Add path workaround for Linux

[egavazzi]

Description

Exactly the same as #226, but for Linux. Tested and found out that it works too.

Related to:

• segfault with Matlab 2022a on Windows #200
• MATLAB code does not run / fails silently #201
• MATLAB.jl crashing when used with Julia 1.10 #220

Before

julia> mat"version"

[1135198] signal 11 (1): Segmentation fault
in expression starting at REPL[3]:1
Allocations: 8630736 (Pool: 8629112; Big: 1624); GC: 9
Segmentation fault (core dumped)

After

julia> mat"version"
"24.2.0.2806996 (R2024b) Update 3"

[tqml]

I don't have a linux machine with Matlab to test it, but isn't Linux using a : instead of ; to separate elements in PATH?

[egavazzi]

Uh-oh, you're completely right. It was working in my minimal example to start Matlab as matlab_libpath was added at the beginning of PATH. But due to the wrong separator, the rest was not interpreted as part of PATH anymore. It should be fixed now.

[tqml]

Nice, thanks for the patch! :)

[egavazzi]

Thank you for the quick review, and for catching my mistake! :)

[musm]

Please use rebase and merge next time.

[sdamera95]

Hello,

Thank you for the quick review, and for catching my mistake! :)

Sorry, the issue still seems to persist on Linux (Ubuntu 22.04) and Matlab 24.2.0.2712019 (R2024b) after the path workaround.

Tested with clean Julia environments across 1.9, 1.10_lts and 1.11 versions.

Also checked the pre--workaround tagged release version of MATLAB.jl(0.8.4) and by building the latest version (0.9.0) of the package locally. The issue seems to persist.

Am I missing something? Please see the error from one of the configs I tested below (all other combinations seem to be identical):

julia> versioninfo()
Julia Version 1.11.3
Commit d63adeda50d (2025-01-21 19:42 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 32 × AMD Ryzen 9 7950X 16-Core Processor
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver4)
Threads: 1 default, 0 interactive, 1 GC (on 32 virtual cores)
Environment:
  LD_LIBRARY_PATH = /usr/lib/x86_64-linux-gnu/gazebo-11/plugins:/opt/ros/humble/opt/rviz_ogre_vendor/lib:/opt/ros/humble/lib/x86_64-linux-gnu:/opt/ros/humble/opt/libphidget22/lib:/opt/ros/humble/lib

(test) pkg> st
Status `~/test/Project.toml`
  [10e44e05] MATLAB v0.8.4

julia> using MATLAB

julia> mat"version"

[938007] signal 11 (1): Segmentation fault
in expression starting at REPL[5]:1
Allocations: 12741729 (Pool: 12740021; Big: 1708); GC: 23
Segmentation fault (core dumped)

Could you please let me know if I am doing something wrong? I'd be happy to provide more info/ test the changes.
Thank you.

[tqml]

Hey @sdamera95
thanks for the message. Can you show the output of MATLAB.matlab_libpath ?
Also could you try removing the LD_LIBRARY_PATH env variable?

[egavazzi]

I think I managed to reproduce the issue and fix it.

On a fresh install, I created two environments. In the first one I installed MATLAB.jl v0.8.4 and in the second one MATLAB.jl v0.9.0 (master branch from github). I installed v0.8.4 before v0.9.0. I got a segfault with both.

I fixed it by rebuilding the package (in the environment with MATLAB.jl v0.9.0) with

pkg> build MATLAB

Then restarted Julia, and it worked.

[egavazzi]

To be honest my first attempt at debugging this was: Installed MATLAB.jl v0.8.4, got a segfault. Removed MATLAB.jl v0.8.4 and installed v0.9.0 instead, got a segfault. Nuked .julia/compiled/v.11/MATLAB/ and .julia/packages/MATLAB/, retried with v0.9.0 and it worked.

Then I remembered that to change matlab version one has to trigger a rebuild with pkg> build MATLAB. So I did what is described in my previous comment, which is more elegant than manually removing things under compiled/ and packages/. However if rebuilding does not work you might want to try the manual removing solution.

[egavazzi]

Fun (and somewhat interesting) fact: after rebuilding MATLAB.jl v0.9.0, the version v0.8.4 is also working. But even better, rebuilding the version 0.8.4 does not bring back the segfault. It just continues to work??

I am not competent enough to understand what is happening here. And the fact that two installations of MATLAB.jl in two different environments affect each other in the first place is also something that confuses me.

[tqml]

That is very odd.
I know the package has a build.jl that searches for MATLAB (either via an ENV variable or in the default install locations), picks the newest version and writes the file system path to a deps.jl file. This file is then imported in the package.
Maybe the different packages reference the same deps.jl file that links to a working MATLAB installation? You can check MATLAB.matlab_libpath, this constant is defined in deps.jl

[egavazzi]

In both environments (with the different versions of MATLAB.jl), I get

julia> MATLAB.matlab_libpath
"/usr/local/MATLAB/R2024b/bin/glnxa64"

The content of deps.jl in .julia/packages/MATLAB/XXX/deps/ is also the same for both:

# This file is automatically generated, do not edit.

function check_deps()
    if libmx_size != filesize(Libdl.dlpath(joinpath(matlab_libpath, "libmx")))
        error("MATLAB library has changed, re-run Pkg.build(\"MATLAB\")")
    end
end

const matlab_libpath = "/usr/local/MATLAB/R2024b/bin/glnxa64"
const matlab_cmd = "exec /usr/local/MATLAB/R2024b/bin/matlab"
const libmx_size = 3600568

I removed everything in .julia/packages/MATLAB/, reinstalled the v0.8.4 to get again the "non-working" version with the segfault. I checked again both MATLAB.matlab_libpath and the content of deps.jl but they are the same as above.