Tokens: 27 in / 10,787 out / $1.6167

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Tokens: 27 in / 24,147 out / $1.8933

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…external to

Tokens: 21 in / 15,406 out / $1.3031

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ifferent lo

Tokens: 27 in / 17,560 out / $1.4334

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ections tab

Tokens: 25 in / 9,342 out / $1.3884

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…PuTTY fixes, dependency updates)

Upstream changes integrated:
- Sessions menu with Ctrl+Left/Right/1-9 tab navigation (SessionsMenu.cs)
- ODBC database connector support (OdbcDatabaseConnector.cs)
- PuTTY window class detection to avoid attaching to dialog windows
- Dependency updates: CEF 145, log4net 3.3.0, NUnit 4.5.0, WebView2 1.0.3800
- BinaryFormatter removal from ObjectListView
- Documentation: variables_reference.rst, keyboard shortcuts
- T4 TextTemplating import (conditional, for VS2026 compatibility)

Conflict resolution:
- Kept our non-blocking timer-based PuTTY window search + added upstream class detection
- Kept our detailed per-DB-type exception handlers + added upstream generic fallback
- Kept our 4-param ODBC connector with Database field support
- Kept our NormalizeType/authType in DatabaseConnectorFactory
- Made T4 TransformOnBuild conditional on VS2026 targets availability
- Added using System.Threading.Tasks to ConnectionInitiator.cs (preexisting bug)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

When Sonnet fails implementation, automatically retry with Opus using
2x timeout and 30 max turns. Passes previous attempt context so Opus
doesn't repeat mistakes. Mirrors existing pattern from chain_triage.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Mark 7 issues as needs_human (mRemoteNG#1182, mRemoteNG#1354, mRemoteNG#1427, mRemoteNG#1631, mRemoteNG#1661, mRemoteNG#1715, mRemoteNG#1752)
- Sync upstream comments and new issue mRemoteNG#3170
- Add chain-context from today's orchestrator session (20+ attempts)
- Clean up old chain-context files from Feb 17

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…eam merge

The upstream merge introduced ConnectionInitiator code referencing
WaitForIPAvailability and WaitForIPTimeout on ConnectionInfo, but the
backing fields and properties were missing from AbstractConnectionRecord.
Adds both properties to resolve CS1061 build errors.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…arly

The two is_last branches inside the AGENT_CHAIN loop did return False
directly, preventing the Opus fallback block after the loop from ever
executing. Changed both to break so control flows to the Opus section.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- SOCKS5 proxy: prevent auth downgrade by only offering method 0x02
  when credentials are configured (cs/user-controlled-bypass)
- iis_orchestrator: fix ReDoS in test output parser regex (py/redos)
- add_PR_2_chlog.yml: add explicit permissions block (contents: write)
- post_2_Reddit.yml: add empty permissions block (no repo access needed)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Equivalent to upstream PR mRemoteNG#3169 (Renovate bot), applied locally on fork.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- actions/checkout v4 → v6
- actions/upload-artifact v4 → v6
- actions/download-artifact v4 → v7
- signpath/github-action-submit-signing-request v1 → v2

Consolidates Dependabot PRs #15, #16, #17, #18.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Cache SecureString in XmlConnectionsDecryptor (avoid repeated ConvertToSecureString)
- Parallel pre-decryption of encrypted fields using Parallel.ForEach
- Static PropertyInfo cache in ConnectionInfoInheritance (avoid repeated reflection)
- Enum parse cache in XmlExtensions (avoid repeated Enum.TryParse)

Root cause: PBKDF2 with 600K iterations runs per encrypted field sequentially.
570 connections × 4 password fields × ~150ms = ~85 seconds.
Parallel decryption reduces this to ~85s / N_cores.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Create per-thread XmlConnectionsDecryptor instances in Parallel.ForEach
to avoid shared BouncyCastle cipher state corruption.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

744 commits since beta.2: 585/838 issues addressed (70%), 2817 tests (0 failures),
7 new protocols (VMRC, MSRA, OpenSSH, Winbox, WSL, Terminal, Serial),
4 security alerts fixed, 81s→ms deserialization, orchestrator v2 with
self-healing supervisor, 67 upstream commits merged.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…→Opus, 12 failure modes

The orchestrator uses only Claude Code (AGENT_CHAIN = ["claude"]) with
Sonnet→Opus model escalation, not 3 separate agents. The supervisor
handles 12 failure modes (FM1-FM12), not 8.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Multiple .slnx files at repo root cause MSB1011 "more than one project
or solution file" when dotnet restore runs without explicit target.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Specs project only declared x86;x64 platforms, causing arm64
self-contained build to fail with PlatformTarget mismatch.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The Specs project needs explicit PlatformTarget=ARM64 for the arm64
self-contained publish. Without it, MSBuild defaults to x86 which
conflicts with RuntimeIdentifier=win-arm64.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…8 MB)

Accumulated debris from v1.79.0→v1.81.0 development (Feb 10-24, 2026):

Root directory (45 files):
- 10x ISSUE_*_FIX_DETAILS.md, TRIAGE_1214.md — issue analysis docs
- 4x standalone C# debug files (TestEncoder, repro_issue_980, etc.)
- 3x test .slnx files (MyTests, MyTests2, mRemoteNGTests)
- 7x text outputs (build_out, config_tests, csv_test_output, etc.)
- 5x VNC/RDP inspection scripts (_inspect_vnc, check_mstsc, etc.)
- 6x ad-hoc build/utility scripts (build_test, hunt-crasher, etc.)
- 5x generated docs (VISUAL_EXAMPLES, PANEL_BINDING_FEATURE, etc.)
- 4x orphan files (issue1944.png, monitor_orch.py, output.json, etc.)
- 9x build logs + Codex temp files
- temp_test_caps/ directory (4 files)

.project-roadmap/ (335 files):
- 210x tmp*.txt/jsonl — orchestrator temp prompt files (498 KB)
- 122x chain-context JSONs for closed issues (471 KB)
- 40x historical/ — v1.79.0 + v1.80.0 execution logs & comments
- 6x root docs (CURRENT_PLAN, CVE assessment, ISSUE_*, email draft)
- orchestrator-status.json.bak

Also deleted (untracked, not in commit):
- mRemoteNGTests/TestResults/ — 296 MB empty UUID directory
- 6x orchestrator log files (11 MB)
- __pycache__/, .auto-claude-security.json, .claude_settings.json, NUL
- test-output.txt, test-results.txt, test-results2.txt, test.ps1, etc.

Preserved: all source code, active scripts (iis_orchestrator.py,
orchestrator_supervisor.py), issues-db, chain-context for open issues,
DEVELOPER_GUIDE.md, LESSONS.md, agents.md, GEMINI.md, build/test runners.

Build verified: 0 errors. No source code affected.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Python bytecode (.pyc) and orchestrator temp outputs should never be
in version control. Added __pycache__/, *.pyc, triage_result.json,
and output.json to .gitignore to prevent future tracking.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

MSBuild passes Platform=ARM64 (uppercase) but the condition checked
for 'arm64' (lowercase). MSBuild conditions are case-sensitive.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The .sln file mapped mRemoteNGSpecs arm64 configurations to x86,
causing NETSDK1032 when MSBuild tried to publish with
RuntimeIdentifier=win-arm64 but PlatformTarget=x86.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…rogress

- Add WaitForIPAvailability/WaitForIPTimeout to test expected properties
  (ConfigWindowGeneralTests + SerializableConnectionInfoAllPropertiesOfType)
- Rewrite run-tests-core.sh v4: parallel execution of 9 groups from shared DLL
  with sequential retry for crashed groups (~33s vs ~8min)
- Add _run_streaming() to orchestrator for real-time test progress in log
- Fix grep -oP to -oE for MSYS2 compatibility
- Increase TEST_TIMEOUT from 300s to 600s

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…lization

- Add TestScope singleton snapshot/restore for test isolation
- Add namespace-level SetUpFixtures (Connection, Integration, Tree, UI)
- Rewrite run-tests-core.sh with MAX_PARALLEL=2 sliding window (0 crashes)
- Add WaitForIPAvailability/WaitForIPTimeout to ConnectionInfoInheritance
- Complete WaitForIP serialization in XML, CSV, and XSD schema
- Update README with 2,916 tests and new 9-group runner architecture

2916/2916 tests passing, 0 failures, 0 crashes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…onfiguration

Tokens: 27 in / 38,821 out / $2.6698

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The supervisor progress report showed 0/246 even after successful
implementations because `processed` only summed triaged+wontfix+
needs_info+duplicate, missing implemented and failed counts.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

… tab

Tokens: 18 in / 16,265 out / $1.2143

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Qodo ignores PRs where head=default branch. Create both head and base
as temporary review/* branches so Qodo triggers correctly.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Qodo Code Review ignores PRs created by bots (github-actions[bot]).
Replace the workflow_dispatch approach with a local bash script that
creates PRs under the user's identity.

Usage: ./qodo-review.sh [commits] [branch]
Cleanup: ./qodo-review.sh cleanup

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>