On 24/02/16 10:19, Sam Thursfield wrote:
I just had a thought regarding the practicality of the "native compile
inside QEMU" approach, I want to share in case I forget again.
We have at least one commercial user of Baserock who aren't able to
install Linux on their developer machines. Currently they run Windows,
and then run Ubuntu 12.04 or 14.04 inside VirtualBox, and then a
Baserock x86_32 chroot inside that, and then they run Morph inside that.
They use 'distbuild' for ARM builds, so the actual ARM builds happen on
a separate ARM server machine.
Do you have any idea if there's any chance of the YBD+Aborginal approach
working for them? I imagine qemu-arm *would* work inside VirtualBox, but
the performance might be too abysmal to contemplate.
Sadly I think it's still pretty common for IT departments in big
companies to prevent employees from installing Linux on their machines,
so we probably do need to consider this case
I could imagine the approach Tristan is taking would work excellently on
a large powerful x86 server (especially with the distcc aspect) - could
be even better if the qemu tcq-multithread work pans out . It also
seems that this is possible to run as non-root from Tristan's recent mail.
So, simplest approach for this class of users would simply to be to
provide logins on a largish x86 build box! (and the box can be managed
by central IT/IS)