Result: blocked
Files: none
Commit: none
Pushed: none
Tests: none run (nothing built); the brief’s probe run on target/fast/nikaia at 843e716 prints the 21 in lines and no item N frame between Value__Contact and Value__Vec_Contact_, though the failure is in list element 1: the Cut drops the backend’s element frame
Gate: green
Deviation: none
Found: none
Needs: architect decision: the list index cannot reach the error without a change to winnow-grammar 4f8e669. (1) src/rt.rs’s repetitions (about lines 1005, 1082, 1248) and src/rt/stream.rs (515, 868) call e.item(n) only on ErrMode::Backtrack; a Cut error takes Err(e) => return Err(e) and loses the index. Every members[S] failure is a Cut (k:Key => ..., fail writes ErrMode::Cut), and so is the probe’s "{" =>, so $.contacts[1]/[0] cannot be told in steps 1 and 2. (2) An item N frame does not say which repetition it came from (digit+ inside an element also pushes item 1), and the template rule __nikaia_list is inlined by the macro (no frame of its own), so the list’s repetition cannot be marked. (3) A frame’s name is a static Label, and an extern rule call maps every error to Backtrack (codegen/expr.rs 1348-1358), so a hand-written list cannot carry it either. The backend must offer one of: (a) repetitions that also number a Cut error, plus a way to mark the list’s repetition (for example a non-inlined rule boundary); or (b) a pattern that pushes a frame computed at run time onto a failing error, Backtrack or Cut (for example frame(expr, p), using Diagnostics::item’s path). Field segments need no backend change: a marker rule per field, __nikaia_members_N_field_NAME wrapping _value’s alternative i, read and dropped in nikaia_std::grammar::parse* before render. I left that half unbuilt on purpose, because a path without its indices ($.contacts.mail) would be wrong
Changelog: none
Owner: none
Disk freed: 0 GB (target/nikaia-project-tests 3.6G, / 11G free; no new cache entries)
Rs2nika: none
