Compilers needed

[traisen]

Just noticed target is arm-unknown-gcc is that normal?

[traisen]

checking target system type... arm-unknown-elf

[skoehler]

That is normal.

[gloomyandy]

Hi,
You could always install something like VirtualBox on your Windows system and install a Linux system inside of that. I suspect this may be a good way forward as it would give you an easy way of having your students use the same environment (you could simply give your students the VM)...

Andy

[skoehler]

I have to say, that it is probably easier to do that Andy suggests.

However, if you want to further debug this, change the MAKEOPTS variable in the script from -j2 to -j1 and run the script again. Please capture the output. Then upload the output somewhere (e.g. pastebin) and post the link here.
To redirect the output, you do the following:
/path/to/build_arm_toolchain.sh 2>&1 |tee logfile.txt

BTW: the script compiles and installs the script inside the current working directory. You should really create a new, empty directory somewhere, cd into that directy, and then execute the script from there. It will create the build, src, install in that folder.

[traisen]

Thanks. I agree I should try this on linux even in virtualbox next. With cygwin at moment it frustrating wondering what is not quite right.

To get on with my real task used the gnuarm toolchain 4.1 and compiled /created a .bin file using make in target/nxt.
And it worked!

I still want to build and use the toolchain Lejos to get the lejos optimizations, but I can move forward and work in parallel on my task and building the lejos specific toolchain.

Thanks again

[skoehler]

To get on with my real task used the gnuarm toolchain 4.1 and compiled /created a .bin file using make in target/nxt.
And it worked!

The question is: does it run?
I once tried to use a different toolchain, and it crashed horribly. No idea why. The problem may be in the linker script, missing initialization of libc, or whatever. I might have been related to an issue related with newlib. I never researched that properly.

[gloomyandy]

Some parts of the firmware also make assumptions about things like floating point memory layout and memory layout in general which may have changed with more recent versions of gcc. Also I seem to remember that we lost a fair bit of performance with more recent versions of gcc. Not really sure why, There are often big jumps in things like code size when you move between different versions as well I suspect that small changes in the code generation logic can have big impacts on some of the leJOS code due to things like the very large switch statement that forms the core of the interpreter.

[traisen]

Hi, I hear you both. And have Ubuntu Virtualbox... is Ubuntu ok? What I am used at moment, but can change. Debian based.

Part of what I need to do I just need code that compiles to make progress.
But sounds like my original plan of compiling all of Lejos and check it works on a NXT is still right way to go.
Thanks.. I'll keep you posted.