[v1] acceptance: update asset hashes for the Bionic aarch64 installer

[PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Paolo Bonzini 5 years, 6 months ago

The kernel and initrd hashes seem to have changed for the Bionic
aarch64 installer, causing BootLinuxConsole.test_aarch64_xlnx_versal_virt
to fail.  Correct them based on the latest.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 tests/acceptance/boot_linux_console.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/acceptance/boot_linux_console.py b/tests/acceptance/boot_linux_console.py
index 73cc69c499..1fcd154a7c 100644
--- a/tests/acceptance/boot_linux_console.py
+++ b/tests/acceptance/boot_linux_console.py
@@ -338,13 +338,13 @@ class BootLinuxConsole(LinuxKernelTest):
         kernel_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
                       'bionic-updates/main/installer-arm64/current/images/'
                       'netboot/ubuntu-installer/arm64/linux')
-        kernel_hash = '5bfc54cf7ed8157d93f6e5b0241e727b6dc22c50'
+        kernel_hash = 'd6d17cc43d14f503493bd38e8277870f2b79132e'
         kernel_path = self.fetch_asset(kernel_url, asset_hash=kernel_hash)
 
         initrd_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
                       'bionic-updates/main/installer-arm64/current/images/'
                       'netboot/ubuntu-installer/arm64/initrd.gz')
-        initrd_hash = 'd385d3e88d53e2004c5d43cbe668b458a094f772'
+        initrd_hash = '7363cd9211b572add76a68df086be24545813b5f'
         initrd_path = self.fetch_asset(initrd_url, asset_hash=initrd_hash)
 
         self.vm.set_console()
-- 
2.26.2

Re: [PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Philippe Mathieu-Daudé 5 years, 6 months ago

On 8/10/20 12:37 AM, Paolo Bonzini wrote:
> The kernel and initrd hashes seem to have changed for the Bionic
> aarch64 installer, causing BootLinuxConsole.test_aarch64_xlnx_versal_virt
> to fail.  Correct them based on the latest.

Thanks for reporting the failure.

> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  tests/acceptance/boot_linux_console.py | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/acceptance/boot_linux_console.py b/tests/acceptance/boot_linux_console.py
> index 73cc69c499..1fcd154a7c 100644
> --- a/tests/acceptance/boot_linux_console.py
> +++ b/tests/acceptance/boot_linux_console.py
> @@ -338,13 +338,13 @@ class BootLinuxConsole(LinuxKernelTest):
>          kernel_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
>                        'bionic-updates/main/installer-arm64/current/images/'

The 'bug' is this link is using "current" which will be a problem each
time Ubuntu update the installer (latest update was last Wed, Aug 5).

Acceptance tests should be stick to a particular set of binary, knew to
have worked once in the past. If a new set works, I'd rather see a new
test added (because it might have a different QEMU coverage).

I'll send a fix of the binary path for these tests.

Regard,

Phil.

>                        'netboot/ubuntu-installer/arm64/linux')
> -        kernel_hash = '5bfc54cf7ed8157d93f6e5b0241e727b6dc22c50'
> +        kernel_hash = 'd6d17cc43d14f503493bd38e8277870f2b79132e'
>          kernel_path = self.fetch_asset(kernel_url, asset_hash=kernel_hash)
>  
>          initrd_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
>                        'bionic-updates/main/installer-arm64/current/images/'
>                        'netboot/ubuntu-installer/arm64/initrd.gz')
> -        initrd_hash = 'd385d3e88d53e2004c5d43cbe668b458a094f772'
> +        initrd_hash = '7363cd9211b572add76a68df086be24545813b5f'
>          initrd_path = self.fetch_asset(initrd_url, asset_hash=initrd_hash)
>  
>          self.vm.set_console()
>

Re: [PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Paolo Bonzini 5 years, 6 months ago

On 10/08/20 11:06, Philippe Mathieu-Daudé wrote:
> The 'bug' is this link is using "current" which will be a problem each
> time Ubuntu update the installer (latest update was last Wed, Aug 5).
> 
> Acceptance tests should be stick to a particular set of binary, knew to
> have worked once in the past. If a new set works, I'd rather see a new
> test added (because it might have a different QEMU coverage).

That seems overkill but yeah the tests should have used
20101020ubuntu543.15 instead of current.  I think this should get into rc4.

Paolo

Re: [PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Philippe Mathieu-Daudé 5 years, 6 months ago

On 8/10/20 11:24 AM, Paolo Bonzini wrote:
> On 10/08/20 11:06, Philippe Mathieu-Daudé wrote:
>> The 'bug' is this link is using "current" which will be a problem each
>> time Ubuntu update the installer (latest update was last Wed, Aug 5).
>>
>> Acceptance tests should be stick to a particular set of binary, knew to
>> have worked once in the past. If a new set works, I'd rather see a new
>> test added (because it might have a different QEMU coverage).
> 
> That seems overkill but yeah the tests should have used
> 20101020ubuntu543.15 instead of current.  I think this should get into rc4.

As our CI is not yet gating, I don't think this justifies delaying the
release (except if there is a critical bugfix on the list).

Maybe it is acceptable to simply merge the CI test fixes before the
release tag, as no code/binary is affected...

> 
> Paolo
>

Re: [PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Paolo Bonzini 5 years, 6 months ago

On 10/08/20 11:35, Philippe Mathieu-Daudé wrote:
>> That seems overkill but yeah the tests should have used
>> 20101020ubuntu543.15 instead of current.  I think this should get into rc4.
> As our CI is not yet gating, I don't think this justifies delaying the
> release (except if there is a critical bugfix on the list).

It is not gating, but it means that acceptance test for 5.1.0 will be
broken forever for everyone that attempts to run them.

Paolo

> Maybe it is acceptable to simply merge the CI test fixes before the
> release tag, as no code/binary is affected...
>

Re: [PATCH] acceptance: update asset hashes for the Bionic aarch64 installer

Posted by Philippe Mathieu-Daudé 5 years, 6 months ago

On 8/10/20 12:37 AM, Paolo Bonzini wrote:
> The kernel and initrd hashes seem to have changed for the Bionic
> aarch64 installer, causing BootLinuxConsole.test_aarch64_xlnx_versal_virt
> to fail.  Correct them based on the latest.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  tests/acceptance/boot_linux_console.py | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/acceptance/boot_linux_console.py b/tests/acceptance/boot_linux_console.py
> index 73cc69c499..1fcd154a7c 100644
> --- a/tests/acceptance/boot_linux_console.py
> +++ b/tests/acceptance/boot_linux_console.py
> @@ -338,13 +338,13 @@ class BootLinuxConsole(LinuxKernelTest):
>          kernel_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
>                        'bionic-updates/main/installer-arm64/current/images/'
>                        'netboot/ubuntu-installer/arm64/linux')
> -        kernel_hash = '5bfc54cf7ed8157d93f6e5b0241e727b6dc22c50'
> +        kernel_hash = 'd6d17cc43d14f503493bd38e8277870f2b79132e'
>          kernel_path = self.fetch_asset(kernel_url, asset_hash=kernel_hash)
>  
>          initrd_url = ('http://ports.ubuntu.com/ubuntu-ports/dists/'
>                        'bionic-updates/main/installer-arm64/current/images/'
>                        'netboot/ubuntu-installer/arm64/initrd.gz')
> -        initrd_hash = 'd385d3e88d53e2004c5d43cbe668b458a094f772'
> +        initrd_hash = '7363cd9211b572add76a68df086be24545813b5f'
>          initrd_path = self.fetch_asset(initrd_url, asset_hash=initrd_hash)
>  
>          self.vm.set_console()
> 

I haven't noticed that because I have these files in the Avocado cache.

The only way to detect such problems is blowing the cache... I certainly
don't want to do that daily. OTOH a developer starting to use the CI or
Avocado locally will hit the problem.

I was thinking about running a job with empty cache for each tags.
Maybe it is safer to do that on a weekly basis a least?

Regards,

Phil.