On 2026-01-23 14:11, Jochen Sprickerhof wrote:
> Hi Aquila,
> 
Hello Jochen,

Thanks a lot for the explanation.
> * [email protected] <[email protected]> [2026-01-23 06:09]:
>>That said, one small drawback is that this makes the sbuild invocation
>>grow with the number of injected binaries, since each extra .deb turns
>>into an additional --add-depends entry. It is workable, but can get a
>>bit unwieldy for sources producing lots of binaries.
> 
> Agreed. You could create a dependency package to pass all the relations, 
> similar to debootsnap, in devscripts, does it.
> 
>>If it makes sense from sbuild's perspective, having an opt-in switch to
>>treat --extra-package versions as preferred during build-dep resolution
>>would make this workflow a bit cleaner, while keeping the current
>>default behavior unchanged.
> 
> The problem here is that you are using the aspcud solver which, from my 
> tests, does not use the apt pinning mechanism. Is there any reason for it?

For ratt/Salsa CI rdep jobs I chose --build-dep-resolver=aspcud
specifically to mimic buildd behaviour as closely as possible. It looks
like buildds also use aspcud (at least for the experimental), for
instance as configured in the buildd puppet sbuild config here:
https://salsa.debian.org/dsa-team/mirror/dsa-puppet/-/blob/production/modules/buildd/templates/sbuild.conf.erb#L7

> 
> Switching to the default, apt, will fix your problem. I am also working on 
> dropping the aspcud (experimental) and aptitude (backports) solver usage in 
> Debian. The apt developers proposed to replace it by apt --no-strict-pinning 
> but I still need to add a proper interface to sbuild for that.

Given you're working on dropping aspcud/aptitude usage in Debian in
favour of apt (with the proposed relaxed pinning behaviour), does it
make sense for us to remove aspcud now from the Salsa CI rdep jobs and
switch to the default apt resolver already? Or would you recommend
keeping aspcud until buildds have switched, to stay aligned with current
buildd semantics?

> 
> Cheers Jochen


Cheers,

Aquila Macedo

Reply via email to