Tracking verified features and aspirational development in the Koru compiler.
These are not failures we haven't got to. Each one is a design question the compiler can't answer for us — the test stays red or parked until it's settled, and the question is written down where the test lives.
Signed runtime `/` and `%`: lower them to `@divTrunc`/`@rem`, or refuse them with a Koru diagnostic that names the operand's signedness?
030_143_signed_runtime_division · failure
How is a source block's type tag spelled — `Source<HTML>`, `Source[HTML]`, or not at all?
210_024_source_scope_capture · todo
Do compiler annotations on TYPE positions live or die — build `[]`-annotation extraction into its own AST field, or drop the surface and allow bare types only?
210_039_file_source_syntax · todo
The same deferred call as 210_039 — annotations on type positions — asked here by regex-in-brackets on branch types.
310_040_literate_annotations · todo
How is a borrow spelled, so `std/string:read` can say "this value lives as long as `s`"?
610_007_reject_dangling_slice · failure
Does Koru own the `:d:N` width semantic, or does the spec keep passing verbatim into Zig's format string?
620_005_width_spec_signed_no_plus · failure
T8 firing granularity: does an installed `! ?persist` arm fire per write, or once per batch — and may its handler hold a live connection?
690_017_store_backend_arms_di · todo
Does a batch insert fire per-row watches N times after the fused apply, or once with a plural payload?
690_019_store_batch_insert_fused_cascade · todo
What does a `vec3` column's watch observe — one whole-vector event, or one event per lane?
690_020_store_compound_field_types · todo
Daily count of verified passing features over time.