Edmund Lodewijks <[email protected]> writes:

> Zig's own from-C path is bootstrap.c, and it needs stage1/zig1.wasm. That 
> file is a precompiled binary, so it cannot stay in the source package. 
> Removing it is the whole point of the work others and I did in an attempt to 
> get Zig into Debian.
>
> So what should a profile build start from?
>
>   (a) The seed I built outside the archive, shipped as an orig-stage0 
> tarball? This works today, but it is a binary. The profile would then test 
> that the plumbing works, not that the starting point is source. My work did 
> show, however, that the seed is identical to upstream's, so there's that.
>
>   (b) What Ivan Carvalho suggests later in this thread (the work he has been 
> doing for a separate bootstrap package in Debian): regenerate the wasm seed 
> on the amd64 builders, then use it to build the other architectures. That 
> keeps everything inside Debian, but it is no longer a single build.
>
> Or, perhaps there is another option that I am not thinking of, that you think 
> would work best?

I believe (a) would be unacceptable but (b) sounds good -- and if
building the *.wasm seed on x86 is feasible, delivering it in an
architecture:all binary package seems fine, and even better than a
manually testable <bootstrap> build profile.

/Simon

Attachment: signature.asc
Description: PGP signature

Reply via email to