Oliver Kieselbach 2b6565b3cb fix(agent): IME matcher line budget charges thread CPU cycles, regex timeouts get one starvation-gated retry - the wall-clock budget and the 1s .NET regex timeout both measured guest-scheduler starvation as match cost on CPU-saturated Hyper-V VMs (session b9f1d134: PS-AGENT-OUTPUT timed out on a line it genuinely matches in 5.8us, factor ~170,000, with 2.5s of process-wide log silence and CPU pinned at 100% during the DeviceSetup script burst; six spurious budget breaks the same pass); MatchLine now charges the per-line budget in cycles the poll thread actually executed (QueryThreadCycleTime, probed once, Stopwatch wall-tick fallback preserving the old behaviour exactly), and a RegexMatchTimeoutException whose attempt consumed < 0.5s of match cost (starvation, not spinning) is retried exactly once - counted into the new regexTimeoutRetries (recovered, NOT skipped work, no degraded event) while an attempt at/over the gate is skipped without a second helping of CPU, so regexTimeouts now means "really burned CPU"; the gate is unreachable in the wall-clock fallback by construction (no retry there); retries surfaced through health snapshot, restart-safe persistence, ime_tracker_degraded, ime_pattern_hits and agent_metrics_snapshot (ime_regex_timeout_retries); known limit: hypervisor-level vCPU steal stays invisible to the guest kernel and still charges cycles - strict improvement over wall clock, not perfect steal accounting ...

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-08-30 21:49:15 +02:00

..

AutopilotMonitor.Agent.V2

fix(security): IME log-pattern pack is anchored and linear (all 90 patterns start with ^, IME-APP-VERSION spells the report-delta JSON out instead of a chained [^}]*? scan) and the ImeLogTracker read loop is bounded — BoundedLineReader (32 MB per line, evidence: real 5.6 MB "Get policies" lines; exact byte bookmarks, no decoder read-ahead loss on cancel), 2 s per-line matching budget across all patterns, held-back unterminated EOF tails (≤1 MB, 1 s settle) instead of dropping them and raw-matching fragments, a complete entry after a dropped one is no longer swallowed as its close, one Warning per pass (oversizedLines/regexTimeouts/lineBudgetBreaks) instead of silent Debug timeouts; unanchored PS-SCRIPT-CONTEXT false positives on [Win32App] SideCarScript… lines (which overwrote the pending platform script's RunContext) are gone; --run-ime-matching assembles multiline entries like the tracker

2026-08-30 14:03:50 +02:00

AutopilotMonitor.Agent.V2.Core

fix(agent): IME matcher line budget charges thread CPU cycles, regex timeouts get one starvation-gated retry - the wall-clock budget and the 1s .NET regex timeout both measured guest-scheduler starvation as match cost on CPU-saturated Hyper-V VMs (session b9f1d134: PS-AGENT-OUTPUT timed out on a line it genuinely matches in 5.8us, factor ~170,000, with 2.5s of process-wide log silence and CPU pinned at 100% during the DeviceSetup script burst; six spurious budget breaks the same pass); MatchLine now charges the per-line budget in cycles the poll thread actually executed (QueryThreadCycleTime, probed once, Stopwatch wall-tick fallback preserving the old behaviour exactly), and a RegexMatchTimeoutException whose attempt consumed < 0.5s of match cost (starvation, not spinning) is retried exactly once - counted into the new regexTimeoutRetries (recovered, NOT skipped work, no degraded event) while an attempt at/over the gate is skipped without a second helping of CPU, so regexTimeouts now means "really burned CPU"; the gate is unreachable in the wall-clock fallback by construction (no retry there); retries surfaced through health snapshot, restart-safe persistence, ime_tracker_degraded, ime_pattern_hits and agent_metrics_snapshot (ime_regex_timeout_retries); known limit: hypervisor-level vCPU steal stays invisible to the guest kernel and still charges cycles - strict improvement over wall clock, not perfect steal accounting

2026-08-30 21:49:15 +02:00

AutopilotMonitor.Agent.V2.Core.Tests

fix(agent): IME matcher line budget charges thread CPU cycles, regex timeouts get one starvation-gated retry - the wall-clock budget and the 1s .NET regex timeout both measured guest-scheduler starvation as match cost on CPU-saturated Hyper-V VMs (session b9f1d134: PS-AGENT-OUTPUT timed out on a line it genuinely matches in 5.8us, factor ~170,000, with 2.5s of process-wide log silence and CPU pinned at 100% during the DeviceSetup script burst; six spurious budget breaks the same pass); MatchLine now charges the per-line budget in cycles the poll thread actually executed (QueryThreadCycleTime, probed once, Stopwatch wall-tick fallback preserving the old behaviour exactly), and a RegexMatchTimeoutException whose attempt consumed < 0.5s of match cost (starvation, not spinning) is retried exactly once - counted into the new regexTimeoutRetries (recovered, NOT skipped work, no degraded event) while an attempt at/over the gate is skipped without a second helping of CPU, so regexTimeouts now means "really burned CPU"; the gate is unreachable in the wall-clock fallback by construction (no retry there); retries surfaced through health snapshot, restart-safe persistence, ime_tracker_degraded, ime_pattern_hits and agent_metrics_snapshot (ime_regex_timeout_retries); known limit: hypervisor-level vCPU steal stays invisible to the guest kernel and still charges cycles - strict improvement over wall clock, not perfect steal accounting

2026-08-30 21:49:15 +02:00

AutopilotMonitor.BootstrapMsi

ci: Bootstrap-MSI build workflow - blob publish behind the download alias

2026-08-18 03:07:08 +02:00

AutopilotMonitor.DecisionCore.Tests

fix(core,backend): every payload-derived string stored into DecisionState is bounded (FactStringBounds, 256) and the reducer-verification replay streams signals — RealmJoinFacts.WithPackageStarted/Completed left PackageId/Version/Scope unbounded and deduped via a linear ordinal compare over up to 200 rows, so a forged 5000-signal stream with ~400 KB near-identical package ids cost minutes of CPU per GET sessions/{id}/reducer-verification while ReducerVerifier held all decoded payloads (~GBs) live; keys are now bounded BEFORE the dedupe scan (forged near-duplicates collapse to one row), Scope normalised to machine/user, and the same bound applied to failed app ids, HelloOutcome, ImeMatchedPatternId, EspAdvisoryFailureCategory and RJ product version/channel incl. WithVersionOverride — never truncating real telemetry (registry key names ≤ 255, GUIDs, enum-like strings); the agent adapter bounds packageId/version at the source; ReducerVerifier replays decode-fold-discard (head-signal check before the journal so stubbed streams stay silent, mid-stream truncation and count-imbalance semantics unchanged, pinned by tests); ISignalRepository.QueryBySessionAsync carries a cumulative PayloadJson budget (SignalQueryLimits, 32 M chars) reflected in the truncated flag of sessions/{id}/signals and reducer-verification; no ingest cap — stored signals stay verbatim

2026-08-30 20:06:51 +02:00

AutopilotMonitor.SummaryDialog

fix: close live catalog/type drift across web, backend and agent (audit P1)

2026-08-13 17:17:51 +02:00

AutopilotMonitor.SummaryDialog.Tests

feat(v2): WG Part-2 architectural cleanup + install-mode handoff fix

2026-05-04 18:36:37 +02:00

LICENSE

update license

2026-03-26 07:02:33 +01:00