fix: normalize INPUT_JOB_NAME hyphen variant so OTLP spans include the actual job name

[Copilot]

OTLP spans were emitting gh-aw.job.setup / gh-aw.job.conclusion instead of gh-aw.<jobname>.setup / gh-aw.<jobname>.conclusion because some runner versions set INPUT_JOB-NAME (hyphen) instead of the standard INPUT_JOB_NAME (underscore) — the same runner quirk already handled for INPUT_TRACE_ID.

Changes

• index.js — normalize INPUT_JOB-NAME → INPUT_JOB_NAME at startup; pass canonical value into setup.sh env and set process.env.INPUT_JOB_NAME before delegating to action_setup_otlp.cjs
• action_setup_otlp.cjs — apply the same two-key lookup as INPUT_TRACE_ID; canonicalize to underscore form in process.env so sendJobSetupSpan (which reads process.env.INPUT_JOB_NAME directly) always finds it
• action_conclusion_otlp.cjs — replace direct process.env.INPUT_JOB_NAME read with normalized lookup before constructing the span name
• setup.sh — explicitly forward INPUT_JOB_NAME to action_setup_otlp.cjs in script mode, alongside the existing INPUT_TRACE_ID

// Before — only underscore form checked
const spanName = process.env.INPUT_JOB_NAME
  ? `gh-aw.${process.env.INPUT_JOB_NAME}.conclusion`
  : "gh-aw.job.conclusion";

// After — hyphen fallback mirrors INPUT_TRACE_ID pattern
const jobName = (process.env.INPUT_JOB_NAME || process.env["INPUT_JOB-NAME"] || "").trim();
const spanName = jobName ? `gh-aw.${jobName}.conclusion` : "gh-aw.job.conclusion";

[Copilot]

Pull request overview

Fixes OTLP span naming so job lifecycle spans consistently include the real job name even when some runner versions expose the input env var as INPUT_JOB-NAME (hyphen) instead of INPUT_JOB_NAME (underscore).

Changes:

• Normalize INPUT_JOB[-_]NAME to the canonical underscore form in the JS entrypoint and setup-span sender.
• Use a normalized job-name lookup in the conclusion-span sender.
• Add tests to validate the hyphen-form job-name behavior for both setup and conclusion spans.

Show a summary per file

File	Description
actions/setup/setup.sh	Forwards INPUT_JOB_NAME into the script-mode action_setup_otlp.cjs invocation.
actions/setup/js/action_setup_otlp.cjs	Normalizes job name from INPUT_JOB-NAME/INPUT_JOB_NAME and sets canonical process.env.INPUT_JOB_NAME.
actions/setup/js/action_otlp.test.cjs	Adds a setup-span test covering the hyphen-form job-name env var.
actions/setup/js/action_conclusion_otlp.cjs	Uses normalized job-name lookup when constructing conclusion span names.
actions/setup/js/action_conclusion_otlp.test.cjs	Extends conclusion-span tests to cover hyphen-form env var and restores env properly.
actions/setup/index.js	Normalizes job-name env var at startup and passes canonical value into setup.sh and action_setup_otlp.cjs.

Copilot's findings

Tip

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

• Files reviewed: 6/6 changed files
• Comments generated: 3

[Copilot]

The fallback order uses process.env.INPUT_JOB_NAME || process.env["INPUT_JOB-NAME"] before trimming, so a whitespace-only INPUT_JOB_NAME will shadow a valid hyphen-form value and resolve to an empty string after .trim(). Consider trimming each candidate first and then selecting the first non-empty value; also delete/clear process.env.INPUT_JOB_NAME when the trimmed result is empty so downstream sendJobSetupSpan() doesn’t treat whitespace as a real job name.

Suggested change

	const inputJobName = (process.env.INPUT_JOB_NAME || process.env["INPUT_JOB-NAME"] || "").trim();
	if (inputJobName) {
	// Normalise to the canonical underscore form so sendJobSetupSpan (which
	// reads process.env.INPUT_JOB_NAME) always finds the value.
	process.env.INPUT_JOB_NAME = inputJobName;
	const inputJobNameUnderscore = (process.env.INPUT_JOB_NAME || "").trim();
	const inputJobNameHyphen = (process.env["INPUT_JOB-NAME"] || "").trim();
	const inputJobName = inputJobNameUnderscore || inputJobNameHyphen;
	if (inputJobName) {
	// Normalise to the canonical underscore form so sendJobSetupSpan (which
	// reads process.env.INPUT_JOB_NAME) always finds the value.
	process.env.INPUT_JOB_NAME = inputJobName;
	} else if (process.env.INPUT_JOB_NAME && !inputJobNameUnderscore) {
	// Clear whitespace-only underscore-form input so downstream consumers do
	// not treat it as a real job name.
	delete process.env.INPUT_JOB_NAME;

[Copilot]

The lookup (process.env.INPUT_JOB_NAME || process.env["INPUT_JOB-NAME"]) occurs before trimming, so a whitespace-only INPUT_JOB_NAME will prevent the hyphen-form value from being used and will fall back to the default span name after .trim(). To make the fallback robust, trim both candidates first and pick the first non-empty trimmed value.

Suggested change

	const jobName = (process.env.INPUT_JOB_NAME || process.env["INPUT_JOB-NAME"] || "").trim();
	const underscoreJobName = (process.env.INPUT_JOB_NAME || "").trim();
	const hyphenJobName = (process.env["INPUT_JOB-NAME"] || "").trim();
	const jobName = underscoreJobName || hyphenJobName;

[Copilot]

This test restores fetchSpy only on the success path; if an assertion throws, the spy can leak into later tests (this file doesn’t have a global vi.restoreAllMocks() in afterEach). Move fetchSpy.mockRestore() into the finally block (or add an afterEach that restores mocks) to keep the suite isolated.