Re: building fewer packages or reduced functionality on i386

On Sat, Aug 15, 2026 at 10:43:39AM +0100, Simon McVittie wrote:
>....
>     Build-Depends:
>       libbasic-dev,
>       libgui-dev <!partial-arch>,
>
>     ...
>
>     Package: foo-gui
>     Build-Profiles: <!partial-arch>
>...
> For example we could classify architectures into those where a GUI might be
> practically useful (everything except s390x) and those where it isn't
> (s390x), or those that could be useful for "heavy" CPU-bound computation
> like scientific software (amd64, arm64) and those that most likely aren't
> (riscv64, loong64).

Loongson is selling loong64 server CPUs with 64 cores (128 threads).

riscv64 is not yet used in HPC, but work on that is ongoing.[1]

Both architectures are pushed in China for technological independence.

> As with any feature reduction, turning off packages per-architecture without
> breaking dependencies needs to start from leaf packages, and work inwards
> towards core packages.

When people are discussing adding work, it is always useful to get
a step back and ask how big the actual benefits would be - especially 
compared to the amount of work suggested.

The major porting pain of i386 has been the excess precision of the x87,
which goes away with a higher baseline.

The major porting pain with s390x is that it is the only remaining 
non-ports big endian architecture, not that GUI parts are built.

Building packages differently on different architectures has always been 
a source of breakages, trying to do the same on all architectures when
reasonably possible is usually the option that requires the least work.

>     smcv

cu
Adrian

[1] https://www.researchgate.net/publication/405221610_Monte_Cimone_v3_Where_RISC-V_Stands_in_High-Performance_Computing