Bug#973287: systemd autopkgtest-virt-lxc failure on arm64

Ryutaroh Matsumoto ryutaroh at ict.e.titech.ac.jp
Fri Nov 27 08:40:49 GMT 2020


Control: retitle -1 systemd autopkgtest-virt-qemu failure on arm64

>> By the way, systemd autopkgtest-virt-qemu works fine if packages with
>> "Priority: important" are included in VM.
>> But packages "Priority: required" is only included, the systemd test
>> script fails.
>> Is it considered as a bug??
> 
> TBH, no idea. This would probably need input from the autopkgtest/debci
> maintainers what kind of environment can be expected by autopkgtests
> and which dependencies need to be listed explicitly.
> Which dependencies are missing, btw?

I have little idea which package is missing/required in this context.
What I see is that, when only "Priority: required" packages are installed
to autopkg VM, at some point systemd test scripts start failing to look up hostnames,
and http://deb.debian.org becomes unable to be found.
/etc/resolv.conf seems destroyed...
This symptom does not appear when "Priority: important" packages are installed.
This is also observed on amd64 QEMU testbed, and not specific to arm64.

If this is considered to be a bug and deserves further investigation, I will
look at it. In such a case maybe cloning this and retitling the cloned one are
suitable.

> module is missing (I think returning 77 should be sufficient)
> Would you like to work on such a patch?

I have little familiarlity with writing autopkgtest scripts,
so I hesitate to take its responsibility...

By the way, to me it seems that when all scripts succeed on autopkgtest-virt-qemu,
there is no/little problem, i.e., we can ignore autopkgtest-virt-lxc errors
provided that there is no error on autopkgtest-virt-qemu.
So I changed the title of this report.

Best regards, Ryutaroh

From: Michael Biebl <biebl at debian.org>
Subject: Re: Bug#973287: systemd autopkgtest-virt-lxc failure on arm64
Date: Fri, 27 Nov 2020 09:23:31 +0100

> Am Freitag, den 27.11.2020, 12:01 +0900 schrieb Ryutaroh Matsumoto:
>> > > > We probably need to cherry-pick two changes
>> > I released 246.6-5 with those changes the other day.
>> 
>> I run autopkgtest 246.6-5 on amd64 and arm64 VM.
>> "upstream" sometimes succeeds and sometimes fails, so I assume it
>> is OK
> 
> Ok, that's at least progress. Thanks for testing.
> 
>> "storage" consistently fails on arm64 VM. The reason is simple as
>> below:
>> modprobe: FATAL: Module scsi_debug not found in directory
>> /lib/modules/5.9.0-3-arm64
>> 
>> linux-image-arm64 does not have scsi_debug kernel module.
>> I have no idea which package is responsible for autopkgtest error of
>> "storage"
> 
> Hm, I wasn't aware that scsi_debug is not built on all architectures.
> I guess it would make sense if the storaged test checks for the
> availability of the scsi_debug module and simply skips the test if the
> module is missing (I think returning 77 should be sufficient)
> Would you like to work on such a patch?
> 
>> By the way, systemd autopkgtest-virt-qemu works fine if packages with
>> "Priority: important" are included in VM.
>> But packages "Priority: required" is only included, the systemd test
>> script fails.
>> Is it considered as a bug??
> 
> TBH, no idea. This would probably need input from the autopkgtest/debci
> maintainers what kind of environment can be expected by autopkgtests
> and which dependencies need to be listed explicitly.
> Which dependencies are missing, btw?
> 
> 
> Regards,
> Michael



More information about the Pkg-systemd-maintainers mailing list