[pjcdawkins]

upsun lint rejected stack: [] for a composable image (Invalid type. Expected: object, given: array), although the platform accepts an empty stack, e.g. for static sites. The platform no longer accepts a non-empty list: stack must be a mapping of runtimes and packages.

• Allow arrays and null for stack in the Flex and Fixed schemas.
• Report a non-empty list from CheckTypes with a specific error ('stack' must be a mapping of 'runtimes' and 'packages', not a list) instead of a schema type error. As a semantic check, it's only shown once the config passes the schema.
• Treat an empty list as an empty stack, and only warn that stack should be specified for a composable image when it is missing or null, not when it is deliberately empty ([], {}, {runtimes: []}).

🤖 Generated with Claude Code

[upsun-dispatch]

Note

Reviewed — No blocking findings · 🔵 1 minor point

🔍 Full review · 7 files reviewed

🔵 Minor point

• internal/lint/types.go:54 — A non-empty stack list now gets an error from checkStackList, but the existing checks still treat it as a real stack, so the same problem is reported twice. isStackEmpty returns false for a non-empty list. With a non-composable type (e.g. type: php:8.4, stack: [php@8.4]), checkRuntime also adds the warning "'stack' is only used with a composable image type" (line 47). With an empty type, line 55 also adds "'type' should be specified (as a composable image) when using 'stack'". In both cases a stack the linter has just rejected as malformed gets a second report.

Verification

• Both schemas use "type": ["object","array","null"], and Flex app and task stacks both point to the single #/definitions/stack, so only lists and null newly pass the schema.
• CheckContent returns on schema errors before runChecks, so the non-empty list error only appears once the config passes the schema, as the description says.
• The composable warning now fires only when the stack is nil (missing or stack:); [], {} and {runtimes: [], packages: []} no longer warn.
• checkStackList runs before the task.Base == "" guard, so tasks using base still get the list check (covered by the types_test stack_list case).

The PR adds table cases in internal/lint/linter_test.go, fixed_test.go and types_test.go for empty stacks and stack lists, and updates two existing cases to the mapping form; CI runs them in the make test-go step of .github/workflows/ci.yml. No test covers a non-empty list combined with a non-composable type or a missing type.

Review details

• Commit: c3c54ca
• Model: claude-opus-5-5

---

Review 1 of 10 for this pull request · View the full run

[upsun-dispatch]

Warning

Changes suggested — 🟡 1 warning

🔁 Incremental · 2 files reviewed

Verification

• Config.Applications and Tasks are maps of struct values, so app.Stack = checkStackList(...) changes a local copy and does not affect the checks that run after CheckTypes.
• For a composable app with a list stack, the empty-map replacement is non-nil, so the extra warning 'stack' should be specified no longer appears.
• For a classic or docker app with a list stack, the replacement counts as empty for isStackEmpty, so the list error is not followed by a duplicate warning that 'stack' is only used with a composable image type.

The stack_list case in TestCheckTypes (internal/lint/types_test.go) now covers composable, classic, untyped and base-task apps with list stacks. Its expected output includes the extra type cannot be empty error reported in the finding, so the test locks that behaviour in.

Review details

• Commits: c3c54ca..287cd22
• Model: claude-opus-5-5

---

Review 2 of 10 for this pull request · View the full run