The segfault issue with MATLAB versions 2022/2023 is gone on Linux platforms

[egavazzi]

I noticed today that MATLAB.jl works perfectly with the new version of MATLAB R2023b.

By curiosity I also checked the former 'problematic' versions (R2023a, R2022b, R2022a) and they all work on my computer. I suspect some update of internals on the MATLAB side: all three versions have received updates in July/August of this year (as can be seen here).

Here is what I manage to get in the REPL:

julia> mat"version"
"9.12.0.2327980 (R2022a) Update 7"

julia> mat"version"
"9.13.0.2320565 (R2022b) Update 6"

julia> mat"version"
"9.14.0.2337262 (R2023a) Update 5"

julia> mat"version"
"23.2.0.2365128 (R2023b)"

So now I don't know what is preferred: removing completely the warning message in README.md, or updating it to inform everything is working as normal now and keeping it for some time 🙂

p.s: this should close issues #200 and #201 🎉

[MariusDrulea]

If the Matlab updates solved the issue, we shall tell the user to update the Matlab in case of failure. Perhaps the update has an ID or something the user can check.

[egavazzi]

I tried to identify the specific update version from which things started to work again. But I am not able to downgrade update version with my university license (only possible with some very expensive professional one apparently, the joy of commercial software). So I guess the best we can do is to tell users to update to the most recent version (I think it is only possible to update to the most recent one anyway).

[egavazzi]

I added a warning to tell the user to update Matlab in case of segfault, as you suggested.
I also tested with older versions of Julia to see if the issue could have come from that side. But everything works, so I think it really is a matter of Matlab updates.

[musm]

Strange my copy still doesn't work and I'm on 2023b on Windows

julia> versioninfo()
Julia Version 1.9.3
Commit bed2cd540a (2023-08-24 14:43 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: 16 × AMD Ryzen 9 5900HS with Radeon Graphics
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-14.0.6 (ORCJIT, znver3)
  Threads: 16 on 16 virtual cores
Environment:
  JULIA_EDITOR = code.cmd
  JULIA_NUM_THREADS = auto

[MariusDrulea]

@musm Have you tried to do an update to Matlab, just in case?

[egavazzi]

@musm Interesting, I will boot on my Windows partition and try. All my tests were done on Linux.

[egavazzi]

Ok it just crashes on Windows for me too... Tested with R2023b and R2023a. R2021b works fine.

[musm]

Good to know. Yes, AFAIK this issue has been isolated to Windows and possibly MacOS

[egavazzi]

I vaguely remember having the issue on Linux also. But I might remember wrong 😅 Actually, I found a comment mentioning the problem on Linux: #200 (comment). Mathworks must have changed something in their build on Linux through an update that fixed the issue. But they didn't change it on Windows.

Then I propose to change the original warning to specify that the problem appears under Windows/MacOS. Maybe I will try to get access to a Mac computer through a colleague to test it there.

[curtcorum]

I am still seeing the segfault with matlab update R2023b Update 6 and 7. And julia 1.9 and 1.10.

Linux green 5.15.0-97-generic #107~20.04.1-Ubuntu SMP Fri Feb 9 14:20:11 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

MATLAB_ROOT=/usr/local/MATLAB/R2023b

(base) curt@green:~$ julia 
               _
   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.9.4 (2023-11-14)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org/ release
|__/                   |

julia> using MATLAB
[ Info: Precompiling MATLAB [10e44e05-a98a-55b3-a45b-ba969058deb6]

julia> mat"2+2"

[96360] signal (11.1): Segmentation fault
in expression starting at REPL[2]:1
Allocations: 2999 (Pool: 2987; Big: 12); GC: 0
Segmentation fault (core dumped)

(base) curt@green:~$ matlab -batch version
ans =
    '23.2.0.2515942 (R2023b) Update 7'

[curtcorum]

It works on Julia 1.9 with 2021b...

julia> mat"version"
"9.11.0.2358333 (R2021b) Update 7"

[egavazzi]

It works on my R2023b but I haven't applied any update (so it is basically Update 0). Now I know I should not update it 😅

julia> mat"version"
"23.2.0.2365128 (R2023b)"

So the problem is back on Linux but seems to be gone on macOS according to #223. Maybe we could change the warning message in the readme to a more general one.