Scope PR test runs to affected apps and cover .github in fullBuildPatterns

[Alexander Holstrup (aholstrup1)]

What & why

BCApps currently runs the full test suite (~316 test apps) on every PR, even when a change touches a single app. The build-optimization logic to scope test runs to only the affected apps already existed in BuildOptimization.psm1 (with unit tests), but nothing in the pipeline called it. This wires it in and fixes a related gap in fullBuildPatterns.

Two changes:

• Wire up per-app test optimization. Invoke-PerProjectTestRun now filters the per-project test-app list through Test-ShouldSkipTestApp, so a PR only runs tests for apps affected by its changed files plus their dependents. Filtering happens before dispatch, so skipped apps never occupy a tenant. This is correctness-preserving: a full run is still forced for fullBuildPatterns matches, unmapped files under src/, non-CI runs, and workflow_dispatch, so the failure mode is always "run too much," never "skip something it shouldn't."

• Cover .github/* in fullBuildPatterns. Previously only two specific workflow files were listed. A change to .github/AL-Go-Settings.json (which defines artifact version, analyzers, target country, and the patterns themselves) mapped to no app, produced an empty affected set, and would have skipped every test. Broadening to .github/* closes that hole and subsumes the two explicit workflow entries.

On a sample of 10 recent open PRs, 6 dropped from 316 test apps to 4 or fewer; the rest correctly stayed full (base-layer or infra changes).

Linked work

Fixes #

How I validated this

• I read the full diff and it contains only changes I intended.
• I built the affected app(s) locally with no new analyzer warnings.
• I ran the change in Business Central and confirmed it behaves as expected.
• I added or updated tests for the new behavior, or explained below why none are needed.

What I tested and the outcome

• Ran the existing BuildOptimization.Test.ps1 Pester suite: 36/37 pass. The one failure (Get-AffectedApps ... Expected 51, but got 56) is a pre-existing hardcoded-count assertion that also fails on main with these changes stashed; it is dependency-graph drift from apps added since the count was written, unrelated to this change.
• Verified .github/AL-Go-Settings.json still parses and ParallelTestExecution.psm1 imports cleanly.
• Replayed the affected-app computation against 10 open PRs to confirm scoping behaves as expected (single-app PRs scope down to 1-4 test apps; base-layer/infra PRs force a full run via fullBuildPatterns).
• No new behavior-specific tests added: the filtering reuses the already-tested Test-ShouldSkipTestApp, and this change is wiring plus a settings entry.

Risk & compatibility

• The optimization only narrows test runs on PR builds; it is a no-op outside CI and whenever a full run is required, so it cannot cause fewer tests to run than intended in those cases.
• Scoping accuracy depends on app.json dependency declarations being correct (already how build order is resolved). A product app with no linked test app resolves to zero affected test apps; that is correct when the app genuinely has no tests, but a follow-up guard could warn when a changed product app has no associated test app.
• Broadening to .github/* means build-irrelevant edits (e.g. PULL_REQUEST_TEMPLATE.md, CODEOWNERS) now force a full run. This is intentional and consistent with the conservative, correctness-first intent of the list.
• Pre-existing stale test count (51 vs 56) is worth fixing separately so the suite goes green.

[github-actions]

Could not find a linked ADO work item. Please link one by using the pattern 'AB#' followed by the relevant work item number. You may use the 'Fixes' keyword to automatically resolve the work item when the pull request is merged. E.g. 'Fixes AB#1234'