Pino Toscano <ptoscano@redhat.com> [2018-08-01, 03:21PM +0200]:
Hi Bjoern,
On Wednesday, 1 August 2018 14:57:07 CEST Bjoern Walk wrote:
And here's the fix for the viriscsitest on big-endian machine like Daniel suggested.
I sent an alternative fix, only for viriscsitest, while you sent this patch.
IMHO virRandomBits() ought to not deal with endianness, since all it sees are random values, without any meaning. The problem was specific to the test, because it assumed a certain order of the data returned by the mocked virRandomBytes().
Yes, I agree. That's why I hesitated sending this fix. But this was also the solution that Daniel suggested and maybe there is value in having the random bit generation endian-safe as well (although I don't see one currently...). I actually don't care how we fix it. I will test your patch and report back.
-- Pino Toscano