Skip to content

feat(observed): add Windows performance counter backend - #817

Open
Ralf Biedert (ralfbiedert) wants to merge 40 commits into
mainfrom
ralfbiedert-observed-windows-counters
Open

Ralf Biedert (ralfbiedert) wants to merge 40 commits into
mainfrom
ralfbiedert-observed-windows-counters

Conversation

@ralfbiedert

@ralfbiedert Ralf Biedert (ralfbiedert) commented Oct 13, 2026 •

Copy link
Copy Markdown
Collaborator

Publish typed observed integer metrics to Windows performance-counter consumers, including PerfSpy deployments, using startup-prepared numeric targets and the existing lazy field visitor.

Event metadata and binding

  • Metadata is inspection-only. Public constructors/setters on event, field, log, metric, routing-entry and source-location metadata are removed; MetricDescriptionBuilder is deleted. No replacement construction API is exposed under interop. This intentionally changes the previous metadata construction API.
  • Macros construct descriptions through hidden helpers; one const helper assigns stable field indices. Generated visitors reuse those descriptions. An absent optional field does not shift later indices. Enrichment has routing information but no static event-field position.
  • EventDescription::fields() supplies the static routed-field inventory. Separate Event::METRICS and metric_value accessors are removed. Fieldless counters remain in EventDescription::metric() and add one per emission.
  • MetricDescription::value_kind() returns required MetricValueKind, not an option. The separate public MetricValue enum is removed: bindings deliver the existing Value, normalizing unsigned measurements without losing integer precision. Invalid values fail explicitly.
  • MetricBindings::bind_event::<E>() prepares targets at startup. Recording selects type identities and field indices, enumerating routed fields but evaluating only bound numeric getters. There is no string matching, hashing, allocation, redaction, or enrichment extraction during successful numeric recording.
  • Backend errors retain typed causes privately, without public conversions from binding/read errors.

Windows definitions and behavior

observed_perf_counters imports an externally authored V2 .man through manifest_file(path)?. category(name) selects a declared set; build() validates exact instrument-name mappings and starts the provider on 64-bit Windows. Supplied provider/set GUIDs, counter IDs and definitions remain authoritative. Unbound declared counters remain zero. There is no XML export, identity generation, or catalog installation at startup.

The supported profile remains a user-mode provider with a custom callback, multiple-instance sets, and inline-named, standard-detail, zero-scale perf_counter_large_rawcount reference counters. The public backend surface remains Error, PerfCounterProcessor and PerfCounterProcessorBuilder.

Unsigned counters add, signed up/down counters adjust a non-negative total, and unsigned gauges replace readings. Overflow/underflow preserve the prior value and surface through tracing and Sink::flush(). Histograms, floats and signed gauges are rejected. Normal sink timestamp/context/sampling behavior remains; use an unsampled sink for deltas. Enrichment opt-out is deferred.

Deployment compiles and installs the same supplied manifest and resource DLL separately; PerfSpy also needs its collection service configured. This imports V2 definitions, not classic .NET custom counter categories.

Direct atomic publication

Commit 15aedd8a replaces the callback snapshot/admission scheme. Perflib references now point directly at the existing, aligned AtomicU64 totals. The ordinary UnsafeCell<u64> publication buffer, snapshot copying, reader gate and sticky callback errors are removed. The custom callback does no work: delayed or overlapping START/END notifications cannot resume a publication copy. Emission retains its existing checked atomic updates, with no additional native calls or operations. Manifest formats are unchanged; no native setter loop is introduced.

The documented FFI argument combines:

This is a Windows-specific scalar-access contract, not a claim that ordinary Rust raw-pointer reads are atomic or that arbitrary mixed-size/native writes are supported. No side payload relies on counter memory ordering. Individual readings are atomic; multiple counters are not a transactional snapshot. .NET's V2 CounterData implementation supplies supporting reference-publication precedent, not a replacement for that contract.

The design guide documents the argument and validation boundaries.

Integration and validation

Prerequisites #821 and #822 are merged. Commit 6de372c2 preserves upstream sampling/sink-builder changes and migrates counter tests, examples and benchmark setup to Sink::builder.

Commit 7ffeb04e merges current main and repairs CI integration with the newly upstreamed sink-interest API. GitHub tested a merge ref containing sink_interest.rs, which was absent from the earlier local head; its tests and doctest still used the removed metadata constructor. They now use macro-generated event descriptions. The same compilation error was verified in Anvil and all four external Linux/Windows dev/release builds; no pipeline checks, assertions or exclusions were weakened.

  • Commit ddd0d444 adds a private injectable manifest opener/reader. Load-once, file-error, malformed-input and byte-budget tests use controlled in-memory inputs without fixture writes or native startup. Public file-loading smoke coverage reads the checked-in manifest unchanged. Unsupported numeric tests assert exact diagnostics so missing manifest names cannot mask a regression. The unused backend tempfile dev-dependency is removed; no public construction API is added.
  • Post-merge all-feature/no-default affected tests, doctests, zero-allocation assertions, full Anvil Clippy, debug/release builds, formatting, generated README and spelling checks pass locally. The formerly failing sink-interest target passes all seven tests, including under Miri. The mutation runner's unmutated compilation also succeeds for observed, observed_macros_impl and observed_perf_counters.
  • The latest merged coverage run passes 470 tests and the unchanged 100% gates for observed, observed_macros_impl and observed_perf_counters. Latest manifest/binding mutations: 7 caught, 1 unviable. Direct-publication native registration/ownership mutations: 11 caught, 1 unviable. No coverage thresholds or mutation exclusions were weakened.
  • Miri additionally passes injected load-once, open/read-error and XML/UTF-8 cases, the 11 native ownership/reference cases, native-ID mapping and concurrent typed emission. The delayed-START regression lets collection read updated values and issue END before START resumes, using controlled coordination rather than sleeps. Mocks model full-width native atomic reads; Miri does not execute Perflib or establish its platform contract.
  • Earlier static-metadata/macro/manifest verification remains documented in commit history. The pinned public external-types checker reported zero errors for the direct-publication API. New CI must validate the pushed head.
  • API docs are rebuilt and opened in the system browser. READMEs are regenerated, not manually edited.
  • Compiling/installing the supplied .man with Windows SDK tools and running the installed-schema consumer scenario remain deployment validation tasks on a suitably provisioned host. The opt-in consumer test is not claimed to have run here. An earlier optional Ubuntu rerun was blocked before compilation by an unavailable kache wrapper; no cache bypass was applied.

Add whole-event numeric metric bindings and native Perflib V2 publishing for integer counters, up/down deltas, and gauges. Generate matching deployment manifests and document Windows counter registration. Use performables for private shared state while preserving existing public Arc interfaces.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 13, 2026 15:27

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The substantial unsafe Windows FFI and callback synchronization require human validation, particularly with the installed-schema consumer test remaining opt-in.

0 open findings

What changed in this PR

Adds typed metric bindings and a native 64-bit Windows Perflib V2 backend for observed.

Changes:

  • Generates typed metric metadata and allocation-free numeric slot access.
  • Adds Windows counter publishing, manifests, validation, tests, and examples.
  • Migrates private sink/enrichment ownership to performables::Arc.
File Description
.spelling Adds Windows counter terminology.
Cargo.toml Registers the backend and dependencies.
Cargo.lock Locks new backend dependencies.
crates/​performables/​src/​arc/​mod.rs Supports unsized standard Arc wrapping.
crates/​observed/​Cargo.toml Registers the counter benchmark.
crates/​observed/​README.md Documents whole-event bindings.
crates/​observed/​benches/​observed_counters.rs Benchmarks bound counter emission.
crates/​observed/​src/​enrichment/​enrich_ext.rs Migrates enrichment ownership.
crates/​observed/​src/​enrichment/​entry.rs Migrates stored values to performables.
crates/​observed/​src/​enrichment/​slot.rs Migrates enrichment slots.
crates/​observed/​src/​event.rs Adds metric metadata and slot access.
crates/​observed/​src/​interop.rs Adds dynamic numeric slot forwarding.
crates/​observed/​src/​lib.rs Documents typed metric bindings.
crates/​observed/​src/​metadata/​metric.rs Defines numeric metric representations.
crates/​observed/​src/​metadata/​mod.rs Exports metric value types.
crates/​observed/​src/​processing/​event_state.rs Forwards metric values during dispatch.
crates/​observed/​src/​processing/​event_view.rs Exposes numeric measurements.
crates/​observed/​src/​processing/​metric_bindings.rs Implements typed event bindings.
crates/​observed/​src/​processing/​mod.rs Exports binding APIs.
crates/​observed/​src/​sink/​core.rs Migrates private sink ownership.
crates/​observed/​tests/​counter_bindings.rs Tests binding and delivery behavior.
crates/​observed_perf_counters/​Cargo.toml Defines the new backend crate.
crates/​observed_perf_counters/​README.md Documents backend deployment and use.
crates/​observed_perf_counters/​examples/​consumer_manifest.rs Generates consumer-test manifests.
crates/​observed_perf_counters/​examples/​publish.rs Demonstrates counter publishing.
crates/​observed_perf_counters/​src/​error.rs Defines backend errors.
crates/​observed_perf_counters/​src/​lib.rs Exposes and documents the backend.
crates/​observed_perf_counters/​src/​native/​consumer_tests.rs Tests native consumer integration.
crates/​observed_perf_counters/​src/​native.rs Implements Perflib publishing.
crates/​observed_perf_counters/​src/​processor.rs Implements processor configuration and updates.
crates/​observed_perf_counters/​src/​schema.rs Builds stable schemas and manifests.
crates/​observed_perf_counters/​tests/​allocations.rs Verifies allocation-free emission.
crates/​observed_perf_counters/​tests/​configuration.rs Tests configuration validation.
crates/​observed_macros_impl/​src/​event.rs Generates metric metadata and accessors.
crates/​observed_macros_impl/​tests/​expansion.rs Adds macro expansion coverage.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__a_fieldless_counter_becomes_the_event_level_metric.snap Captures fieldless-counter generation.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__a_macro_substituted_type_is_seen_through_its_invisible_group.snap Captures macro-substituted numeric access.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__an_instrument_looks_through_parentheses_around_its_value_type.snap Captures parenthesized numeric types.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__an_instrument_without_a_name_is_named_after_the_event.snap Captures default instrument naming.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__every_instrument_kind_records_the_field_it_names.snap Captures all metric representations.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__whole_event_metadata_does_not_generate_conflicting_associated_keys.snap Covers conflicting metric names.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__whole_event_metadata_preserves_non_identifier_counter_names.snap Covers non-identifier names.
crates/​observed_macros_impl/​tests/​snapshots/​expansion__whole_event_metadata_preserves_raw_fields_and_instrument_names.snap Covers raw fields and names.

🧠 Review effort: Balanced


💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@codecov

codecov Bot commented Oct 13, 2026 •

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.0%. Comparing base (2be413c) to head (f218f85).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files
@@            Coverage Diff            @@
##            main     #817      +/-   ##
=========================================
  Coverage   99.9%   100.0%              
=========================================
  Files        808      241     -567     
  Lines      68565    15971   -52594     
=========================================
- Hits       68556    15971   -52585     
+ Misses         9        0       -9     
Flag Coverage Δ
linux 100.0% <100.0%> (ø)
linux-arm 100.0% <100.0%> (ø)
scheduled ?
windows 100.0% <100.0%> (+<0.1%) ⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Exercise numeric defaults, pointer-width code generation, counter identities, sticky errors, callback faults, native startup failures and teardown diagnostics. Add a private native API boundary so lifecycle and reference-cell tests run under Miri without Perflib syscalls. Keep native allocation and OS integration checks intact, run portable arithmetic tests on Linux, and limit Linux mutations to code actually compiled there. Preserve the 100 percent coverage gate and exclude only the equivalent default-constructor mutation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 13, 2026 20:50

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The substantial unsafe Windows FFI and callback synchronization require final human review, especially with the installed-schema consumer test remaining opt-in.

0 open findings

🧠 Review effort: Balanced

Preserve the counter backend's equivalent-mutation and Linux platform rules alongside the updated metabench rules from main. Retain the base branch's dependency and benchmark changes.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 13, 2026 20:59

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The substantial unsafe Perflib lifecycle and synchronization code warrants final human review, especially while the installed-schema consumer test remains opt-in.

0 open findings

🧠 Review effort: Balanced

Declare alloc_tracker only for 64-bit Windows, matching the native allocation test. Linux no longer receives an unused dev-dependency; Windows retains the allocation-free emission contract.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 14, 2026 01:50

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

Native callback contention and tracing while holding internal locks can permanently freeze collection or deadlock reentrant subscribers.

3 open findings

🧠 Review effort: Balanced

… tracing

Separate atomic collection admission from the provider registry so registry contention can reuse a frozen snapshot and matching collection-end notifications never need that lock. Return retryable busy only while a snapshot writer is active. Store sticky processor errors before tracing and release both ordinary and poisoned guards before invoking subscribers. Add contention recovery, concurrent reference-read, reentrant subscriber, and panic-retention regressions.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 14, 2026 02:14

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

Manifest validation permits generated counter help text to exceed its configured UTF-16 limit.

1 open finding
3 resolved since last review

🧠 Review effort: Balanced

Validate the final description after appending unit metadata, preserving the existing component checks and UTF-16 limit. Cover the reported overflow, exact-limit manifest output, and one-code-unit overflow for ASCII and supplementary Unicode units.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings October 14, 2026 02:23

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Native Windows FFI and callback synchronization warrant human review, especially with the installed-schema consumer test remaining opt-in.

0 open findings

1 resolved since last review

🧠 Review effort: Balanced

Comment thread crates/performables/src/arc/mod.rs Resolved
Comment thread crates/observed/src/metadata/metric.rs Resolved
Comment thread crates/observed/src/sink/core.rs Resolved
Extract the zero-copy standard Arc conversion for unsized slices and trait objects, with regression tests, as a prerequisite for #817.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: c45a6d66-dac9-4b55-beda-d5786efb3c2d

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

The regression tests accept generic configuration failures, so duplicate-destination and missing-field cases can pass for unrelated errors without validating their intended diagnostics.

2 open issues

Review effort: Balanced

Distinguish missing manifest, category, and instance diagnostics. Require the Bytes Sent duplicate-destination error rather than accepting any invalid configuration.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings November 2, 2026 19:37

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Exposing Rust AtomicU64 storage directly to Perflib still requires native-safety review of the Windows-specific read/shutdown contract and an installed-schema consumer run, neither established by mocks or Miri.

No open issues 🎉

2 resolved since last review

Review effort: Balanced

Update async-executor to the compatible non-yanked 1.14.1 release. Exercise the allocator head-lease handoff and Tokio startup abort paths without depending on thread scheduling. Keep the existing coverage and dependency gates unchanged.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings November 3, 2026 13:42

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Direct AtomicU64 publication to Perflib requires Windows maintainer review of native-read and shutdown guarantees, while SDK manifest compilation and installed-schema consumer validation remain unrun.

No open issues 🎉

Review effort: Balanced

Resolve the lockfile conflict by keeping main's removal of obsolete executor dependencies while preserving counter-backend dependencies.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings November 6, 2026 08:45

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

The configuration rejection tests only check the supplied name, so they could pass for an unintended failure instead of verifying the exact expected diagnostics.

1 open issue

Review effort: Balanced

Assert complete primary errors for unknown counter sets and unmapped instruments so another configuration failure cannot satisfy these tests.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7
Copilot AI balanced review requested due to automatic review settings November 6, 2026 08:57

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

AtomicU64 publication and shutdown safety depend on a Windows-specific FFI contract requiring maintainer review, while SDK manifest compilation and installed-consumer validation remain unverified.

No open issues 🎉

1 resolved since last review

Review effort: Balanced

Keep the new workspace release versions and exact observed macro pins, preserve the counter backend entry, and regenerate the observed README from its documentation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The unsafe Perflib boundary relies on Windows atomic-read and shutdown guarantees that mocks cannot establish, requiring platform-owner review and the still-unrun installed-schema consumer validation.

No open issues 🎉

Review effort: Balanced

Keep the Gamma migration and allocator handoff regression from main. Adapt the incoming dynamic-event test to the private metadata constructors and disambiguate the native backend dependency in the merged lockfile.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Direct Perflib access to Rust AtomicU64 storage requires human review of the native FFI safety contract and installed-manifest consumer validation on Windows, neither of which mocks or Miri establish.

No open issues 🎉

Review effort: Balanced

@github-advanced-security github-advanced-security AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cargo-gamma found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.

Adopt the exact upstream executor test fix from c27b005 as approved by the operator. Tighten disabled-event, nonzero-slot, native lifecycle, startup, and XML diagnostic assertions. Document only observed equivalent mutations, invariant-only error paths, and native-boundary or timeout limitations beside their exact source sites.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The unsafe Perflib backend requires a Windows maintainer to assess native atomic-access and shutdown-safety guarantees and validate installed-manifest consumer readings on a provisioned Windows host.

No open issues 🎉

Review effort: Balanced

@github-actions

github-actions Bot commented Nov 13, 2026 •

Copy link
Copy Markdown

⚠️ SemVer check advisory

Potential breaking changes

cargo semver-checks flagged the following on this PR. This is informational -- breaking changes between commits are expected; the major-version bump happens at release time, not on every PR.

observed

    Checking observed v0.30.1 -> v0.30.1 (no change; assume minor)
     Checked [   0.019s] 202 checks: 201 pass, 1 fail, 0 warn, 58 skip

--- failure inherent_method_missing: pub method removed or renamed ---

Description:
A publicly-visible method or associated fn is no longer available under its prior name. It may have been renamed or removed entirely.
        ref: https://doc.rust-lang.org/cargo/reference/semver.html#item-remove
       impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.51.0/src/lints/inherent_method_missing.ron

Failed in:
  LogDescription::new, previously in file crates/observed/src/metadata/log.rs:24
  SourceLocation::new, previously in file crates/observed/src/metadata/source.rs:19
  MetricFieldEntry::dimension, previously in file crates/observed/src/metadata/field.rs:51
  MetricFieldEntry::instrument, previously in file crates/observed/src/metadata/field.rs:57
  FieldDescriptor::new, previously in file crates/observed/src/metadata/field.rs:102
  FieldDescriptor::log_only, previously in file crates/observed/src/metadata/field.rs:108
  MetricDescription::new, previously in file crates/observed/src/metadata/metric.rs:58
  LogFieldEntry::new, previously in file crates/observed/src/metadata/field.rs:17
  EventDescription::new, previously in file crates/observed/src/metadata/event.rs:32

     Summary semver requires new major version: 1 major and 0 minor checks failed
    Finished [   0.047s] observed

Keep the upstream testable Tokio startup coordinator and preserve this branch's existing startup regression checks. Adapt the incoming observed oracle fixtures to the hidden metadata constructors without changing their assertions.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 740c130d-6b52-40ad-92aa-da53562338f7

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

The unsafe Perflib backend needs Windows/Perflib maintainer review because its native atomic-read contract is platform-specific and installed-manifest consumer validation remains unrun.

No open issues 🎉

Review effort: Balanced

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants