[Nix-dev] Cross-compilation and intrusiveness

Ludovic Courtès ludo at gnu.org
Wed Nov 18 10:48:14 CET 2009


Saluton Lluís!

Llus Batlle <viriketo at gmail.com> writes:

> You can view the changes in this commit at:
>    https://svn.nixos.org/viewvc/nix?rev=18398&view=rev

[...]

> --- nixpkgs/branches/stdenv-updates/pkgs/development/libraries/readline/readline6.nix	2009-11-17 21:14:57 UTC (rev 18397)
> +++ nixpkgs/branches/stdenv-updates/pkgs/development/libraries/readline/readline6.nix	2009-11-17 22:58:48 UTC (rev 18398)
> @@ -8,7 +8,7 @@
>      sha256 = "1pn13j6f9376kwki69050x3zh62yb1w31l37rws5nwr5q02xk68i";
>    };
>  
> -  propagatedBuildInputs = [ncurses];
> +  propagatedHostInputs = [ncurses];

[...]

>    gcc43_real = lowPrio (wrapGCC (makeOverridable (import ../development/compilers/gcc-4.3) {
> -    inherit fetchurl stdenv texinfo gmp mpfr noSysDirs;
> +    inherit fetchurl gmp mpfr noSysDirs;
> +    stdenv = stdenvNoCross;
> +    texinfo = texinfoNoCross;
>      profiledCompiler = true;
>    }));

I haven’t followed your recent developments regarding cross-compilation
(sorry), but this looks to me like it’s getting very intrusive.

Do we really need ‘propagatedHostInputs’?  What does it mean?  Do we
really need ‘NoCross’ variants of packages?

(Besides, I think it’d have been nice to use a separate branch for this.
But then again, I can understand that SVN provides incentives against
branching...)

Thanks,
Ludo’.



More information about the nix-dev mailing list