site  contact  subhomenews

T2 on the Raspberry Pi3

December 02, 2016 — BarryK
I have used T2 to compile Puppy and Quirky from source, for i486, i686 and x86_64 target architectures.
In the Puppy days, there was Wary and Racy Puppy, with Quirky there was the April series.

I posted about the online Bones repository for T2, for building Quirky April, back in Jan. 2015:

The web page that introduces Bones and explains how to download T2 for April, is here:

I haven't looked at T2 much in the last couple of years, but one thing on my to-do list has been to try a native ARM build.

Which is what I am doing right now. Now that I have my 1TB hard drives working on the Pi3, there was my T2 project on the drive, tempting me to have a go.

I am using the old T2 in my Bones repository, rather than the latest from, as mine is heavily customised. Will think about upgrading T2 later.

I chose the "embedded" target in T2, and "cortex-a8" CPU as there isn't a "cortex-a7" option, nor anything else that looks reasonable for the Pi2 and Pi3. Sticking with 32-bit.

Let it run overnight, had to get up a few times to fix something. This morning, it is still in "Stage 0", which is the cross-toolchain building stage. But, it is doing well, I am actually quite surprised.


T2 compile got to stage 1, glibc failed but I fixed that.

Then dietlibc failed, and this is where I am stuck. Heaps of undefined symbols.

There was one that I fixed, as documented back in January 2016:

After that, I got overwhelmed by them. Back in Jan, it was a x32 cross-build, this time it is a armv7 native build. Same problems.

In fact, I tried to compile dietlibc directly in Quirky, same problems.

Dietlibc does not like the headers in Quirky! Fiddled around for awhile, have given up.

The same problem that is causing dietlibc to be broken in Quirky, is also the same reason why dietlibc won't compile in T2.

Tags: linux