Skip to content

[v0.17 proof] Measure Windows linker time from an explicit boundary #1635

Description

@TheGreenCedar

Context

Windows package timing in the 0.16.3 release train reported linker evidence from a textual match that also accepted the Rust build line Compiling time v0.3.47. That is the dependency named time, not a linker duration.

The package graph, archive identity, and Windows Vulkan runtime proof all passed. The defect is telemetry-only and was explicitly accepted as a post-release follow-up.

Acceptance

  • Measure linker wall time from an explicit timed boundary or structured build trace, not an unscoped substring search.
  • Keep cache restoration, Rust compilation, native setup, linking, and packaging as distinct reported intervals.
  • A fixture containing only Compiling time v0.3.47 must not satisfy linker-timing evidence.
  • A captured real linker invocation/duration must satisfy the selector.
  • Timing remains observational; missing timing must not invalidate an otherwise authenticated package.
  • Use focused workflow-policy tests and the required hostile mutation. Do not dispatch broad proof from the support PR.

Refs #1611
Refs #1179

v0.17 issue-map role — 2026-08-01\n\nThis issue remains the exact-candidate linker-timing evidence tracker. The workflow/harness source repair is isolated in #1709; its implementation PR closes #1709 only. Do not use #1635 itself as a stack layer.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workingepic:performance-opsPerformance, scale, sidecar SLA, and artifact operationsphase:next-waveActive next implementation wavepriority:p2Important follow-up for next release wavesaga:codestory-intelligenceCodeStory product-grade intelligence sagatype:harnessBenchmark or verification harness work

    Projects

    Status
    Todo

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions