strance error when running Nova + Genode + Seoul (scenarion : seoul-auto.run)

[Alexander Boettcher]

Hi,

On 14.10.2015 12:44, Parfait Tokponnon wrote:
> Yes, I am running Ubuntu from within a VM (VirtualBox). Is there any
> problem related to it? I didn't know we cannot test the platform from
> within a VM...

developing within a VM is no issue, but actually running the result from
within a VM is adventurous - at least - most the time, based on
practical experience.

In your case you have 3 level of virtualization/emulation nested - you
are running Ubuntu in Virtualbox (1), then you run Genode/Nova inside
Qemu (2) ontop of (1) and finally you run a Linux VM with Seoul VMM (3)
on top of (2).

Beside the obviously waste of CPU time and increasing your idle time,
you have to rely on all upper levels to work correctly. Such a
assumption and in principal such a development setup just leads to
various types of frustrations - I would try to avoid that.

> 
> And Yes, when I had invoked qemu directly with the provided command in the
> log file it worked too.
> So what could be the explanation?

Our run tool sets for automatically tested scripts (as seoul-auto.run) a
timeout until it must have been finished. Since you say it works for you
when you are calling qemu directly - it could be that the execution time
in your 3-level nested virtualization setup is too short.

You may try to increase the timeout in seoul-auto.run (search for
run_genode_until) or try using seoul-net.run instead, which is mainly
the same setup but without timeout.

Does this change things for you ?

If not, you would have to find out why tcl/expect stops working with a
eof (in <genode-dir>/tool/run/run) error when waiting for qemu output.



Regards,

Alex.