test(rust): enable passing ot.th/ot.rv tracestate sampling scenarios [rust@milan.garnier/otel-consistent-sampling]

[MilanGarnier]

Motivation

Stacked on #7372 (the ot.th/ot.rv conformance scaffolding). Enables the Rust entries against dd-trace-rs's wire implementation (branch milan.garnier/otel-tracestate-ot-th), verified locally against a rust@0.5.1-dev+c36937d build (axum weblog) across every scenario (DEFAULT + all five sample-rate scenarios).

What does this PR do?

Flips all tests/test_otel_tracestate_sampling.py Rust entries from missing_feature to >=0.5.0-dev in manifests/rust.yml, and backfills 3 previously-unlisted classes (Test_ForwardInboundOtUnchangedWhenDropped, Test_ThOnlyDoesNotFabricateRv, Test_ThOnlyDoesNotFabricateRvWhenDropped) as missing_feature across the other 8 language manifests to preserve the "disabled for every tracer until it ships" invariant. OTLP entries (test_tracing_otlp.py, APMAPI-2172) are untouched per scope.

Local verification results (ai generated draft not reviewed)

• DEFAULT scenario: 6/9 selected tests pass. The 3 Test_ForceKeepClearsTh (A4) sub-tests fail, but not because of an ot.th gap: the local weblog used for verification (a not-yet-merged Rust Axum weblog) has no AppSec/WAF integration, so the Arachni-user-agent-triggered force-keep this test relies on never fires. The underlying "clear th, forward inherited rv" wire logic it exercises is already unit-tested and passing (tracecontext.rs::force_keep_clears_th_forwards_rv).
• Rate scenarios: only SAMPLING (rate 0.5) passes outright. OTEL_SAMPLING_RATE_0_01/0_1/0_2/0_99 all fail on a th mismatch of a few ULPs against TH_BY_RATE — two independent rounding bugs, not one:
  • TH_BY_RATE[0.1] in the test fixture itself is stale (e6666666666668); the exact-fraction value is e6666666666666, which matches both what Rust emits and dd-trace-rs's own corrected example.
  • For rate 0.01, Rust emits one ULP above the exact-fraction floor (fd70a3d70a3d71 vs fd70a3d70a3d70) — a rounding discrepancy in the unreleased libdd-sampling::OtelConsistentSampling threshold computation dd-trace-rs depends on, not in dd-trace-rs itself.

Both issues need a fix (one in this repo's fixture, one upstream in libdatadog) before these entries are actually green. Opening as draft now since the manifest/scope work itself is done and the base branch is expected to move; will flip out of draft once the rounding fixes land and a clean local rerun confirms all scenarios pass.

Base

Targets ben.db/feat-otel-tracestate-sampling-scaffolding (stacked). Retarget to main once #7372 merges.

[github-actions]

CODEOWNERS have been resolved as:

manifests/rust.yml                                                      @DataDog/apm-rust

[datadog-official]

 

⚠️ Warnings

🚦 65 Pipeline jobs failed

DataDog/system-tests | Amazon_Linux_2023_amd64.CON: [test-app-ruby-container]     

DataDog/system-tests | Amazon_Linux_2_arm64.INS: [test-app-ruby-container]     

DataDog/system-tests | Debian_11_amd64.INS: [test-app-ruby-container]     

View all 65 failed jobs.

ℹ️ Info

No other issues found (see more)

🧪 All tests passed
❄️ No new flaky tests detected

🔄 Datadog auto-retried 101 jobs - 67 passed on retry

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: af225e6 | Docs | Datadog PR Page | Give us feedback!}