Re: QEMU build dependencies for new board

[Peter Maydell]

On Mon, 23 Nov 2020 at 17:31, Ancuta, Cristian
<cristian.ancuta@intel.com> wrote:
> I’ve implemented a new CPU target in ./target/arch_name and I’m also trying 
> to add a new board to emulate that target on in system mode in 
> ./hw/arch_name. The board is based on the versatilepb, but I’ll gradually be 
> removing all the arm implementation from it, aiming for a minimal 
> implementation with just my custom cpu architecture, system bus, main memory 
> and an UART.

Incidentally, versatilepb is a really bad board to start with
as a template, because it's one of the oldest we have, and it
does a lot of things in ways that work but which aren't how
we'd recommend writing a new board model today. You might be
better off looking at something added more recently.

> I’ve also added all the necessary stuff required by minikconf.py in
> ./default-configs/<target>-softmmu.mak,

This suggests you're not basing this on current head-of-git,
because this is default-configs/targets/<target>-softmmu.mak
and default-configs/devices/<target>-softmmu.mak now.

> ./hw/<arch>/Kconfig and a source entry in ./hw/Kconfig.

> The problem is that the build system is bringing in other files
> that I didn’t specify (CONFIG_A15MPCORE, CONFIG_A15MPCORE,
> CONFIG_9MPCORE, etc. ), and I’m not sure how they’re ending
> up in ./build/<target>-softmmu/config_device.mak:

This shouldn't happen, but it's not really possible to identify
the exact problem since you don't provide your code. I would
try first doing a complete build from scratch (ie delete
the build directory) in case the problem is that there are
stale files in the build tree that are getting picked up.
Otherwise re-double-check your default-config files to make
sure they really don't have any CONFIG_whatever or "SELECT whatever"
in that they shouldn't. (Looking at the list of devices, an
accidental "select REALVIEW" would have that effect.)

thanks
-- PMM