[bug] t27o emits 55 specs that t27c cannot compile either -- undeclared names, negative enum(u32) values, _ identifiers, >>=, open ranges #10905

Summary

55 of the 1,216 specs t27o emitted on the roadmap repositories (step 1 of #10879) do not compile under the reference t27c test-report either: the converter writes text that is not valid t27. A converter must refuse what it cannot express, never emit an invalid spec.

Expected vs actual (counts from the step-1 run, master 57aa3bc)

Steps to reproduce

tri t27o <file.rs> on the examples, then t27c check / t27c test-report on the output.

Fix

Each class either converts to valid t27 (with a test that the output compiles under t27c AND runs under t27b) or is refused with its reason. A test per class, a negative control.

Boundary

• May touch: specs/tools/t27o/rust_convert.t27 (and zig_convert.t27 if a Zig spec is in the 55), their gen files and seals.
• Must not touch: anything under specs/tri/t27b/ or cli/t27b/ (the t27b side is t27b for t27o: ExprIdentifier(undeclared) -- unlocks 24 of 1216 t27o specs (#10879) #10897 and its siblings).

Links

#10879, #10522, #10883.

User Scenarios

• Given a Rust file containing an undeclared name, when t27o converts it, then the converter either emits valid t27 that compiles under t27c or refuses with a clear reason.
• Given a Rust file containing a negative value in an enum(u32), when t27o converts it, then the converter either emits valid t27 that compiles under t27c or refuses with a clear reason.
• Given a Rust file containing _ as an identifier, c >>= 1, or path[..i], when t27o converts it, then the converter either emits valid t27 that compiles under t27c or refuses with a clear reason.

Requirements

• FR-001: The converter MUST refuse to emit specs with undeclared names instead of producing invalid t27.
• FR-002: The converter MUST refuse to emit specs with negative values in enum(u32) instead of producing invalid t27.
• FR-003: The converter MUST refuse to emit specs using _ as an identifier instead of producing invalid t27.
• FR-004: The converter MUST refuse to emit specs containing >>= or open ranges like [..i] instead of producing invalid t27.
• FR-005: Each refusal MUST include a diagnostic naming the unsupported construct and the source location.
• FR-006: For each of the five problem classes, a test case MUST exist that passes when the converter refuses or emits valid t27, and fails on the current behavior.

Success Criteria

• t27c check specs/tools/t27o/rust_convert.t27 exits 0
• t27c test-report specs/tools/t27o/rust_convert.t27 exits 0
• 5 new negative-control tests in the converter test suite each produce a refusal diagnostic for their respective problem class (undeclared name, negative enum u32, _ identifier, compound shift, open range)
• Running tri t27o on the 55 problematic inputs from RESULTS.txt produces zero specs that fail t27c check