Segmentation fault (signal 11) in Julia 1.12.3 on macOS arm64 when using Plots + SymPy in JupyterLab Desktop

[r-de-r]

Dear Julia Developers,

I would like to report a reproducible segmentation fault issue I am encountering with Julia 1.12.x on macOS.

Summary

When using Julia 1.12.2 or 1.12.3 on macOS (Apple Silicon) in JupyterLab Desktop, the Julia kernel frequently crashes with:

signal 11 (2): Segmentation fault: 11

This occurs even when no plotting function is executed; merely loading the Plots package appears to trigger instability that later causes a crash during SymPy operations.

The same workflow worked reliably on Julia 1.11.x.

---

Environment

• Julia Version: 1.12.3 (also reproduced on 1.12.2)
• OS: macOS (arm64, Apple M1)
• Jupyter: JupyterLab Desktop
• Threads: JULIA_NUM_THREADS=1
• BLAS vendor: :lbt
• Python: system Python used via PyCall / SymPy
• Homebrew: completely removed from the system

versioninfo() output:

Julia Version 1.12.3
Commit 966d0af0fdf (2025-12-15)
Platform Info:
  OS: macOS (arm64-apple-darwin)
  CPU: Apple M1
  LLVM: libLLVM-18.1.7
  Threads: 1

---

Minimal Reproducible Example

The following code reliably reproduces the crash in JupyterLab:

using Plots
using SymPy

@syms R::positive r1::positive x1::positive r2::positive y2::positive

eq1 = x1^2 + r1^2 - (R + r1)^2
eq2 = (x1 - R)^2 + r1^2 - (R - r1)^2
eq3 = (R - r2)^2 + y2^2 - (R + r2)^2
eq4 = r2^2 + y2^2 - (R - r2)^2

res = solve([eq1, eq2, eq3, eq4], (R, r1, x1, y2))[1]

ans_r1 = res[2]   # segmentation fault occurs here

If using Plots is commented out, the crash does not occur.

---

Additional Observations

• The crash happens even if no plotting backend (e.g. pyplot()) is explicitly selected or used.
• The issue reproduces outside Jupyter as well, though it is more frequent inside JupyterLab Desktop.
• Setting BLAS.set_num_threads(1) does not help.
• Clearing ~/.julia/compiled does not help.
• Removing Homebrew and external BLAS/OpenMP libraries does not help.

---

Workaround

Switching from Plots.jl to PlotlyJS.jl completely avoids the issue.
This suggests a possible interaction problem involving Plots.jl, Python-based backends, or library initialization order in Julia 1.12.x.

---

Thank you very much for your time and for developing Julia.
Please let me know if further diagnostic information would be helpful.

Best regards,
r-de-r