On 19/09/2023 12:21, Jan Beulich wrote:
>
>
> On 19.09.2023 11:53, GitLab wrote:
>>
>>
>> Pipeline #1009404353 has failed!
>>
>> Project: xen ( https://gitlab.com/xen-project/xen )
>> Branch: staging ( https://gitlab.com/xen-project/xen/-/commits/staging )
>>
>> Commit: ea36ac0d (
>> https://gitlab.com/xen-project/xen/-/commit/ea36ac0de27c2a7c847a2a52c3e0f97a45864d81
>> )
>> Commit Message: xen/ppc: Enable full Xen build
>>
>> Bring ppc's Mak...
>> Commit Author: Shawn Anastasio
>> Committed by: Jan Beulich ( https://gitlab.com/jbeulich )
>>
>>
>> Pipeline #1009404353 (
>> https://gitlab.com/xen-project/xen/-/pipelines/1009404353 ) triggered by
>> Ganis ( https://gitlab.com/ganis )
>> had 5 failed jobs.
>>
>> Job #5118269375 ( https://gitlab.com/xen-project/xen/-/jobs/5118269375/raw )
>>
>> Stage: build
>> Name: debian-bullseye-gcc-ppc64le-debug-randconfig
>
> This and ...
>
>> Job #5118269256 ( https://gitlab.com/xen-project/xen/-/jobs/5118269256/raw )
>>
>> Stage: analyze
>> Name: eclair-x86_64
>> Job #5118269373 ( https://gitlab.com/xen-project/xen/-/jobs/5118269373/raw )
>>
>> Stage: build
>> Name: debian-bullseye-gcc-ppc64le-randconfig
>
> ... this imo can't be expected to work. Is it really useful to run randconfig
> tests on ports which are only in the process of being brought up?
>
>> Job #5118269370 ( https://gitlab.com/xen-project/xen/-/jobs/5118269370/raw )
>>
>> Stage: build
>> Name: debian-bullseye-gcc-ppc64le-debug
>> Job #5118269369 ( https://gitlab.com/xen-project/xen/-/jobs/5118269369/raw )
>>
>> Stage: build
>> Name: debian-bullseye-gcc-ppc64le
>
> These two, otoh, look to be a result of the tests pre-seeding xen/.config with
> CONFIG_DEBUG settings, followed by making the olddefconfig goal. That, aiui,
> isn't picking up xen/arch/*/configs/*_defconfig, which at this point is
> mandatory for PPC (and likely is going to be so also for RISC-V once the full
> build is enabled there), at least as far as some of the option disables there
> go.
>
> I think this wants switching to making the defconfig goal, and substituting
> CONFIG_DEBUG in the resulting .config. Due to x86'es and Arm's defconfig-s
> all being empty, this ought to be no change in what exactly is being tested
> there.
Apart from CONFIG_DEBUG there are other options we add using EXTRA_XEN_CONFIG
that might result in new options becoming visible and thus triggering a prompt
without olddefconfig.
So if at all, I think the flow should be:
defconfig
replacements
olddefconfig
make
~Michal
