1. Most turns: finding why [1, 2] has no element type. About 8 greps went into tracing it: ListLit led to check_core_list_literal, then LitInt turned out to be Unknown, and the cause was finally the open-number machinery (ADR-285 D34) in check_numbers3.nika and check_core_stmt.nika:720-770, 1560. The brief’s hint (“integer-literal or not-yet-defaulted type, or Ty::Unknown”) was right but named no place to look.
2. Looked up, could have been handed:
   • The fact that integer literals are Ty::Unknown until each function’s open numbers are decided after its body.
   • The existing open_elements branch in the for checker, about 50 lines below the spot the brief pointed to; the fix belongs there.
   • How num::value is defined (the Value trait), which told me the nested rows can’t work with a copy alone.
3. Checks run:
   • One lower-std/build cycle.
   • The old and new binary on the probe, plus an edge probe (mut list, _, printing).
   • A standalone rustc copy of num::value for the nested case.
   • tests/compiler --both-settings, the guard set, clippy and pre-push-check.

   None was wasted. The rustc copy saved a build: it showed the nested rows fail for a second reason.

4. One change: where the probe matrix shows a declared-versus-untyped split, the brief should name the inference pass that types untyped values. Here that is ADR-285 D34 open numbers: flow.open_numbers and open_elements, and their one-level element flag. The architect could also have spotted in advance that the nested rows need D34 one level deeper and split them into their own step.

Process note: Disk freed was filled in after commit.nika ran. The report file is not among the committed files, so nothing committed holds the DISK placeholder.

This site is open source. Improve this page.