Re: [PATCH v3 00/23] rust: fix CI + allow older versions of rustc and bindgen

["Alex Bennée" <alex.bennee@linaro.org>]

Daniel P. Berrangé <berrange@redhat.com> writes:

> On Sun, Oct 27, 2024 at 10:01:26AM +0300, Michael Tokarev wrote:
>> I think this is the wrong direction (ie, backwards).
>> 
>> Sacrificing current code to be compatible with old stuff feels wrong.
>> Especially for really old, like rustc in debian bookworm.
>> 
>> bookworm has rustc-web (and a few related packages) which is regular
>> rustc version 1.78, just renamed.  It is regular bookworm, not backports.
>> It has some packages disabled (compared to regular rust) and is a hack,
>> but it exists and can be used for now (dunno if it is sufficient for
>> qemu though).
>> 
>> Also debian has backports mechanism, which also can be used for qemu -
>> I can try back-porting regular rust (and llvm) to bookworm.
>> 
>> I think this is a better way (at least a way forward) than trying to
>> move backwards.
>> 
>> But generally, what is the reason to support debian stable?  I understand
>> the CI thing, - we need a way to test stuff.  For this, I'd say a better
>> alternative would be to target debian testing (currently trixie), not
>> debian stable.
>
> The stable distros are what our community of contributors are usually
> using, as few people want non-released bleeding edge distros as their
> primary development platform.
>
> Custom installing latest upstream pieces is not a user friendly position
> to take. Occassionally it is unavoidable, but it is something to be
> avoided wherever practical.

At least rustup makes this reasonably easy for the rust bits. We do rely
on the excellent Debian backports for getting QEMU quickly into testing
images but I was assuming we would have trixie before --enable-rust
became mandatory so I'm not too worried if bookworm is the outlier for
old versions.

>
> With regards,
> Daniel

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro