Re: CVS commit: pkgsrc/math/suitesparse

["Dr. Thomas Orgis" <thomas.orgis%uni-hamburg.de@localhost>]

Am Tue, 17 Jun 2025 12:28:30 +0100
schrieb Jonathan Perkin <jperkin%pkgsrc.org@localhost>:

> Setting it to an unsupported value has the same effect as not setting it 
> at all.  pkglint would have warned you prior to commit:
> 
>    WARN: Makefile:15: "fortran8" is not valid for USE_LANGUAGES. 

(maybe there _should_ be a fortran2008 value, like fortran90,
fortran95, fortran2003 … ?)

> So you are effectively removing "fortran" from USE_LANGUAGES.

Of course pkglint is a help, but I do wonder if the USE_LANGUAGES
system could be made more robust to fail on such obvious errors.

I reverted the change again to plain fortran (which can mean anything
now from 90 to 2023). So that leaves me with the question how Kevin ended up with

> The issue that was happening was it was unable to link against its
> own libraries, namely, libAMD. I did _not_ try using -rpath on it
> yet.

> (hard to tell as the package fails on illumos and macOS 
> daily bulk builds for different reasons)

Hm, should I look at those failures and guess or is the reason/fix
clear to you? I can only test things on Linux/x86_64.


Alrighty then,

Thomas

-- 
Dr. Thomas Orgis
HPC @ Universität Hamburg