Re: [PATCH] Fix gfortran (?)

[Ludovic Courtès]

Ricardo Wurmus <address@hidden> skribis:

> with this patch applied I can build randomjungle without having to add
> the “lib” output of gfortran to the package inputs.

‘randomjungle’ currently fails at configure time like this:

--8<---------------cut here---------------start------------->8---
checking for pow... no
checking for pow in -lm... no
configure: WARNING: cannot find library containing definition of pow
--8<---------------cut here---------------end--------------->8---

Is this due to the fact that libgcc_s.so is not found?  I vaguely recall
discussions like this.

> From f3d97b7f1f2b597f7f74b2f59212f241a87fa602 Mon Sep 17 00:00:00 2001
> From: Ricardo Wurmus <address@hidden>
> Date: Mon, 21 Mar 2016 14:09:45 +0100
> Subject: [PATCH] gnu: gfortran: Disable separate "lib" output.
>
> * gnu/packages/gcc.scm (gfortran): Disable separate "lib" output.
> ---
>  gnu/packages/gcc.scm | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/gnu/packages/gcc.scm b/gnu/packages/gcc.scm
> index 1df5150..341c30c 100644
> --- a/gnu/packages/gcc.scm
> +++ b/gnu/packages/gcc.scm
> @@ -474,7 +474,8 @@ as the 'native-search-paths' field."
>  
>  (define-public gfortran
>    (custom-gcc gcc "gfortran" '("fortran")
> -              %generic-search-paths))
> +              %generic-search-paths
> +              #:separate-lib-output? #f))

This is easier than augmenting the “lib” spec.

The downside is that things compiled with gfortran would retain a
reference to the whole ‘gfortran’ package (the closure of which is
217 MiB, vs. 72 MiB for gfortran:lib.)

Ludo’.