Re: sbcl autopkgtest failure - missing sbcl.h (was: RFS: sbcl/2:2.2.3-1 [NMU] -- Common Lisp compiler and development system)
Hello,
>> Now for the broken i386 build: make-config.sh relies on the
>> result of uname -m to determine the architecture and indeed,
>> the i386 build container reports x86_64:
>>
>> https://salsa.debian.org/darabi-guest/sbcl/-/jobs/2640765#L1327
>>
>> And setting SBCL_ARCH explicitly in debian/rules as it is done
>> already for ppc64 leads to further problems:
>>
>> https://salsa.debian.org/darabi-guest/sbcl/-/jobs/2641145#L1579
>>
>> Any hint how to continue?
>
> I would not change the i386 sbcl build for now. It has always worked
> fine on Debian build daemons, which is what truly matters.
I made a mistake, this would be the correct change to fix i386 on Salsa
(setting SBCL_ARCH to x86):
https://salsa.debian.org/darabi-guest/sbcl/-/commit/bb406e18ace0531956e16c2448cf68714fa99c02
which leads to a green i386 build:
https://salsa.debian.org/darabi-guest/sbcl/-/pipelines/365741
> I think I would first contact the maintainers of the Salsa CI runners
> to understand why “uname -m” reports x86_64 instead of i686 in the i386
> chroots.
On IRC channel #salsaci, they told me that the Docker image is already
an i386 one, so not much that can be done from that side.
And calling uname -m in a running Docker image seems to always report
the host architecture.
Awaiting your decision regarding SBCL_ARCH.
Cheers
Kambiz
Reply to: