Bun v1.3.5 crash: integer overflow panic on Windows with high memory usage
Resolved 💬 3 comments Opened Jan 31, 2026 by admin-meshlogic Closed Jan 31, 2026
Description
Claude Code consistently crashes with a Bun runtime panic on Windows 11 when memory usage reaches approximately 14-15GB RSS. The crash occurs during normal operations (API calls, web searches, /resume) with the error "integer does not fit in destination type".
Environment
- Claude Code Version: 2.1.27
- Bun Version: 1.3.5 (bundled)
- OS: Windows 11 (Build 26100.7623)
- CPU: x64 (baseline) with SSE4.2, AVX, AVX2
- System RAM: 34GB
Crash Details
Bun v1.3.5 (1e86cebd) Windows x64 (baseline)
Elapsed: 29184ms | User: 11531ms | Sys: 7546ms
RSS: 14.50GB | Peak: 14.50GB | Commit: 15.50GB | Faults: 4492302
panic(thread 15676): integer does not fit in destination type
oh no: Bun has crashed. This indicates a bug in Bun, not your code.
Reproduction Steps
- Launch Claude Code on Windows 11:
claude --dangerously-skip-permissions - Perform any operation that triggers API calls (asking a question, web search, /resume)
- Claude Code crashes when RSS approaches ~14-15GB
Observations
- Crash is 100% reproducible - happens on every session within 1-2 API calls
- Same crash signature every time (identical bun.report URL)
- Reinstalling Claude Code via
winget install Anthropic.ClaudeCodedoes not fix the issue - Memory usage climbs rapidly before crash
- Crash occurs regardless of session history size (tested with minimal sessions-index.json)
Bun Crash Report URL
Expected Behavior
Claude Code should operate normally without crashing due to integer overflow in Bun runtime.
Workaround Attempts (all failed)
- Reinstalling Claude Code
- Clearing/trimming session history
- Running with minimal configuration
- Running from clean PowerShell session
Additional Context
This appears to be a Bun 1.3.5 bug specifically affecting Windows with high memory usage scenarios. The integer overflow panic suggests a 32-bit integer is being used where 64-bit is needed for memory calculations.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗