On Wednesday, 1 August 2018 15:24:05 CEST Daniel P. Berrangé wrote: > On Wed, Aug 01, 2018 at 03:08:07PM +0200, Pino Toscano wrote: > > viriscsitest tries to ensure the interface IQN used is a specific one, > > checking later on that it is the same all during the whole test. Since > > the IQN generation involves random bytes, viriscsitest got a fake > > virRandomBytes from the virrandommock helper library, setting static > > values. virRandomBits(), called by virStorageBackendCreateIfaceIQN(), > > always requests 8 random bytes, chopping off the ones not requested by > > the caller -- this meant that on big endian machines it would chop bytes > > from the wrong size of the data buffer, and thus not returning the > > expected numbers. > > > > As a fix, do not rely on the mock virRandomBytes, but provide an own > > version of it: this version will fill the values in the expected order, > > depending on the endianness of the system. This way, the result of > > virStorageBackendCreateIfaceIQN() will be what the test actually > > expects. > > > > Signed-off-by: Pino Toscano <ptoscano redhat com> > > Mailman is having deliver problems so you might not have received > the patch that was sent earlier for this problem: > > https://www.redhat.com/archives/libvir-list/2018-August/msg00039.html > > I prefer that approach since it avoids the endianness problems for > all future usage too. I replied to that patch, as IMHO there is no "endianness" for random data, and thus virRandomBytes() ought to not do any manipulation. -- Pino Toscano
Description: This is a digitally signed message part.