lubuntu-qa team mailing list archive

[Jonathan Marsden <jmarsden@xxxxxxxxxxx>]

On 07/24/2013 06:07 PM, Phill Whiteside wrote:

> I think that everyone missed what I mentioned. That is the 
> virtualisation of a 32 bit processor from a VM running on a 64 bit
> host. In kvm, i can choose from various pentium models, basic kvm32
> etc. etc. And whilst VM's can never take the place of actual
> hardware, when we need some i386 iso's testing to get them released,
> getting a VM to as near as to a 32 bit system to try them on is far
> better than just 'ticking' the box and saying it works.

So you want the virtualization environment to not show CPU flags
indicating the CPU is 64-bit capable?  Or to trap on all 64bit
instructions?  Both?

At first thought, all this really tests is that the compiler used for
the i386 code generation did not accidentally generate 64bit
instructions... is that what you are wanting to test?

Can't you run just file on all binaries installed and verify they are
32bit i386 binaries, and be done with it?  Something like:

  file /bin/* /usr/bin/* |grep executable |grep -v 'script\|32-bit'

would list any 64-bit executables in those directories, for example.

Can you point to a Launchpad bug report which this kind of "don't test
i386 on a 64bit capable CPU, you MUST test on an i386-only capable CPU"
testing approach would have found, which testing an i386 image on a
64-bit capable CPU (real or virtual) would have missed?  I need a real
example to better understand what you are expecting to gain.

We don't force i386 image testers to test on 32bit-only CPU hardware, do
we?  So why would we need to require testers using VMs to use
32-bit-only VMs?  We should be consistent about this, if indeed it is an
issue, as you seem to be suggesting it is.

Jonathan