Re: [PATCH v2 7/7] tests/avocado: add boot_xen tests

[Cleber Rosa <crosa@xxxxxxxxxx>]

On Thu, Feb 18, 2021 at 10:43:54AM +0100, Philippe Mathieu-Daudé wrote:
> On 2/17/21 9:46 PM, Cleber Rosa wrote:
> > On Thu, Feb 11, 2021 at 05:19:45PM +0000, Alex Bennée wrote:
> >> These tests make sure we can boot the Xen hypervisor with a Dom0
> >> kernel using the guest-loader. We currently have to use a kernel I
> >> built myself because there are issues using the Debian kernel images.
> >>
> >> Signed-off-by: Alex Bennée <alex.bennee@xxxxxxxxxx>
> >> ---
> >>  MAINTAINERS                  |   1 +
> >>  tests/acceptance/boot_xen.py | 117 +++++++++++++++++++++++++++++++++++
> >>  2 files changed, 118 insertions(+)
> >>  create mode 100644 tests/acceptance/boot_xen.py
> 
> >> +class BootXen(BootXenBase):
> >> +
> >> +    @skipIf(os.getenv('GITLAB_CI'), 'Running on GitLab')
> >> +    def test_arm64_xen_411_and_dom0(self):
> >> +        """
> >> +        :avocado: tags=arch:aarch64
> >> +        :avocado: tags=accel:tcg
> >> +        :avocado: tags=cpu:cortex-a57
> >> +        :avocado: tags=machine:virt
> >> +        """
> >> +        xen_url = ('https://deb.debian.org/debian/'
> >> +                   'pool/main/x/xen/'
> >> +                   
> >> 'xen-hypervisor-4.11-arm64_4.11.4+37-g3263f257ca-1_arm64.deb')
> >> +        xen_sha1 = '034e634d4416adbad1212d59b62bccdcda63e62a'
> > 
> > This URL is already giving 404s because of a new pacakge.  I found
> > this to work (but yeah, won't probably last long):
> > 
> >         xen_url = ('http://deb.debian.org/debian/'
> >                    'pool/main/x/xen/'
> >                    
> > 'xen-hypervisor-4.11-arm64_4.11.4+57-g41a822c392-2_arm64.deb')
> >         xen_sha1 = 'b5a6810fc67fd50fa36afdfdfe88ce3153dd3a55'
> 
> This is not the same package version... Please understand the developer
> has to download the Debian package sources, check again the set of
> downstream changes between 37 and 57. Each distrib number might contain
> multiple downstream patches. Then the testing has to be done again,
> often enabling tracing or doing single-stepping in gdb. This has a
> cost in productivity. This is why I insist I prefer to use archived
> well tested artifacts, rather than changing package URL randomly.
>

I understand it's not the same version... but from my different and
limited PoV it was the obvious thing to suggest during a review.  Of
course using stable archived versions is much better (I believe Alex
will look into that for these packages).

Best,
- Cleber.

Attachment: signature.asc
Description: PGP signature