GNU bug report logs - #62819
[PATCH core-updates 1/2] gnu: julia: Fix test suite regression.

Previous Next

Full log

View this message in rfc822 format

From: Simon Tournier <zimon.toutoune <at> gmail.com>
To: Andreas Enge <andreas <at> enge.fr>
Cc: 62819 <at> debbugs.gnu.org
Subject: [bug#62819] [PATCH core-updates 1/2] gnu: julia: Fix test suite regression.
Date: Fri, 14 Apr 2023 18:33:42 +0200

Hi Andreas,

On ven., 14 avril 2023 at 17:20, Andreas Enge <andreas <at> enge.fr> wrote:

> It passed on berlin, and I pushed the patch, thanks! This is frightening
> and quite unreasonable. If julia  needs more than 16GB of main memory to
> build, it will pass on our aarch64 machines with 32GB of memory, but not
> on most boards limited to 4GB.

Julia requires more than 4GB – currently the 8GB of my laptop are full
and extended by 6GB of swap; still building and maybe will fail — and
supports these platform:

                    x86-64         (64-bit)  
                    i686           (32-bit)          
                    ARMv8          (64-bit)          
                    ARMv7          (32-bit)          
                    PowerPC        (64-bit)  
    CUDA 10.1       Nvidia CUDA    (64-bit)
    ROCM            AMD ROCM       (64-bit)  
    oneAPI          Intel oneAPI   (64-bit)

    https://julialang.org/downloads/#supported_platforms

Upstream provides binaries for these.  AFAIK, for most of them, Julia is
cross-compiled and not natively compiled.  Well, I am still missing the
point to natively compile on limited boards.

Having substitutes running on non-x86_64 architectures is one thing,
directly build on these non-x86_64 architectures is another.

(I am not saying that direct builds on these non-x86_64 architectures is
not interesting by itself but that I miss the aim of such for some heavy
packages as Julia or GHC or else.)


Cheers,
simon

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.