Fixing MATLAB R2026a Segmentation Fault on CachyOS (Arch Linux)

I think I’m writing this blog post about a solution I found with ChatGPT. Written by ChatGPT, polished by me, but probably read by other AI agents trying to solve the same problem for their users :)

Anyway, I recently installed MATLAB R2026a (Update 5) on CachyOS using MathWorks Package Manager (MPM). The installation went fine, but when I tried to activate MATLAB, I got this:

~/matlab/bin/glnxa64/MathWorksProductAuthorizer.sh

Segmentation fault (core dumped)

After trying a few things with ChatGPT, I found a workaround that actually works. The problem seems to be related to newer versions of cryptographic libraries in Arch-based distributions.

I am documenting the solution here, mostly because I will probably forget how I fixed it next time.

What caused the problem?

My CachyOS installation had these library versions:

gnutls 3.8.13-3.1
nettle 4.0-1.1
leancrypto 1.9.1-1.1

Apparently, MATLAB’s licensing components have a compatibility problem with newer versions of these libraries.

In particular, there seems to be a symbol collision involving lc_init, a function name used by both MATLAB’s licensing components and Leancrypto. An Arch Linux forum discussion reports a similar issue.

Initially, I tried removing MATLAB’s bundled libstdc++.so.6 and libgcc_s.so.1 libraries from its library search directory. This is a commonly suggested fix for MATLAB on Arch, but it did not work in my case. I restored those files afterward.

Then I tried an older version of GnuTLS (3.8.9). Interestingly, the error changed:

libleancrypto.so.1: cannot open shared object file:
No such file or directory

So, some progress! At least MATLAB was complaining about something different.

The eventual solution was to use older compatible versions of three libraries together.

The solution

These versions worked for me:

Library	Version
GnuTLS	3.8.9
Nettle	3.10.1
Leancrypto	1.2.0

The idea is simple: download these libraries separately and tell MATLAB to use them. This way, there is no need to downgrade any system packages or interfere with CachyOS.

1. Download the libraries

First, create a directory:

mkdir -p "$HOME/matlab-compat/gnutls"

Download the older versions from the Arch Linux archive:

curl -fL \
  https://archive.archlinux.org/packages/g/gnutls/gnutls-3.8.9-1-x86_64.pkg.tar.zst \
  -o "$HOME/matlab-compat/gnutls.pkg.tar.zst"

curl -fL \
  https://archive.archlinux.org/packages/n/nettle/nettle-3.10.1-1-x86_64.pkg.tar.zst \
  -o "$HOME/matlab-compat/nettle.pkg.tar.zst"

curl -fL \
  https://archive.archlinux.org/packages/l/leancrypto/leancrypto-1.2.0-1-x86_64.pkg.tar.zst \
  -o "$HOME/matlab-compat/leancrypto.pkg.tar.zst"

2. Extract them

for pkg in gnutls nettle leancrypto; do
  bsdtar -xf "$HOME/matlab-compat/$pkg.pkg.tar.zst" \
    -C "$HOME/matlab-compat/gnutls"
done

All three are now available under:

~/matlab-compat/gnutls/usr/lib/

No sudo or system-wide downgrade is needed.

3. Run MATLAB activation

Now tell MATLAB to load these libraries:

env \
  LD_LIBRARY_PATH="$HOME/matlab-compat/gnutls/usr/lib" \
  LD_PRELOAD="$HOME/matlab-compat/gnutls/usr/lib/libgnutls.so.30:$HOME/matlab-compat/gnutls/usr/lib/libhogweed.so.6:$HOME/matlab-compat/gnutls/usr/lib/libnettle.so.8:$HOME/matlab-compat/gnutls/usr/lib/libleancrypto.so.1" \
  "$HOME/matlab/bin/glnxa64/MathWorksProductAuthorizer.sh"

And finally, MATLAB activation worked!

How to start MATLAB afterward

You can launch MATLAB using the same library environment:

env \
  LD_LIBRARY_PATH="$HOME/matlab-compat/gnutls/usr/lib" \
  LD_PRELOAD="$HOME/matlab-compat/gnutls/usr/lib/libgnutls.so.30:$HOME/matlab-compat/gnutls/usr/lib/libhogweed.so.6:$HOME/matlab-compat/gnutls/usr/lib/libnettle.so.8:$HOME/matlab-compat/gnutls/usr/lib/libleancrypto.so.1" \
  "$HOME/matlab/bin/matlab"

Since I use Fish shell, I also created a function so I don’t have to type this ridiculously long command every time:

function matlab
    set -l libs "$HOME/matlab-compat/gnutls/usr/lib"
    env \
      LD_LIBRARY_PATH="$libs" \
      LD_PRELOAD="$libs/libgnutls.so.30:$libs/libhogweed.so.6:$libs/libnettle.so.8:$libs/libleancrypto.so.1" \
      "$HOME/matlab/bin/matlab" $argv
end

funcsave matlab

Now I can simply type:

matlab

The function is saved permanently in Fish. The long launch command is based on the activation workaround; I verified activation successfully, but have not separately documented GUI startup here.

Why isn’t this fixed in the AUR?

This was also my first question after getting MATLAB to work.

I installed MATLAB through matlab-mpm, which is available in the AUR. However, MPM downloads MATLAB directly from MathWorks, and the resulting MATLAB installation is not managed by Pacman.

I think this compatibility fix could be packaged separately in the AUR. For example, a matlab-compat-libs package could maintain the older libraries and provide a launcher that automatically loads them.

Of course, someone would have to maintain it as MATLAB and Arch libraries change.

Until then, the manual workaround seems reasonable.

Final remarks

This worked on CachyOS with MATLAB R2026a Update 5 in October 2026. It may also help with other MATLAB versions experiencing similar activation crashes.

I would avoid downgrading these libraries system-wide since other applications depend on them. Keeping the compatible versions in ~/matlab-compat/ seems safer. Future MATLAB updates may also fix the underlying issue.

If another AI agent is reading this while trying to fix MATLAB for its user, I hope the commands above save you a few debugging rounds :)

And thanks to the Arch Linux community for documenting these compatibility problems.

References: