feat(capabilities): inject container-id/entity-id/external-env headers

[bengl]

Native (wasm) trace export never sent Datadog-Container-ID / Datadog-Entity-ID / Datadog-External-Env: libdatadog's entity detection (libdd-common entity_id) is gated #[cfg(unix)] and inert on wasm32, and DD_EXTERNAL_ENV is unreachable from wasm. System-tests expect these headers.

What

Detect the entity headers in the Node HTTP transport (crates/capabilities/src/http_transport.js) — Node can read /proc and process.env — and rewrite the Rust-rendered HTTP/1.1 request head to carry them. This is the wasm analog of the headers native libdatadog adds via Endpoint::set_standard_headers.

• detectEntityHeaders() mirrors dd-trace-js's exporters/common/docker.js (the proven legacy-exporter path) and libdd-common's compute_entity_id (ci-<container_id> else in-<cgroup_inode>); external-env from DD_EXTERNAL_ENV. Cached (values are process-stable).
• applyEntityHeaders() rewrites the pre-rendered head: it drops any existing line for a name it sets — so libdatadog's empty datadog-container-id is replaced, not duplicated — appends the detected headers, and preserves the request line + framing (Host, Content-Length) byte-for-byte (latin1 round-trip). Returns the head untouched when nothing is detected or the head is malformed.
• DD_EXTERNAL_ENV is rejected if it contains CR/LF or non-ASCII, to avoid header injection (native libdatadog rejects invalid bytes via HeaderValue).

Why here (not Rust)

The head is rendered in Rust but wasm cannot read /proc/env; the Node transport is the only layer with both the rendered head and OS access. No upstream-libdatadog or dd-trace-js change is required — consumers pick this up on the next release.

Tests

test/http_transport.js (pre-existing observer + UDS tests preserved): detection (docker id, in-<inode> fallback, external-env, CR/LF rejection, empty case), head rewrite (append + framing preservation, empty-container-id dedup, no-op, malformed passthrough), and a real-transport E2E asserting datadog-external-env lands on the wire through the rewritten head.

• node --test test/http_transport.js → 16 passing
• node --test test/pipeline.js (rebuilt wasm) → 49 passing
• eslint clean

[github-actions]

Overall package size

Self size: 30.08 MB
Deduped: 30.08 MB
No deduping: 30.08 MB

Dependency sizes

| name | version | self size | total size | |------|---------|-----------|------------|

_{🤖 This report was automatically generated by heaviest-objects-in-the-universe}

[Copilot]

Pull request overview

Adds WASM-compatible injection of Datadog entity headers by detecting container/entity/external-env values in the Node HTTP transport layer (which can read /proc and process.env) and rewriting the Rust-rendered HTTP/1.1 request head to include them, with accompanying unit and E2E tests to satisfy system-test expectations.

Changes:

• Implement detectEntityHeaders() (container-id/entity-id via cgroup parsing; external-env via DD_EXTERNAL_ENV with CR/LF + non-ASCII rejection) plus memoized caching.
• Implement applyEntityHeaders() to drop/replace any existing lines for injected header names and append detected headers while keeping the rest of the head intact.
• Add tests covering detection, head rewriting behavior (including dedup of empty container-id), and a real-transport E2E asserting headers appear on the wire.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
crates/capabilities/src/http_transport.js	Adds entity header detection + caching and rewrites the Rust-rendered request head to inject/replace these headers during httpRequest.
test/http_transport.js	Adds unit tests for detection and head rewriting, plus an end-to-end transport test validating header injection on the wire.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rochdev]

This should be an event that is handled by the binary instead of altering the headers from the outside. It's fine for now to unblock, but this should be addressed as a follow-up, and ideally a TODO should be added.