Open mharradon opened 8 months ago
A few thoughts:
Thanks for the quick reply! To be clear regarding (1), my primary issue here is with running the conda-forge distribution of Julia. I don't install Julia from main in either example (the latter error may be irrelevant due to channel mixing). My primary concern is the warning of improper library linking when running this conda-forge Julia distribution in the most basic installation.
Again, the problem is that the suitesparse-feedstock is out of date. There's nothing to be done to fix this in this repository.
Thanks. I do need to get around to changing the build to adapt for this. Keeping pinging me if this is not fixed.
Hi @minrk It would be great if it's fixed. Thanks!
I just encountered https://github.com/JuliaSparse/SparseArrays.jl/issues/542 - is this the same issue?
Can you show conda env export
? The latest build is not done with cholmod 4, so I think you aren't getting the build where this is fixed
Here (clean new environment and then conda install julia
):
$ conda env export
name: jupyter
channels:
- conda-forge
- intel
- defaults
dependencies:
- _libgcc_mutex=0.1=conda_forge
- _openmp_mutex=4.5=2_gnu
- arpack=3.9.1=nompi_h77f6705_101
- bzip2=1.0.8=hd590300_5
- c-ares=1.28.1=hd590300_0
- ca-certificates=2024.6.2=hbcca054_0
- curl=8.8.0=he654da7_0
- git=2.45.2=pl5321ha099dd3_1
- gmp=6.3.0=h59595ed_1
- icu=73.2=h59595ed_0
- julia=1.10.4=hf18f99d_1
- keyutils=1.6.1=h166bdaf_0
- krb5=1.21.2=h659d440_0
- libblas=3.9.0=22_linux64_openblas
- libcblas=3.9.0=22_linux64_openblas
- libcurl=8.8.0=hca28451_0
- libedit=3.1.20191231=he28a2e2_2
- libev=4.33=hd590300_2
- libexpat=2.6.2=h59595ed_0
- libgcc-ng=13.2.0=h77fa898_10
- libgfortran-ng=13.2.0=h69a702a_10
- libgfortran5=13.2.0=h3d2ce59_10
- libgit2=1.8.1=he8d1d4c_1
- libgomp=13.2.0=h77fa898_10
- libhwloc=2.10.0=default_h5622ce7_1001
- libiconv=1.17=hd590300_2
- liblapack=3.9.0=22_linux64_openblas
- libnghttp2=1.58.0=h47da74e_1
- libopenblas=0.3.27=pthreads_h413a1c8_0
- libopenblas-ilp64=0.3.27=pthreads_h384dd9e_0
- libopenlibm4=0.8.1=hd590300_1
- libssh2=1.11.0=h0841786_0
- libstdcxx-ng=13.2.0=hc0a3c3a_10
- libunwind=1.6.2=h9c3ff4c_0
- libutf8proc=2.8.0=h166bdaf_0
- libxcrypt=4.4.36=hd590300_1
- libxml2=2.12.7=hc051c1a_1
- libzlib=1.3.1=h4ab18f5_1
- mbedtls=3.5.1=h59595ed_0
- metis=5.1.0=h59595ed_1007
- mpfr=4.2.1=h9458935_1
- ncurses=6.5=h59595ed_0
- openblas-ilp64=0.3.27=pthreads_h3d04fff_0
- openlibm=0.8.1=hd590300_1
- openssl=3.3.1=h4ab18f5_0
- p7zip=16.02=h9c3ff4c_1001
- pcre2=10.44=h0f59acf_0
- perl=5.32.1=7_hd590300_perl5
- suitesparse=7.7.0=hf4753ba_1
- tbb=2021.12.0=h297d8ca_1
- xz=5.2.6=h166bdaf_0
- zlib=1.3.1=h4ab18f5_1
- zstd=1.5.6=ha6fb4c9_0
prefix: /home/obk/anaconda3/envs/jupyter
Yes, I think I am starting to understand. That message doesn't quite mean what it says - SparseArrays expects a particular version of suitesparse that it would usually be compiled against, but doesn't check what it was actually compiled against. It mistakenly claims that this is the version julia is compiled against, which is not true, but it does appropriately warn that the version is significantly different from what it expects. Since it's different by a major version, it's reasonable to warn that things might not work.
Our patch is enough to get the current suitesparse to be loaded and everything seems to work, but this hardcoded version check is still there.
Solution to issue cannot be found in the documentation.
Issue
Installing Julia just from conda-forge onto an Intel Mac results in CHOLMOD warning. I am not sure if what it states is true, as apparently this warning has been erroneous in the past:
When not forcing conda-forge (which I expect to have issues anyway) I get:
Installed packages
Environment info
Thanks for taking a look!