Skip to content

CrashScope 1.1.0

Choose a tag to compare

View all tags
@marcelosofficial-ctrl marcelosofficial-ctrl released this 11 Sep 19:46
· 16 commits to main since this release
v1.1.0
341487a

CrashScope 1.1.0

Publication provenance

  • validated private engineering commit: e51943f6b7aecaf98223ef30cbec10ef94c1eba1
  • privacy-safe public tag commit: 341487a0f82c15523611a92fb2829a16cf97d2a7
  • exact shared Git tree: 908cb6d12da1c360a663ce85b9a95c2469464c13
  • portable SHA-256: 4ac728d03d634218add25d94e69a1a55ddd3283b6afc786dc155e1e7e15360a9
  • installer SHA-256: eb0acdab0f1da5ae5fbaeceadbd3691ee188f0d42204ddffc9527aa38f28062f

The full engineering history remains private in CrashScope-development because historical commit metadata contained personal author/committer email information. The public tag preserves the exact validated source tree without exposing that history.

CrashScope 1.1.0 release notes

CrashScope 1.1.0 extends the frozen 1.0 line with a generic external-evidence provider architecture and ConfigTrace integration while preserving the local-first, privacy-first, evidence-first, and low-overhead design.

Public publication is still intentionally deferred until the coordinated release sweep. The exact final v1.1.0 tag target is the locally frozen 1.1 release commit produced after validation; no tag or GitHub Release is created by the local release-engineering batches.

What changed from 1.0

Generic external evidence providers

CrashScope now has a generic EvidenceEvent provider boundary for process-isolated evidence sources. Provider evidence is mapped into the existing incident model as bounded chronological evidence rather than as a separate causal system.

The central rule remains:

A nearby event is correlation evidence unless stronger evidence supports a causal claim.

For example, CrashScope can say that a renderer setting changed seconds before an incident. It does not claim that the setting change caused the incident.

ConfigTrace 1.0.1 integration

CrashScope 1.1 bundles ConfigTrace as its first external sidecar provider:

  • ConfigTrace is OFF by default;
  • the user explicitly selects one existing absolute configuration root;
  • ConfigTrace starts only while a workload is actively monitored;
  • the sidecar stops with the workload;
  • journals live under CrashScope-owned local application data;
  • nearby semantic configuration changes are imported as Context evidence;
  • ConfigTrace failure is isolated and must not take CrashScope down;
  • portable and installed layouts use the same bundled providers\ConfigTrace\configtrace.exe path.

Bundled ConfigTrace provenance:

  • version: 1.0.1
  • source commit: b629c970dfc14fca5df1e0ef2b0d1d07d0d8c56c
  • Windows EXE SHA-256: fe1c470a58402e82e97ee529c6a6b02822430da70e65ffc5fc5a71359ad4e521
  • license: MIT
  • bundled license path: providers\ConfigTrace\LICENSE.txt

ConfigTrace 1.0.1 includes a privacy fix discovered during CrashScope integration: sensitive-key recognition now handles camelCase/PascalCase names such as apiToken, accessToken, refreshToken, clientSecret, sessionId, and APIKey. Real JSONL watcher validation proved that sensitive values remain redacted while change fingerprints continue to work.

Settings and dashboard

CrashScope settings schema v3 adds:

  • configTraceEnabled
  • configTraceRootPath

Existing 1.0-era settings migrate with ConfigTrace disabled and no configured root.

The dashboard adds a ConfigTrace control that:

  • saves one absolute configuration root;
  • prevents enabling the provider without a valid saved root;
  • explains that ConfigTrace only runs during monitored workloads;
  • keeps the provider disabled by default.

Manual diagnostic markers

User-requested safe diagnostic markers now use the same bounded provider-evidence mapping as live incidents. ConfigTrace evidence attached to a marker remains Context; the marker classification stays UserDiagnosticMarker.

Validation status

The 1.1 integration checkpoint passed:

  • 239 .NET tests passed, 0 failed
  • ConfigTrace Rust formatting/tests/build validation
  • ConfigTrace 1.0.1 real watch-journal privacy smoke
  • portable sidecar packaging/hash/provenance checks
  • real ConfigTrace OFF/ON lifecycle validation
  • real semantic config-change journal capture
  • secret absent from the raw ConfigTrace journal
  • secret absent from CrashScope incident JSON
  • exactly one ConfigTrace evidence item observed in the end-to-end marker validation
  • that evidence item mapped as Context
  • bundled sidecar started with the monitored workload and stopped with it
  • real user CrashScope state restored after validation

A short idle integration measurement during the pre-release A11 gate recorded approximately:

  • ConfigTrace OFF: Agent average CPU 0.29%, Agent peak working set 97.63 MB
  • ConfigTrace ON: Agent average CPU 0.1937%, Agent peak working set 99.74 MB
  • ConfigTrace sidecar average CPU 0% in that sample, peak working set 4.95 MB

Those are targeted integration measurements, not the final 1.1 performance seal. Final release validation separately rechecks performance and cross-machine behavior.

Packaging

The self-contained 1.1 portable and installer packages include:

CrashScope.exe
wwwroot/
providers/
  ConfigTrace/
    configtrace.exe
    LICENSE.txt
LICENSE.txt
THIRD-PARTY-NOTICES.md
BUILD-INFO.txt

The installer recursively installs the same validated portable payload. It remains per-user, non-admin for normal operation, does not create a Windows service, does not force Start with Windows, and preserves CrashScope user data.

Not included in 1.1

GapTrace/network evidence is intentionally not part of CrashScope 1.1. The generic provider boundary is designed to support future evidence sources without making them a requirement for this release.

Code signing is also not implemented in 1.1.0; unsigned-build SmartScreen reputation warnings may occur.

Final local release validation

CrashScope 1.1 local release validation is complete.

Validated release boundary:

  • exact v1.1.0 release/tag target: e51943f6b7aecaf98223ef30cbec10ef94c1eba1
  • exact portable SHA-256: 4ac728d03d634218add25d94e69a1a55ddd3283b6afc786dc155e1e7e15360a9
  • exact installer SHA-256: eb0acdab0f1da5ae5fbaeceadbd3691ee188f0d42204ddffc9527aa38f28062f
  • automated .NET tests: 239/239 PASS

Installed upgrade validation passed for the exact 1.0.0 -> 1.1.0 path:

  • settings schema v2 -> v3 migrated correctly;
  • existing Auto Assist and retention settings were preserved;
  • ConfigTrace was introduced disabled with a null root by default;
  • a persisted 1.0 diagnostic-marker incident survived the upgrade;
  • startup state survived the upgrade;
  • 1.1 uninstall removed the application and owned startup value while preserving user data;
  • the original pre-validation user state was restored exactly.

Final main-PC performance/cleanliness measurements:

  • ConfigTrace OFF: CrashScope 0.2365% average CPU, 91.71 MB average working set, 94.58 MB peak working set, 34.76 MB average private memory;
  • ConfigTrace ON with an active controlled workload: CrashScope 0.1966% average CPU, 95.36 MB average working set, 97.31 MB peak working set, 34.06 MB average private memory;
  • ConfigTrace sidecar: 0% measured average CPU, 4.96 MB average working set, 4.99 MB peak working set, 0.91 MB average private memory;
  • stale WebSocket frames: 0;
  • stream delivery misses: 0;
  • ConfigTrace journal created successfully;
  • plaintext test secret in the journal: not found;
  • sidecar start/stop lifecycle: PASS.

Genuine second-PC validation also passed:

  • Windows 10;
  • Intel Core i5-3210M;
  • Intel HD Graphics 4000;
  • non-elevated validation;
  • exact portable SHA matched;
  • automated validator exited 0;
  • final manual dashboard inspection: PASS.

The branch may contain a later documentation-only commit recording these completed gates. That later documentation commit is intentionally not the v1.1.0 binary release boundary. The release/tag target remains the exact runtime commit above so source provenance matches the binaries that were actually validated.

Public release publication is complete for this version.

Assets 5
Loading