Stop reporting success when the SDK is exporting nothing — TS, Go, Python, .NET

[brentrager]

Problem

The Rust SDK told a production service it was fine while it emitted nothing for months (#84). The other four SDKs all had the same bug. bootstrap() reported installed: true whenever it was not explicitly disabled — including when no OTLP endpoint was configured, so telemetry had nowhere to go. The no-endpoint branch was completely silent: it warned about missing AUTH but not about the far more consequential missing DESTINATION.

Per-language verdict

I verified each language independently rather than assuming the bug ported.

SDK	Verdict	Evidence
TypeScript	🔴 Bug present — worse than Rust	bootstrapped = { installed: true, otel, stopRefresh } set unconditionally; no endpoint check anywhere. Worse: setupOtelSdk falls back to new OTLPTraceExporter({ headers }) with no url, which OTel defaults to http://localhost:4318 — so a container doesn't no-op, it retries into the void forever.
Go	🔴 Bug present	result = BootstrapResult{Installed: true, Otel: otelHandle} regardless of endpoint. SetupOtelSDK correctly skips exporters when the endpoint is empty, leaving every provider nil — so the handle already knew the truth and nothing asked it.
Python	🔴 Bug present — worse than Rust	BootstrapResult(installed=True, ...) unconditionally. Same localhost fallback as TS (OTLPSpanExporter(headers=...) with no endpoint). The bug had already left a note about itself: python/tests/conftest.py exists to silence the resulting "connection refused to localhost:4318" spam.
.NET	🔴 Bug present	return Cache(new BootstrapResult { Installed = true, Otel = otel }) unconditionally. Like Go, ObservabilitySdk.Build gates each exporter on a non-empty endpoint, so providers are null and nobody checked.
Rust	✅ Already fixed	#84, merged as d40f623 — this PR mirrors its semantics and wording.

What changed in each

1. An honest status flag — exporting (TS, Python), Exporting (Go, .NET), matching Rust's exporting. installed/Installed keeps its old meaning and now says so honestly in its doc comment instead of implying more.

   Two deliberate per-language differences:

   • Go derives it from the handle, not the endpoint strings — an endpoint whose exporter failed to construct leaves every provider nil, and that is just as much "not exporting" as having no endpoint.
   • .NET counts traces + metrics only. Setup builds exporters for exactly those two; a LogsEndpoint alone is consumed by the ILoggingBuilder extension, so claiming Exporting on it would be the same lie in a new place.

2. A loud warning when no endpoint is configured, wording matched to Rust's so the five SDKs read alike. It names the variable to set AND offers SMOOAI_OBSERVABILITY_DISABLED=true, so an intentional no-op can be declared rather than inferred from absence.

3. Both halves tested in all four. With only one asserted, an implementation that hard-codes either value passes.

Three tests enshrined the misleading shape and now assert the honest one (same as the Rust one #84 fixed): Go's TestBootstrapInstallsClientAndCapture, Python's test_never_raises_on_bad_config, .NET's Run_NeverThrows_OnBadConfig — each asserted installed while nothing had a destination. Each now also pins the warning and clears the OTEL_EXPORTER_OTLP_* env vars, so "no endpoint" means no endpoint from any source rather than whatever the CI runner happens to export.

Mutation table

Every new assertion was mutation-checked: apply the mutation, confirm the specific test fails, restore, verify byte-identical via sha256 on raw bytes.

#	Mutation	Test that failed	Exit	Residue
TS-M1	exporting hard-coded true	reports exporting=false and warns loudly… — expected true to be false	1	none
TS-M2	exporting hard-coded false	reports exporting=true and stays quiet… — expected false to be true	1	none
TS-M3	no-endpoint warning removed	reports exporting=false… — expected '' to contain 'NO OTLP ENDPOINT CONFIGURED'	1	none
GO-M1	Exporting hard-coded true	TestBootstrapInstallsClientAndCapture	1	none
GO-M2	Exporting hard-coded false	TestBootstrapExportingWhenEndpointConfigured	1	none
GO-M3	no-endpoint warning removed	TestBootstrapInstallsClientAndCapture	1	none
PY-M1	exporting hard-coded True	test_never_raises_on_bad_config — assert True is False	1	none
PY-M2	exporting hard-coded False	test_exporting_true_when_endpoint_configured — assert False is True	1	none
PY-M3	no-endpoint warning removed	test_never_raises_on_bad_config — 'NO OTLP ENDPOINT CONFIGURED' not in stderr	1	none
NET-M1	Exporting hard-coded true	Run_NeverThrows_OnBadConfig — Assert.False() Failure	1	none
NET-M2	Exporting hard-coded false	Run_WithEndpoint_ReportsExporting — Assert.True() Failure	1	none
NET-M3	no-endpoint warning removed	Run_NeverThrows_OnBadConfig — Assert.Contains() Failure	1	none

12/12 mutations killed the intended test and only that test. Residue was verified with python3 over raw bytes (not grep) across all touched files: 0 hits.

Two pre-existing .NET CI failures (second commit)

Both reproduce on a stashed clean origin/main tree and are unrelated to this change — but the dotnet lane only runs when dotnet/** changes, so this is the PR that has to face them.

1. dotnet format --verify-no-changes exits 2 on CrashChild.cs — 14 WHITESPACE errors, a braced switch-case body indented one level short. Identical on the clean tree. Fixed by running dotnet format on that one file; pure indentation.

2. OtelSetupTests.Setup_IsIdempotent is a flake — 3 of 8 full-suite runs failed on the clean tree. ObservabilitySdk._installed is a process-wide static and three classes call ResetForTests() on it, but only BootstrapTests was in a collection. xUnit parallelizes across collections, so the other two ran concurrently and a foreign reset landed between that test's two Setup() calls — wiping the very install guard the test exists to assert. It hid because it passes 6/6 when the suite is filtered down.

   Fixed by putting all three classes in one non-parallel collection (OtelGlobalStateCollection, renamed from "Bootstrap" since it guards the OTel singleton, not bootstrap). 10/10 full-suite runs green afterwards.

Verification

Every gate run as CI runs it, judged by exit code captured immediately after:

TS   typecheck 0 · lint 0 · test 0 (263 passed) · build 0 · format:check 0   (oxfmt, not prettier)
Go   gofmt -l → empty · vet 0 (., fiber, gin) · test 0 (., fiber, gin)
Py   ruff check 0 · ruff format --check 0 · pytest 0 (78 passed)
.NET restore 0 · build 0 · test 0 (80 passed) · format --verify-no-changes 0

Changeset added (minor).

🤖 Generated with Claude Code

[changeset-bot]

🦋 Changeset detected

Latest commit: fafaba9

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@smooai/observability	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR