Hi Julien,
On Sat, Jul 15, 2017 at 08:25:17AM +0200, Julien Yann Dutheil wrote:
> Yes, indeed. The other test that somehow fails was known to us. It is
> actually badly designed because it relies on simulations, and has a 5%
> change or so to fail just by chance. I guess that is time to do sthg about
> that.
It would be very sensible to make this at least a non-failing test if it
bears the chance to fail.
> Yet I am puzzled about the mips thing: the failling test performs
> some calculation which just gives a very different result on this
> architecture, and I have no idea why :s
Me neither. Perhaps there is a chance to consult
debian-mips@lists.debian.org.