Piecing together a development environment
It has been a while since my last post for good reason I have been mighty busy. Now that things have settled down I have finally gotten the chance to start to piece together my embedded development environment. Embedded development is a quite an interesting beast in that many of the development concepts are quite behind standard desktop development. Overall I have come to believe it is this way because quite honestly embedded development is incredibly low level. There are really no huge api's in existence because abstractions really do not help with portability as no matter how well abstracted you still need heavy modifications for cross target support due to various CPU and peripheral features being located at different memory addresses etc. So in this respect I think there was never really a need to build massively robust software tools to develop on typical 8 bit and 32 bit micros.
So my particular development platform of choice is my new MacBook Pro. This machine is amazing quite a beast. The reason I chose a Mac over a PC with Windows is quite simple. Despite Windows having quite a following in the IDE department for embedded development Windows is still a very gimped platform. Every embedded toolchain for instance uses make files under the hood and these are GNU makefiles running on GNU make. The various IDE vendors ported make over to windows themselves and distribute it with the IDE. This actually makes the build process quite slow because make was really designed around POSIX. As I said previously embedded development still uses quite a few old concepts and the main reasoning behind this is the arcane architectures and the need to be able to select where code goes in memory and it just so happens that GCC, Make, and Linker Files are still the best way to do this. So my main reason for choosing make was the "It Just Works" system with the strong UNIX core that provides POSIX features and a powerful terminal like bash. It really is a win win as you no longer have to worry about crap breaking, not working at all, or various hardware incompatibilities that come with Linux which is getting better but still horrible.
So now that the machine is out of the way we need tools to use. The first obvious tool you need is a GCC cross compiler for ARM. For those that do not know a cross compiler is a compiler that runs on one system type say a PC but instead of generating machine code for that machine it generates machine code for a different architecture which allows you to even do embedded development at all. Without cross compilers you would never really be able to develop for these small chips as you typically can't run a PC like OS on the chip to compile your code. This is a simple task all you need to do is download the compiler set which includes everything you need like GDB, GCC, G++, Linker, Assembler, etc... All you do is download, extract to a directory and add the compiler to your path and you are done.
The next task is needing a GDB server for GDB to connect to for remote debugging. In order to debug hardware related code it needs to run on the hardware. You also need to be able to get the binary burned into the chips memory. Most ARM development boards come with a programming/debugging module on it already. This module can typically burn the chip on the development half of the board or also burn to a external chip via certain pin hookups. Still to operate these features you need another piece of software. In my case for maximum compatibility and to be able to use the same tool for possibly different chips I chose OpenOCD. On Linux/Mac/Windows OpenOCD needs to be compiled. There are sites that provide binaries for Windows but this often is not needed because the vendor usually has a tool ready for windows. On linux/mac OpenOCD or a tool someone else wrote like stlink made by a ST employee is required. On Mac open OCD can be taken care of quickly with the homebrew package utility. This allows for not only a debugging server but also a interface to burn your code to the chip.
Over all that is all that is needed besides driver code like CMSIS or Vender supplied libraries. When I say driver code it is not what people think of as a driver. All driver code is are various source files and headers which pre map peripheral and cpu memory addresses for the chip in question. Think of it more like a very tiny and low level API. Then you need the programming manuals, reference manuals, and datasheets.
As for IDE's on windows there are tons of choices. Many are quite expensive but there are a few free ones that work relatively well. On any platform you can easily use Eclipse with CDT and maybe 1 or 2 embedded plugins to handle this. Then there is always the non IDE route using a text editor like Emacs or VIM. This is a decent option considering you are not really working with large and confusing API's like you would be in C++, Java, or C#. The api's are very slim so "intellisense" is not paramount. I have not chosen what I am going to use on this front quite yet. Like always there are heated debates in this camp some saying Eclipse is the way to go and others saying Vim and Emacs are the way to go because you should know how your tools work for when stuff breaks.
I am not much for heated debates so I will figure out what I want to do here I will probably end up going with Eclipse because quite honestly I hate having to configure every little tiny piece of my editors.
That is all for now have fun and write awesome code.