[libvirt] [PATCH 24/24] tests: qemumonitorjson: Test extraction of iSCSI device node names

Peter Krempa pkrempa at redhat.com
Wed Aug 2 13:39:25 UTC 2017

On Wed, Aug 02, 2017 at 13:50:49 +0200, Bjoern Walk wrote:
> So, this test fails on S390 because the actual test in
> testBlockNodeNameDetect is dependent on the ordering of the entries in
> the hash table, which is different on big endian machines.
> There are two other tests which have multiple results in the backing
> chain, 'same-backing' and 'relative', but their keys are reasonable
> similar that we got lucky and the hashes are ordered the same.
> I don't suspect that code that uses the tested function,
> qemuBlockNodeNameGetBackingChain, is dependent on the ordering of keys
> so this is actually only a problem in the test suite.

Yes, the hash function is non-deterministic. I'm working on mocking it
so that it uses a simpler algorithm which will be endian independent.

> Sorry for the late notice for this.

Well, since the issue is not in this series, but the other tests needing
specific ordering were lucky enough to be ordered the same way, it was
rather hard to catch.

> Bjoern

> --
> libvir-list mailing list
> libvir-list at redhat.com
> https://www.redhat.com/mailman/listinfo/libvir-list

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20170802/83a81402/attachment-0001.sig>

More information about the libvir-list mailing list