[BUG] Segmentation fault (SIGSEGV) on install and startup — CachyOS Linux

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?

Both the native installer's own install step and every subsequent invocation of claude crash immediately with a segmentation fault (exit code 139). The crash is fully deterministic — it happens at the exact same instruction pointer address (ip 0000000001d10458) across separate runs. error 4 decodes to a user-mode read of an unmapped page (null pointer dereference) at address 0 — not random memory corruption, but something specific to this environment.

Claude Code is built on Bun (JavaScriptCore engine). Segfaults at address 0x0 on startup are a known crash pattern in Bun/JSC, usually tied to JIT code generation or memory-mapping assumptions that don't hold on non-standard kernels. CachyOS ships a custom kernel/scheduler (BORE) and distro-level compiler optimizations that differ from mainstream distros (Ubuntu/Fedora/Debian), which may conflict with how the JIT allocates executable memory (W^X / RWX mappings). Disabling the JIT via BUN_JSC_useJIT=0 did not prevent the crash, ruling out the most common JIT-specific workaround.

What Should Happen?

Installer completes and claude --version / claude doctor run normally.

Error Messages/Logs

Installer output:

Setting up Claude Code...
bash, linha 226: 7226 Falha de segmentação   (imagem do núcleo gravada) "$binary_path" install ${TARGET:+"$TARGET"}
Installation was killed before it could finish (exit code 139).


dmesg output (two separate crashes, same instruction pointer both times):

[   32.510354] claude[2745]: segfault at 0 ip 0000000001d10458 sp 00007ffc9cb40960 error 4 in 2.1.243[1b0f458,1953000+3b33000] likely on CPU 9 (core 4, socket 0)
[   32.510367] Code: cc cc 55 48 89 e5 48 89 fe 48 8b 05 82 19 78 03 48 89 f9 48 c1 e9 1d 48 8b 84 c8 80 00 00 00 89 f1 c1 e9 0d 81 e1 f8 ff 00 00 <48> 8b 3c 08 48 85 ff 74 1e 48 8b 07 64 48 33 04 25 00 00 00 00 75

[  652.461011] claude-2.1.243-[7226]: segfault at 0 ip 0000000001d10458 sp 00007ffe8e361c70 error 4 in claude-2.1.243-linux-x64[1b0f458,1953000+3b33000] likely on CPU 5 (core 6, socket 0)
[  652.461024] Code: cc cc 55 48 89 e5 48 89 fe 48 8b 05 82 19 78 03 48 89 f9 48 c1 e9 1d 48 8b 84 c8 80 00 00 00 89 f1 c1 e9 0d 81 e1 f8 ff 00 00 <48> 8b 3c 08 48 85 ff 74 1e 48 8b 07 64 48 33 04 25 00 00 00 00 75

Steps to Reproduce

1. Clean environment: remove any prior npm-based install (npm uninstall -g @anthropic-ai/claude-code, rm -rf ~/.local/share/claude)
2. Run the native installer:

``
curl -fsSL https://claude.ai/install.sh | bash
``

1. Installer downloads successfully, then crashes during its own install step (see Error Messages/Logs)
2. Running the installed binary directly (claude --version, claude doctor, etc.) crashes the same way
3. Try the known Bun/JavaScriptCore JIT workaround — does not help:

``
set -gx BUN_JSC_useJIT 0
curl -fsSL https://claude.ai/install.sh | bash
``

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.243 (Claude Code)

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Other

Additional Information

Diagnostic checks already ruled out:

• Architecture (uname -m): x86_64 — matches downloaded binary
• libc (ldd --version): ldd (GNU libc) 2.44 — glibc, not musl
• Memory (free -h): 15Gi total, 10Gi available — not an OOM kill (that gives exit 137, not 139)
• AVX support (grep avx /proc/cpuinfo): present
• Virtualization (systemd-detect-virt): none — bare metal, not a VM/container
• BUN_JSC_useJIT=0 workaround: tried, no change

Environment: CachyOS Linux (ID=cachyos, Arch-based, custom BORE kernel/scheduler and distro-level compiler optimizations), shell: fish, arch: x86_64.

Note: while filling out this form, GitHub flagged several potential duplicates with what looks like the exact same crash signature (same instruction pointer 0x0000000001d10458 in free), including #89371, #89369, and #89360.

Happy to run further diagnostics (strace, gdb on a core dump, testing other BUN_JSC_* flags, testing on the mainline kernel instead of linux-cachyos) if maintainers can suggest specific commands.