fix(workflows): state the equal-priority tie-break in workflow resolve output

[jawwad-ali]

Problem

workflow resolve prints the layer list under a bare header:

console.print("Layers (highest precedence first):")

collect_all_layers sorts by (priority, source) ascending, which puts the winning layer first only while priorities differ. On a tie the sort is alphabetical by source, while the merge gives the conflict to the last id — exactly as documented:

Equal-priority overlays are applied alphabetically by ID, with the last ID winning conflicts. — docs/reference/workflows.md:126

So on a tie the header states the opposite of the outcome printed directly beneath it, in the one command whose documented purpose is explaining which overlay contributed or overrode a step.

Reproduction on current main (c173bf1)

DIFFERENT priorities (alpha=5, beta=10)
   listed first : project:alpha
   actual winner: echo FROM-ALPHA
   header correct? YES

EQUAL priorities (both 10)
   listed first : project:alpha
   actual winner: echo FROM-BETA
   header correct? NO  <-- inverted

The full output contradicted itself two lines apart:

Layers (highest precedence first):
  • [project-overlay] project:alpha (priority=10)
  • [project-overlay] project:beta (priority=10)
Step attribution:
  • build: project:beta          <-- beta won, alpha was labelled highest

Fix — the label, not the order

The ordering is deliberate: test_workflow_resolve_equal_priority_layers_sort_by_source pins it, and its own comment acknowledges that the alphabetically-last layer is the one that wins. Reordering the list would fight a decision the maintainers made on purpose, so this spells the tie-break out instead:

Layers (highest precedence first; on equal priority the last ID wins):
  • [project-overlay] project:alpha (priority=10)
  • [project-overlay] project:beta (priority=10)
Step attribution:
  • build: project:beta

Now self-consistent, and the behaviour is untouched.

Verification

• Fail-before / pass-after: 1 new-vs-baseline failure with the source reverted to upstream/main → 26 passed with the fix.
• The new test asserts the header and the attribution together, so it pins the label against the real outcome rather than against a hard-coded string alone.
• No existing test asserts the old header text (checked: git grep "highest precedence first" -- tests/ is empty), and test_workflow_resolve_equal_priority_layers_sort_by_source passes unchanged.
• uvx ruff@0.15.0 check src tests → clean

No behaviour change — output text only.

Note on overlap: this touches overlays/_commands.py, as does my #4141, but a different function (workflow_resolve vs workflow_overlay_add). Happy to rebase whichever lands second.

---

Written with assistance from Claude Code. Bug found, reproduced, and verified by me on current main.

🤖 Generated with Claude Code