[BUG] Frequent segfault crash in JSC GC (SlotVisitor::visitChildren) on v2.1.50

Resolved 💬 7 comments Opened Feb 21, 2026 by arunendra2024 Closed Mar 31, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Title:
Frequent segfault crash in JSC GC (SlotVisitor::visitChildren) on v2.1.50

Body:
## Bug Report

Claude Code version: 2.1.50 (verified via claude --version)
Install method: Standalone installer
OS: RHEL 9.4, Linux x86_64
Kernel: 5.14.0-427.13.1.el9_4.x86_64
Platform: linux
Shell: bash
Terminal: TERM_PROGRAM=claude (running inside Claude Code's own terminal)
Memory: 62GB total, ~45GB available

## Description

Claude Code is crashing frequently with a segmentation fault. This is happening across multiple concurrent sessions. The following crash report was generated by Claude Code upon crashing:

Crash report: https://bun.report/1.3.10/B_11423d3cmgEugogC4oi5wE+uzP41/ykFw8rzkF4lzzkF+iqikF867ukFm0gvkFowp5wEiguiBA2AA

## Crash Details (from the crash report link above)

  • Crash type: Segfault at address 0x00000000
  • Location: JSC::SlotVisitor::visitChildren in ClassInfo.h:115
  • Trigger: GC heap traversal during parallel drain (drainFromShared)
  • Frequency: Frequent, across multiple concurrent sessions

## Kernel Error Logs (from dmesg)

traps: claude[1706483] trap invalid opcode ip:3044dd4 sp:7fff153e0080 error:0 in 2.1.41[2ae7000+38cd000]
traps: claude[2098774] trap invalid opcode ip:3049514 sp:7ffffc46b0c0 error:0 in 2.1.50[2ae8000+3920000]
traps: claude[2099755] trap invalid opcode ip:3049514 sp:7ffd448da900 error:0 in 2.1.50[2ae8000+3920000]
traps: claude[2097729] trap invalid opcode ip:3049514 sp:7ffc16d49380 error:0 in 2.1.50[2ae8000+3920000]

Key observations from the logs:

  • The crash is a trap invalid opcode, not just a segfault — the binary is executing an illegal CPU instruction
  • Same instruction pointer ip:3049514 across all three v2.1.50 crashes — this is a deterministic bug, not random memory corruption
  • The issue existed in v2.1.41 as well (ip:3044dd4) at a slightly different offset, suggesting a long-standing bug
  • Three crashes from v2.1.50 within minutes of each other (PIDs 2097729, 2098774, 2099755)

## What I verified during the session

  • Confirmed v2.1.50 is the latest on npm registry — no upgrade available
  • Confirmed ample memory available (~45GB free), ruling out OOM
  • Found 5 versions installed locally (2.1.41 → 2.1.50), suggesting the issue persists across upgrades
  • No crash logs found in standard Claude Code log directories
  • Kernel logs confirm repeated crashes at the same instruction pointer

## Expected Behavior

  • Claude Code should run stable sessions without crashing mid-conversation
  • The runtime should not execute illegal CPU instructions
  • If recovery is needed, Claude Code should handle it gracefully instead of a raw trap/segfault

## Actual Behavior

  • Claude Code crashes abruptly during normal usage with trap invalid opcode
  • No warning or error message before the crash
  • Sessions terminate immediately, interrupting ongoing work
  • User must manually resume via claude code --resume
  • Happening repeatedly across multiple concurrent sessions and across versions

## Steps to Reproduce

  1. Run Claude Code v2.1.50 on RHEL 9.4 (x86_64), bash shell
  2. Use normally across multiple sessions
  3. Crash occurs intermittently with the trap/segfault shown above

---
Filed by Claude Code (self-report)
Model: claude-opus-4-6 | Date: 2026-02-22
Environment: CLAUDECODE=1, CLAUDE_CODE_ENTRYPOINT=cli
Process: PID 2100112, invoked via claude code --resume
Parent shell: TERM_PROGRAM=claude, SHELL=bash

What Should Happen?

  • Claude Code crashes abruptly with a segfault during normal usage
  • No warning or error message before the crash
  • The crash terminates the session immediately, interrupting ongoing work
  • The user has to manually resume the session via claude code --resume
  • This is happening repeatedly across multiple concurrent sessions
  • Post resume it occurs gain and a new session is observed in the resume and it doesn't have any memory

Error Messages/Logs

## Kernel Error Logs (from `dmesg`)

  traps: claude[1706483] trap invalid opcode ip:3044dd4 sp:7fff153e0080 error:0 in 2.1.41[2ae7000+38cd000]
  traps: claude[2098774] trap invalid opcode ip:3049514 sp:7ffffc46b0c0 error:0 in 2.1.50[2ae8000+3920000]
  traps: claude[2099755] trap invalid opcode ip:3049514 sp:7ffd448da900 error:0 in 2.1.50[2ae8000+3920000]
  traps: claude[2097729] trap invalid opcode ip:3049514 sp:7ffc16d49380 error:0 in 2.1.50[2ae8000+3920000]

Steps to Reproduce

  1. Run Claude Code v2.1.50 on RHEL 9.4 (x86_64)
  2. Use normally across multiple sessions
  3. Crash occurs intermittently with the segfault linked above

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.50 (verified via claude --version)

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Other

Additional Information

CPU: x86_64_baseline (may lack newer instruction set extensions)
Kernel: 5.14.0-427.13.1.el9_4.x86_64
Install: Standalone (not npm), symlinked from ~/.local/share/claude/versions/
Versions on disk: 2.1.41, 2.1.42, 2.1.47, 2.1.49, 2.1.50
Crash report: https://bun.report/1.3.10/B_11423d3cmgEugogC4oi5wE+uzP41/ykFw8rzkF4lzzkF+iqikF867ukFm0gvkFowp5wEiguiBA2AA

dmesg output showing repeated crashes:
traps: claude[1706483] trap invalid opcode ip:3044dd4 sp:7fff153e0080 error:0 in 2.1.41[2ae7000+38cd000]
traps: claude[2098774] trap invalid opcode ip:3049514 sp:7ffffc46b0c0 error:0 in 2.1.50[2ae8000+3920000]
traps: claude[2099755] trap invalid opcode ip:3049514 sp:7ffd448da900 error:0 in 2.1.50[2ae8000+3920000]
traps: claude[2097729] trap invalid opcode ip:3049514 sp:7ffc16d49380 error:0 in 2.1.50[2ae8000+3920000]

Note: "invalid opcode" suggests the binary may be using CPU instructions not supported on this x86_64_baseline processor. Same instruction pointer (ip:3049514) across all v2.1.50 crashes confirms this is deterministic, not random
corruption.

Filed by Claude Code (claude-opus-4-6) during a live session — environment verified via CLAUDECODE=1, CLAUDE_CODE_ENTRYPOINT=cli, PID 2100112.

View original on GitHub ↗

This issue has 7 comments on GitHub. Read the full discussion on GitHub ↗