fix: detect git branch on detached-HEAD CI checkouts (SDK-7009) by harshit-browserstack · Pull Request #63 · browserstack/wdio-browserstack-service

harshit-browserstack · 2026-07-20T09:30:01Z

What is this about?

Fixes new CI builds reporting no branch in their VCS info, causing them to drop out of branch-based dashboards and filters even when on master.

Root cause: CI checks out a commit SHA directly (git checkout <sha>), leaving a detached-HEAD repo (customer build runs in a container at git root /app). There, .git/HEAD holds a raw SHA instead of ref: refs/heads/<branch>, so git-repo-info returns branch: null. getGitMetaData() copied info.branch verbatim into version_control.branch with no fallback → payload carried branch: null. Local checkouts sit on a real branch, so they were unaffected — matching the ticket's two same-SHA snippets.

Fix: in getGitMetaData() (packages/browserstack-service/src/util.ts), when git-repo-info yields no branch, fall back to (1) the CI provider's branch env var — covering the CI systems getCiInfo() already detects, plus an explicit BROWSERSTACK_GIT_BRANCH override — then (2) a best-effort git for-each-ref --points-at HEAD backstop. Values are sanitized via normalizeBranchName(). Fully graceful; never throws.

Verified by running the real compiled getGitMetaData() (real git-repo-info@2.1.1, real fs) in a real detached-HEAD repo: before → branch: null (reproduced symptom), after → branch: "master". 3 unit tests added (getGitMetaData suite 5/5); lint + build (tsc) pass; no new regressions.

Related Jira task/s

SDK-7009

Release (mandatory for every PR — required for the `ready-for-review` label)

Version bump: (required — tick exactly one)

minor (backwards-compatible feature)
patch (bug fix or other small change)

Release notes type: (optional)

New Feature
Bug Fix
Other Improvement

Release notes (customer-facing): (optional but encouraged)

Fixed the git branch not being reported for builds run in CI (detached-HEAD checkouts), so builds again appear under branch-based filters and dashboards.

Release notes (internal): (required — engineer-facing; what actually changed / why)

getGitMetaData now falls back to CI branch env vars (+ BROWSERSTACK_GIT_BRANCH override) and a git for-each-ref --points-at HEAD backstop when git-repo-info returns no branch on a detached-HEAD checkout. Adds normalizeBranchName sanitizer. Changeset included (patch).

Checklist

Ready to review
Has it been tested locally?

PR Validations

Run Tests: Comment RUN_TESTS to trigger sanity tests.

CI systems check out a commit SHA directly (`git checkout <sha>`), leaving the repo in a detached-HEAD state. There, `.git/HEAD` holds a raw SHA rather than `ref: refs/heads/<branch>`, so git-repo-info returns no branch and `getGitMetaData` reported `version_control.branch = null`. Builds then dropped out of branch-based dashboards and filters even though they were on master. Add a branch fallback in `getGitMetaData`: when git-repo-info yields no branch, resolve it from the CI provider's branch env var (covering the CI systems getCiInfo already detects, plus an explicit BROWSERSTACK_GIT_BRANCH override), then from a best-effort `git for-each-ref --points-at HEAD` backstop. Values are sanitized via normalizeBranchName (strips refs/heads|origin/, rejects unsafe input). Fully graceful — never throws, degrades to the prior behaviour. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

github-actions · 2026-07-20T10:34:27Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

github-actions · 2026-07-20T10:35:29Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

harshit-browserstack · 2026-07-20T10:54:01Z

✅ Good to go

All prior findings are resolved. The one remaining item from an earlier round — CI_MARKER_KEYS in packages/browserstack-service/tests/util.test.ts potentially drifting from getCiInfo()'s real gating vars — was fixed:

CI_MARKER_KEYS is now exhaustive against every env var getCiInfo() gates on.
The two CI-dependent tests (ignores a generic BRANCH env var when NOT in a recognised CI, uses a generic GIT_BRANCH var only inside a recognised CI) now assert utils.getCiInfo() is null/non-null before the branch assertion — any future drift between the list and getCiInfo() fails loudly instead of silently passing.

Update on current head (c4891798): the only change since the prior approved head (c119ae03) is in packages/browserstack-service/src/util.ts — gitForEachRef's execFile import moved from module-top to a lazy const { execFile } = await import('node:child_process') inside the function, fixing a CI test-isolation regression around module-load timing. Production branch-resolution behavior is byte-identical to the previously approved head. CI is green (Build & test node 18.20/20/22, Lint).

No blocking findings remain.

github-actions · 2026-07-20T10:54:10Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent reported blocking findings (🔴) on the current head commit — fix them and re-run the agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

…tests - BROWSERSTACK_GIT_BRANCH now overrides even a git-repo-info-resolved branch (moved ahead of info.branch), matching its documented "override" role. - GitLab: prefer CI_COMMIT_BRANCH over CI_COMMIT_REF_NAME (the latter is the tag name on tag pipelines). - Split CI branch env vars into provider-specific (trusted directly) vs generic (BRANCH/GIT_BRANCH/…, only trusted when getCiInfo() detects a CI) to stop a stray local env var being misreported as the branch. - Add CodeFresh CF_BRANCH (parity with getCiInfo detection). - git backstop: query refs/heads (lstrip=2) then refs/remotes (lstrip=3) as two ordered calls, yielding bare branch names for ANY remote (not just origin) and making heads-before-remotes explicit rather than relying on refspec order. - normalizeBranchName: denylist unsafe/illegal ref chars instead of an allowlist, so legal names (e.g. containing '#') are no longer dropped; kept internal (un-exported) so this stays a patch. - Tests: cover the git for-each-ref backstop (mocked spawnSync) and the null-result path; assert override precedence; all env mutations via a save/restore helper (try/finally) to prevent cross-test leakage. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

github-actions · 2026-07-20T11:03:12Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

github-actions · 2026-07-20T11:24:46Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent reported blocking findings (🔴) on the current head commit — fix them and re-run the agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

…op, tests - getBranchFromGit no longer passes git-repo-info's `.root` (which can point at the main repo dir under a git worktree); the backstop runs in process.cwd(), the actual checkout being instrumented. - Convert the git for-each-ref backstop from synchronous spawnSync to async promisified execFile, and make it lazy (query refs/heads, return early, only then refs/remotes) with a 3s timeout — no blocking subprocess on getGitMetaData's async hot path. - Tests: add coverage for the generic-var anti-collision gate (BRANCH ignored outside a recognised CI; GIT_BRANCH honoured inside Jenkins), CI-specific vs generic priority, and the remotes fall-through; env helper now also neutralises CI-detection markers so getCiInfo() is deterministic under ambient CI. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

github-actions · 2026-07-20T11:32:55Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

github-actions · 2026-07-20T11:46:00Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent reported blocking findings (🔴) on the current head commit — fix them and re-run the agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

Round-3 review: the test's CI_MARKER_KEYS list (used to force getCiInfo() null in the generic-var-gating tests) didn't cover every branch the real getCiInfo() checks (Concourse, Google Cloud, Shippable, some CodeBuild/Wercker vars), so a stray ambient CI marker could silently flip those tests to the wrong branch. - Complete CI_MARKER_KEYS against every gating var in getCiInfo(). - Add self-checking guards — assert utils.getCiInfo() is null (outside-CI test) / non-null (Jenkins test) before the branch assertion — so any future drift between the list and the real detector fails loudly instead of silently. Test-only change; production code unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

github-actions · 2026-07-20T11:49:32Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

github-actions · 2026-07-20T11:55:11Z

🟢 SDK PR Review gate is green — the SDK PR Review Agent has given a GTG for this PR (the ready-for-review label is present and the latest SDK PR Review Agent run reports success on the current head commit).

A native GitHub reviewer approval is still separately required by branch protection before this PR can merge — this check does not substitute for that.

github-actions · 2026-07-20T11:57:34Z

🟢 SDK PR Review gate is green — the SDK PR Review Agent has given a GTG for this PR (the ready-for-review label is present and the latest SDK PR Review Agent run reports success on the current head commit).

A native GitHub reviewer approval is still separately required by branch protection before this PR can merge — this check does not substitute for that.

…d&test) The module-top `import { execFile }` + `promisify(execFile)` broke every test that partial-mocks `node:child_process` (e.g. tests/cli/frameworks/index.test.ts mocks it as `{ spawn }` only) — `execFile` was undefined at util.ts load, so those suites failed to load with "No execFile export is defined on the mock". Import child_process lazily inside gitForEachRef (within its existing try/catch) so module load never references execFile. It's only touched when the git backstop actually runs (covered by util.test.ts's full mock); suites that never hit the backstop are unaffected, and a missing/partial mock degrades to []. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

github-actions · 2026-07-20T15:31:47Z

🔴 SDK PR Review gate is red. Pending:

The SDK PR Review Agent has not reviewed the current head commit yet — run the SDK PR Review Agent.

It turns green once the latest SDK PR Review Agent run reports GTG on the current head commit. A native reviewer approval is separately required by branch protection before merge.

github-actions · 2026-07-20T15:39:06Z

🟢 SDK PR Review gate is green — the SDK PR Review Agent has given a GTG for this PR (the ready-for-review label is present and the latest SDK PR Review Agent run reports success on the current head commit).

A native GitHub reviewer approval is still separately required by branch protection before this PR can merge — this check does not substitute for that.

github-actions · 2026-07-20T15:39:53Z

🟢 SDK PR Review gate is green — the SDK PR Review Agent has given a GTG for this PR (the ready-for-review label is present and the latest SDK PR Review Agent run reports success on the current head commit).

A native GitHub reviewer approval is still separately required by branch protection before this PR can merge — this check does not substitute for that.

harshit-browserstack requested a review from a team as a code owner July 20, 2026 09:30

harshit-browserstack requested review from AakashHotchandani and osho-20 July 20, 2026 09:30

harshit-browserstack added the ready-for-review label Jul 20, 2026

Merge branch 'main' into fix/sdk-7009-ci-branch-detection

6fa51e6

harshit-browserstack added ready-for-review and removed ready-for-review labels Jul 20, 2026

Uh oh!

Conversation

harshit-browserstack commented Jul 20, 2026 • edited by atlassian Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What is this about?

Related Jira task/s

Release (mandatory for every PR — required for the ready-for-review label)

Checklist

PR Validations

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

harshit-browserstack commented Jul 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

github-actions Bot commented Jul 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

harshit-browserstack commented Jul 20, 2026 •

edited by atlassian Bot

Loading

Release (mandatory for every PR — required for the `ready-for-review` label)

harshit-browserstack commented Jul 20, 2026 •

edited

Loading