v0.14.0

What's Changed

• Bump DiagnosedBenchmarks sample to benchmarkdotnet 0.13.0 by @pedrosakuma in #358
• ci: strace-based Linux crash-repro job for SampleProfiler flake (refs #147) by @pedrosakuma in #359
• ci: stop strace -f from hanging on leaked tracees in crash-repro job by @pedrosakuma in #360
• ci: add non-ptrace LD_PRELOAD mmap-trace crash-repro job by @pedrosakuma in #361
• test: un-quarantine 3 hardcoded-skip CPU sampler tests via SkipOnLinuxCiFact by @pedrosakuma in #362
• chore: sweep repo-slug URLs to dotnet-diagnostics after rename by @pedrosakuma in #363
• docs: add v0.13.0 output-examples gallery + reorganize docs hub by track by @pedrosakuma in #364
• docs: reframe ptrace gate as cross-surface (MCP + CLI; BDN immune) by @pedrosakuma in #366
• docs: add practical authorization reference (docs/authorization.md) by @pedrosakuma in #368
• docs: scrub RFC 0001 citations to authorization.md; remove RFC 0001 by @pedrosakuma in #369
• docs: normalize tool-reference to English + repair drilldown section by @pedrosakuma in #370
• docs: scrub RFC 0002 citations and delete the RFC by @pedrosakuma in #371
• docs: remove 5 design-spike docs, keep the orchestrator spec by @pedrosakuma in #372
• chore(ghcr): publish container under the bundle name dotnet-diagnostics by @pedrosakuma in #373
• docs: slim root README + docs hub into drill-down structure by @pedrosakuma in #374
• docs: scrub last RFC 0001 dangling references by @pedrosakuma in #376
• ci: skip docs-only PRs + cache Docker layers + trim crash-dump uploads by @pedrosakuma in #377
• docs: add offenders-report example to output-examples by @pedrosakuma in #378
• feat: emit loadBase on native frames for ASLR-resilient handoff by @pedrosakuma in #379
• Persist per-row dispersion stats in KeyMatrixRow (#347 W3-2) by @pedrosakuma in #380
• Retire SampleDiffer: route typed N=2 diff through comparable projectors by @pedrosakuma in #381
• Normalize Portuguese "bootstrap implícito" to English (#389) by @pedrosakuma in #396
• feat: Wave A auto-hints (regex backtracking, channel write backpressure, active-timer-count) by @pedrosakuma in #399
• feat(cli): Wave A ergonomics — progress spinner, query redirect, dump path disclosure by @pedrosakuma in #400
• Add GCHeapStats per-gen + pinned-object trend view to gc-events (#384) by @pedrosakuma in #398
• Trim unified-tool [Description] bloat; move enums to discriminator params (#385) by @pedrosakuma in #397
• feat(cli): opt-in --launch child-launch dev mode for zero-privilege ptrace attach (#365) by @pedrosakuma in #382
• feat: add collect_events(kind=kestrel) Kestrel HTTP-server collector by @pedrosakuma in #402
• Add CLI shell completion scripts by @pedrosakuma in #401
• feat(networking): outbound-network collector (collect_events kind=networking) by @pedrosakuma in #404
• Add startup EventPipe collector by @pedrosakuma in #403
• docs(changelog): cut 0.14.0 (Phase 12 Wave A + Wave B) by @pedrosakuma in #405

Full Changelog: v0.13.0...v0.14.0

v0.13.0

What's Changed

• Add cpu + allocation demo benchmarks to DiagnosedBenchmarks sample by @pedrosakuma in #356
• Attribute allocation origin (call site), not just type by @pedrosakuma in #357

Full Changelog: v0.12.0...v0.13.0

v0.12.0

What's Changed

• feat(benchmarkdotnet): add cpu kind with per-frame hotspots to the diagnoser by @pedrosakuma in #354
• chore(benchmarks): consume published dotnet-diagnostics-benchmarkdotnet 0.11.0 by @pedrosakuma in #353
• feat(benchmarkdotnet): add allocation kind (per-type GCAllocationTick) to the diagnoser by @pedrosakuma in #355

Full Changelog: v0.11.0...v0.12.0

v0.11.0

What's Changed

• chore(deps): Bump the actions-minor-patch group across 1 directory with 2 updates by @dependabot[bot] in #306
• chore(deps): Bump azure/setup-helm from 4.3.0 to 5.0.0 by @dependabot[bot] in #274
• docs: review for two deliverables (CLI + MCP server) by @pedrosakuma in #312
• feat: CPU drilldown views (top-methods/by-module/by-namespace/hot-path/caller-callee) by @pedrosakuma in #316
• feat: GC drilldown views (timeline/longestPauses/byGeneration) by @pedrosakuma in #317
• feat: events catalog (collect_events kind=catalog) by @pedrosakuma in #318
• feat: DATAS analysis (collect_events kind=datas) by @pedrosakuma in #319
• C1: ComparableSnapshot substrate + datas/counters projectors by @pedrosakuma in #328
• feat(comparison): N-ary SnapshotDiffer engine (C2 #324) by @pedrosakuma in #329
• C4: extend compare_to_baseline for comparable snapshots by @pedrosakuma in #331
• Add CLI comparable snapshot compare by @pedrosakuma in #332
• C5: add N-way query_snapshot journey diffs by @pedrosakuma in #333
• C6 journey diff resource links by @pedrosakuma in #334
• docs(comparison): comparative + N-way trend investigation playbook (C7 #322) by @pedrosakuma in #335
• W2-0 scaffold: registry-driven diff gating by @pedrosakuma in #342
• W2-A: comparable projectors for heap/cpu/native/allocation (N-ary) by @pedrosakuma in #343
• W2-B: comparable projectors for contention (key-set) + threadpool (scalar) by @pedrosakuma in #344
• W2-C: expose dispersion mode on query_snapshot/compare_to_baseline + CLI by @pedrosakuma in #345
• refactor: drop DotnetDiagnosticsMcp prefix; scope Mcp to the MCP server (refs #346) by @pedrosakuma in #349
• docs: BenchmarkDotNet diagnoser shipping design (refs #346) by @pedrosakuma in #348
• build: ship DotnetDiagnostics.Core as a referenceable NuGet library (refs #346) by @pedrosakuma in #350
• feat(benchmarkdotnet): in-process diagnoser library by @pedrosakuma in #351
• feat(benchmarkdotnet): port DiagnosedBenchmarks sample to the in-process lib by @pedrosakuma in #352

Full Changelog: v0.10.0...v0.11.0

v0.10.0

What's Changed

• Improve tool discoverability and triage memory coverage (#280) by @pedrosakuma in #281
• Native (unmanaged) allocation sampler over perf (collect_sample kind=native-alloc) by @pedrosakuma in #282
• [CLI Phase 0] Relocate AddDiagnosticCoreServices Server to Core by @pedrosakuma in #289
• Extract host-neutral process-inspection use cases into Core (#285) by @pedrosakuma in #290
• Add diag host-mode spike inside the Server binary (#287) by @pedrosakuma in #291
• feat(cli): standalone DotnetDiagnosticsMcp.Cli (Core-only) — processes + capabilities by @pedrosakuma in #292
• Extract collect orchestration into Core; wire CLI collect (#288 PR2) by @pedrosakuma in #293
• Extract attach-guard + symbol-path validation into Core (#288 PR3a) by @pedrosakuma in #295
• Harden flaky RuntimeConfig + ThreadPool live tests (#294) by @pedrosakuma in #296
• Extract heap-inspection + process-dump into Core; wire CLI inspect-heap + dump (#288 PR3b) by @pedrosakuma in #297
• CLI query + get-bytes commands (Core-only) — #288 PR4 by @pedrosakuma in #298
• Remove Server diag spike + package standalone CLI (#288 PR5) by @pedrosakuma in #299
• CLI usability: project MCP-audience hints into CLI vocabulary (Tier 1) by @pedrosakuma in #303
• CLI usability Tier 2: de-MCP-ify capability/threadpool notes, per-subcommand --help (#302) by @pedrosakuma in #304
• Add stateful session REPL to the Core-only CLI (#300 PR1) by @pedrosakuma in #305
• Extract Core-owned dead-PID handle eviction (#300 PR2) by @pedrosakuma in #307
• CLI session heap query: extract projection views to Core (#300) by @pedrosakuma in #308
• Extract cpu-sample call-tree query to Core for CLI session reuse by @pedrosakuma in #309
• Extract thread-snapshot + off-cpu queries to Core for CLI session reuse by @pedrosakuma in #310
• feat(cli): bind a session target pid once (strand C of #300) by @pedrosakuma in #311

Full Changelog: v0.9.1...v0.10.0

v0.9.1

What's Changed

• fix(core): Windows smoke-test robustness (process enum, contention/threadpool call stacks, EventPipe start timeout) by @pedrosakuma in #278

Full Changelog: v0.9.0...v0.9.1

v0.9.0

What's Changed

• feat(symbols): resolve native/unresolved addresses to (module, rva | unmapped) by @pedrosakuma in #276

Full Changelog: v0.8.0...v0.9.0

v0.8.0 - GC Overlay & Diagnostic Improvements

Phase 13: GC Overlay & Diagnostic Improvements

New Features

GC Overlay for Activities - Correlate GC pause events with ActivitySource spans to identify requests impacted by garbage collection:

# Capture activities and GC simultaneously
collect_events(kind="activities") → act-handle
collect_events(kind="gc") → gc-handle

# Correlate to find impacted spans
query_collection(handle="act-handle", view="gc-overlay", gcHandle="gc-handle")

Output includes:

• Impacted activities ranked by GC pause percentage
• Per-activity overlap metrics (pauseMs, pausePercent)
• Summary totals (impacted count, total GC overlap time)

Improvements

Contention Diagnostics on Linux - Improved hints when collect_events(kind="contention") returns no events on Linux (known EventPipe limitation). Now explicitly suggests:

• collect_thread_snapshot() followed by query_snapshot(view="lock-graph") to see contended SyncBlocks with waiter counts

Technical Notes

• LTTng-based contention was evaluated but requires CAP_SYS_ADMIN which is too invasive for sidecar deployments
• SyncBlock-based approach works well with existing CAP_SYS_PTRACE requirement
• Allocation call-tree already existed via collect_sample(kind="allocation") + query_snapshot(view="call-tree")

Full Changelog: v0.7.2...v0.8.0

v0.7.2 — Exception & Latency Indicators

What's New

Extended TopIndicators — Two new metrics for error and latency diagnostics:

Indicator	Normal	Elevated	High	Critical
exception-count	0	1-10	10-50	>50
request-duration-p95	<100ms	100-500ms	500ms-2s	>2s

These surface error spikes and latency issues in the triage view even when the overall verdict is "healthy", enabling proactive optimization.

Full Changelog

v0.7.1...v0.7.2

Full Changelog: v0.7.1...v0.7.2

v0.7.1 — Ranked Top Indicators

What's New

Ranked Top Indicators for Proactive Optimization

The inspect_process(view="triage") now always returns TopIndicators — a ranked list of the most notable metrics even when the system is healthy. This enables proactive optimization, not just reactive firefighting.

Example output:

{
  "verdict": "healthy",
  "severity": "Healthy",
  "topIndicators": [
    {"name": "time-in-gc", "value": 8, "unit": "%", "score": 28, "level": "elevated"},
    {"name": "cpu-usage", "value": 5, "unit": "%", "score": 3, "level": "normal"},
    {"name": "alloc-rate", "value": 15, "unit": "MB/s", "score": 12, "level": "elevated"}
  ]
}

Score bands:

Metric	Normal	Elevated	High	Critical
cpu-usage	<30%	30-70%	70-90%	>90%
time-in-gc	<5%	5-15%	15-30%	>30%
queue-length	<10	10-50	50-200	>200
contention	<3	3-10	10-50	>50
alloc-rate	<10MB/s	10-50	50-100	>100
gen2-gc	0	1-2	3-9	≥10

This helps identify optimization opportunities before they become problems and understand the application's performance profile at a glance.

Full Changelog: v0.7.0...v0.7.1

Full Changelog: v0.7.0...v0.7.1