Description
pear.exe v3.0.0 crashes immediately and deterministically on every invocation on Windows 11, including the simplest possible command (pear --version). No output is produced before the crash. This happens whether invoked directly (AppData\Local\Programs\pear\pear.exe) or via the npm-installed pear wrapper (AppData\Roaming\npm\pear.cmd, which just spawns the same binary) — both paths hit the identical crash.
Environment
- OS: Windows 11 Home, build 10.0.26200
- CPU: 11th Gen Intel Core i5-1135G7 (Tiger Lake) — AVX2/AVX-512 capable, ruling out an instruction-set mismatch
- pear version: 3.0.0 (
latest dist-tag — confirmed current at time of report)
- Node.js: v24.18.0
- npm: 11.16.0
pear.exe SHA256: A07EC55E03DFAA6EEDFA96633AB4072A823AE9B59AA90CE60C68213817C83200
Steps to reproduce
(also reproduces identically with pear touch, npx pear --version, and invoking pear.exe directly with a full path)
Actual result
Process exits immediately with exit code -1073740791 (0xC0000409 / STATUS_STACK_BUFFER_OVERRUN) and no stdout/stderr output at all. This exit code is also used by __fastfail()/abort() on Windows, and critically, that path bypasses Windows Error Reporting entirely — so no crash dump or WER event is ever created.
Expected result
Normal CLI output, e.g. the version string for pear --version.
What I've ruled out
- Stale/corrupted install — reinstalled via two separate methods (npm global install and the
pear.sh bootstrap installer); same crash both times.
- Corrupted local state — renamed
%APPDATA%\pear aside entirely; crash persists with a fully fresh state directory.
- Missing VC++ Redistributable — confirmed present.
- Windows Defender / real-time protection — confirmed via
Get-MpComputerStatus that RealTimeProtectionEnabled was genuinely False at test time (not just toggled in the UI, which Tamper Protection can silently revert); crash was byte-for-byte identical with protection off.
- CPU instruction-set incompatibility — CPU supports AVX2/AVX-512.
- Outdated pear version —
3.0.0 is the current latest on npm at time of testing.
- npm-wrapper vs standalone binary discrepancy — traced
pear.cmd's source; it just spawns the exact same pear.exe, so both entry points are hitting the identical crash in the identical binary.
Given the crash is 100% reproducible on the very first command with no state involved yet, this looks like a genuine platform-specific bug in the 3.0.0 Windows build rather than an environment/AV issue on this particular machine. Happy to provide further diagnostics (WinDbg trace, etc.) if that would help narrow it down.
Description
pear.exev3.0.0 crashes immediately and deterministically on every invocation on Windows 11, including the simplest possible command (pear --version). No output is produced before the crash. This happens whether invoked directly (AppData\Local\Programs\pear\pear.exe) or via the npm-installedpearwrapper (AppData\Roaming\npm\pear.cmd, which just spawns the same binary) — both paths hit the identical crash.Environment
latestdist-tag — confirmed current at time of report)pear.exeSHA256:A07EC55E03DFAA6EEDFA96633AB4072A823AE9B59AA90CE60C68213817C83200Steps to reproduce
(also reproduces identically with
pear touch,npx pear --version, and invokingpear.exedirectly with a full path)Actual result
Process exits immediately with exit code
-1073740791(0xC0000409/STATUS_STACK_BUFFER_OVERRUN) and no stdout/stderr output at all. This exit code is also used by__fastfail()/abort()on Windows, and critically, that path bypasses Windows Error Reporting entirely — so no crash dump or WER event is ever created.Expected result
Normal CLI output, e.g. the version string for
pear --version.What I've ruled out
pear.shbootstrap installer); same crash both times.%APPDATA%\pearaside entirely; crash persists with a fully fresh state directory.Get-MpComputerStatusthatRealTimeProtectionEnabledwas genuinelyFalseat test time (not just toggled in the UI, which Tamper Protection can silently revert); crash was byte-for-byte identical with protection off.3.0.0is the currentlateston npm at time of testing.pear.cmd's source; it just spawns the exact samepear.exe, so both entry points are hitting the identical crash in the identical binary.Given the crash is 100% reproducible on the very first command with no state involved yet, this looks like a genuine platform-specific bug in the 3.0.0 Windows build rather than an environment/AV issue on this particular machine. Happy to provide further diagnostics (WinDbg trace, etc.) if that would help narrow it down.