Gioele Barabucci <[email protected]> writes:

> On 11/11/26 13:22, Edmund Lodewijks wrote:
>> On 2026/11/11 13:18, Jonathan Dowland wrote:
>>>> Let me ask this question though: what is the purpose of packaging Zig
>>>> into Debian? In my opinion, upstream developers prefer to obtain Zig
>>>> directly from upstream. I think the purpose of packaging Zig into
>>>> Debian is not for Zig users but so that Zig can be used as a
>>>> development package to build other Debian packages. In this case, you
>>>> don't need any cross-compilation support or any of these vendored
>>>> dependencies. They can be completely omitted from the package. This
>>>> hamstrings Zig as a compiler tool, so it would be appreciated if it
>>>> were made clear to Zig users that this package is not for them. But I
>>>> think that is a path worth exploring for Debian, if the goal is to
>>>> build other packages.
>> For the first upload, however, I think a stripped down version would
>> be a great start. And then we could always add the cross-compiling
>> parts later again.
>
> Or have two packages: zig-{core,minimal,basic} and zig.
>
> (At the beginning these could be two separate source packages, then
> they could be merged into a single source package from which two
> binary packages can be built.)
>
> In this way Debian packages can Build-Depend on the smaller package
> and users can install the full-blown package (when it will become
> available).

Please consider a <bootstrap> build profile instead, so that we will be
able to re-bootstrap zig for new platforms in the future.  Then we can
continously test that zig still is possible to bootstrap using current
packages, into something that builds the real zig.  Maybe upstream
doesn't handle this today, but modelling like this could result in
upstream prioritizing this ability.  Which is important for new
architectures in the future.

/Simon

Attachment: signature.asc
Description: PGP signature

Reply via email to