Context
Specialization-aware CFG replay now has deterministic depth-first traversal, reports an identical source diagnostic once per script even when multiple type/scalar variants contain it, and reports a shared .pt defect independently for each failing script.
Those mechanics are intentional, but the user-facing presentation policy is not yet explicit.
Decide and document
- whether replay order should remain call-closure DFS order or be globally sorted by source file/location
- whether the CLI trailer should say that errors came from the script and its reachable code rather than only
error compiling scriptFile ...
- how a shared library/template diagnostic should be grouped when multiple scripts fail in one compiler invocation
- preserve per-script failure semantics; do not globally suppress a diagnostic and accidentally make later scripts appear successful
Acceptance tests
- two scripts reaching the same failing template retain independent failure results
- diagnostic source tokens continue to point at the
.pt file
- within one script, concrete and scalar companion variants produce one identical diagnostic
- the selected ordering and trailer wording are pinned end to end
This is diagnostic-policy polish, not a correctness blocker for PR #89.
Context
Specialization-aware CFG replay now has deterministic depth-first traversal, reports an identical source diagnostic once per script even when multiple type/scalar variants contain it, and reports a shared
.ptdefect independently for each failing script.Those mechanics are intentional, but the user-facing presentation policy is not yet explicit.
Decide and document
error compiling scriptFile ...Acceptance tests
.ptfileThis is diagnostic-policy polish, not a correctness blocker for PR #89.