diff --git a/.claude/memory/MEMORY.md b/.claude/memory/MEMORY.md index c432a3d1..34c1cd6b 100644 --- a/.claude/memory/MEMORY.md +++ b/.claude/memory/MEMORY.md @@ -229,3 +229,4 @@ - [EDR auto-isolates GuruRMM PowerShell](project_edr_rmm_autoisolation_fp.md) — Datto "Exfiltration Over HTTP" rule auto-isolates benign GuruRMM scripts fleet-wide (vwp-qbs 4h outage); watcher + narrow suppression deployed, policy fix pending Mike - [Background tasks — no shell `&`](feedback_background_task_no_ampersand.md) — with run_in_background:true, run the command in the foreground of the shell; adding `&`/`disown` forks + exits 0 instantly, orphaning a blocking wait so it never delivers (hit twice building ask-forum) - [ask-forum — human-in-the-loop via #ct-forum](../skills/ask-forum/SKILL.md) — ask a teammate a question in the private ct-forum Discord forum, get their human answer back in-session; forum-only, any human can answer, run the --wait as a proper background task +- [Claude Code Bun crash -> pin 2.1.110](reference_claude_code_bun_crash_pin.md) - v2.1.114+ segfaults on Win11 at new-session start; npm pin 2.1.110 + DISABLE_AUTOUPDATER=1; unpin on #55219 fix diff --git a/.claude/memory/reference_claude_code_bun_crash_pin.md b/.claude/memory/reference_claude_code_bun_crash_pin.md new file mode 100644 index 00000000..350ce841 --- /dev/null +++ b/.claude/memory/reference_claude_code_bun_crash_pin.md @@ -0,0 +1,26 @@ +--- +name: reference_claude_code_bun_crash_pin +description: Claude Code v2.1.114+ (Bun-based builds) segfault at startup on Windows 11 — fix is pin to 2.1.110 + DISABLE_AUTOUPDATER; unpin when upstream fix ships +type: reference +--- + +Claude Code v2.1.114+ ships as a Bun-compiled binary; on some Windows 11 builds (seen on +Howard-Home, Win11 build 26100, CC v2.1.207, 2026-07-13) NEW interactive sessions crash at +TUI init with `panic: Segmentation fault at address 0x3C` / "oh no: Bun has crashed". +Already-running sessions keep working (binary loaded in memory); only fresh launches die. +Tracked: anthropics/claude-code #55219 (+ matching upstream Bun issue). `--version`/`--help` +still work. It is a runtime bug inside the shipped binary — no config/cache/reinstall of the +same version helps. + +**Fix (verified working 2026-07-13, Howard-Home):** +1. `npm install -g @anthropic-ai/claude-code@2.1.110` (last pure-Node build; installs fine + even while an older session is running — the EPERM "cleanup" warning is just npm failing + to delete the locked running exe, the install itself succeeds). +2. Pin it — in `~/.claude/settings.json` add: + `"env": { "DISABLE_AUTOUPDATER": "1" }` + NOTE: a field named `autoUpdate` does NOT exist in the settings schema (validation + rejects it); the env-var route is the correct kill switch. +3. **Unpin when fixed:** watch issue #55219; then `npm i -g @anthropic-ai/claude-code@latest` + and remove the env entry. + +Do NOT switch to the native installer as a workaround — it is also Bun-based (same crash). diff --git a/errorlog.md b/errorlog.md index f12594db..0383b026 100644 --- a/errorlog.md +++ b/errorlog.md @@ -19,6 +19,14 @@ Categories (the `[type]` tag): _(none)_ = skill/command execution failure · +2026-07-13 | Howard-Home | bash/env | [friction] local Windows Defender killed /mingw64/bin/curl (Permission denied) mid-session on HOWARD-HOME; user disabled Defender to continue - RMM dispatch loop needs curl; consider Defender exclusion for Git-for-Windows mingw64 bin + +2026-07-13 | Howard-Home | mailbox/link-extraction | picked wrong INKY-wrapped link from Anthropic sign-in email - selected the url8792.mail.anthropic.com tracker (the Contact Support link, lands on support.claude.com) instead of the domain=claude.ai magic link; fix = decode ALL inky blobs and select the claude.ai/magic-link URL [ctx: email=secure-link-login mailbox=mike@azcomputerguru.com] + +2026-07-13 | Howard-Home | remediation-tool/get-token | [friction] vault_path not set in ~/.claude/identity.json on Howard-Home AGAIN - needed VAULT_ROOT_ENV=/d/vault inline (same as 7/2 session); fix identity.json or make get-token.sh resolve the repo-local identity.json [ctx: ref=2026-07-02 session log Problems Encountered] + +2026-07-13 | Howard-Home | mailprotector | HTTP 404 POST https://emailservice.io/api/v1/users/find_by_address: "Not found" [ctx: cmd=find-user] + 2026-07-13 | Howard-Home | packetdial | HTTP 404 GET https://pbx.packetdial.com/ns-api/v2/domains/farwestdrilling: {"code":404,"message":"Not Found"} [ctx: cmd=domain] 2026-07-13 | Howard-Home | packetdial | HTTP 404 GET https://pbx.packetdial.com/ns-api/v2/domains/fw.91912.service: {"code":404,"message":"Not Found"} [ctx: cmd=domain] diff --git a/session-logs/2026-07/2026-07-13-howard-billing-reconcile-june26-july10.md b/session-logs/2026-07/2026-07-13-howard-billing-reconcile-june26-july10.md index c33450e9..e34ee254 100644 --- a/session-logs/2026-07/2026-07-13-howard-billing-reconcile-june26-july10.md +++ b/session-logs/2026-07/2026-07-13-howard-billing-reconcile-june26-july10.md @@ -87,3 +87,13 @@ Entered the internal-labor fortnight ticket following the #32480 precedent: **#3 Final tally for the day: 5 tickets billed — Cascades #32538 (16.0h block, inv #68035 $0), Sif-Oidak #32526 (1.0h, inv #68037 $150), VWP #32543 (1.0h block, inv #68038 $0), Prairie Schooner #32544 (3.0h, inv #68039 $450, ticket held In Progress), ACG internal #32546 (28.0h, inv #68040 $0). Six explicit no-bill decisions (Scileppi 7/2+7/6, Dataforth MYDATA -> Mike's #32501, Goldstein OpenVPN, Four Paws cleanup, IMC, Grabb & Durando). Note on the 11:11 sync's vault push: the two vault commits (VWP-ROSE workstation login 7bcff18, Reliant channa-email swept into 95a2630) came from CONCURRENT Claude sessions on this machine — not this session. Both are legit credential adds; my sync's catch-all committed/pushed the pending one. No action needed. + +## Update: 15:15 PT — Claude Code login rescue + Bun crash pin + +Two afternoon incidents, both resolved: + +**1. Claude.ai login / MCP re-auth code hunt.** Howard's Claude Code needed re-auth; the "verification code" email to mike@azcomputerguru.com never appeared. Findings: (a) Anthropic's login email for this flow contains ONLY a magic sign-in link — no code; the code is displayed on the claude.ai page AFTER opening the link. (b) EOP message trace (Get-MessageTraceV2 via exchange-op adminapi — Get-MessageTrace is deprecated; investigator-exo token 401s on adminapi, exchange-op works) proved all 4 Anthropic emails DELIVERED with ~2-3 min lag — nothing quarantined. (c) azcomputerguru.com is NOT behind Mailprotector/CloudFilter (only gurushow.com is, account 16085); filtering is INKY (annotation-only) + EOP. No whitelist needed. (d) Wrong-link gotcha logged: the email has TWO INKY-wrapped tracker links — the url8792.mail.anthropic.com one is "Contact Support" (lands on support.claude.com); the real sign-in is domain=claude.ai -> https://claude.ai/magic-link#... — decode ALL INKY blobs (zlib(base64url(t param after "h."))) and pick the claude.ai one. Newest email finally forwarded to howard@azcomputerguru.com via Graph /forward (202). + +**2. Claude Code v2.1.207 Bun segfault on new sessions (Howard-Home).** New interactive sessions crashed at TUI init (panic: Segmentation fault at address 0x3C, Bun 1.4.0); the already-running session kept working. Known issue anthropics/claude-code #55219 (Bun-based builds since ~2.1.114). Fix applied + verified: npm downgrade to 2.1.110 (pure Node; installs fine next to a running session — EPERM cleanup warning is cosmetic) + DISABLE_AUTOUPDATER=1 via env block in ~/.claude/settings.json (the "autoUpdate" settings field does not exist — schema validation rejects it). Fresh session confirmed opening clean on 2.1.110. Memory written: .claude/memory/reference_claude_code_bun_crash_pin.md (+ MEMORY.md index). Unpin when #55219 fix ships. Native installer NOT a workaround (also Bun-based). + +Errorlog entries this stretch: syncro async-invoice quirk, remediation-tool/get-token vault_path friction (repeat, ref 7/2), mailbox link-extraction wrong-link.