everyone to profit
from such changes sooner then later. :)