The gate I added could never have run, and enum was not the only key the converter drops #5275

Re-filed from #2724. An attempt claimed that issue on 2026-09-24 and left an empty branch, and a dispatch row that has spent its retry ceiling keeps its issue for good - so no bee could take it again, however much the brief improved. The work below is carried over unchanged.

Three follow-ups. Two of them found something wrong with earlier work in this same session.

1. Two numbers that were wrong in two places each

lsp/language.t27 and lsp/protocol.t27 each carried a length written into both the declared type and the assertion, so the two agreed with each other and with nothing else -- neither could be checked while the files did not compile.

Settled by counting, not by preference. The strings are right; the counts were wrong. Their siblings of identical shape -- "t27"->3, "python"->6, "exit"->4, "textDocument/didClose"->21 -- were all correct and passing, which is what made the two stand out.

lsp/language.t27 now runs 14 tests, all green -- a file whose tests had never executed at all. lsp/protocol.t27 is 30 passing, 1 failing, and that one stays: message_buffer_create reports a capacity it cannot hold (#2719), which is an API decision.

2. The self-check gate had no Zig -- it could not have run once

I wired harness_selfcheck.py into the prove job earlier this session. It sits at step 10 of 70. Nothing installs Zig before it, and nothing in that job installs Zig at all.

A gate that cannot execute is the decoration it was written to prevent. Fixed by putting setup-zig immediately ahead of it.

I nearly made it worse: my first move today was to add a second self-check step further down the same job -- a duplicate, also without Zig. Caught by listing the job's step names before committing.

Version matters here and the repo disagrees with itself. The only other Zig in CI is phi-loop-ci.yml, pinned to 0.13.0; every figure the run scan has produced was measured on 0.16.0. They are not interchangeable for this purpose -- usingnamespace was removed in 0.16 and the @typeInfo tags were renamed -- so the gate pins 0.16.0 and says why in a comment.

3. enum: was not the only key the converter flattens

Counted structurally across all 366 ancestors -- every entry under a type whose value is a nested block:

variants: is the commoner of the two bare-name keys, and I fixed the rarer one first. Only 3 reached this corpus, because most of those ancestors have no route -- so the number that matters for a future conversion is 41, not 3.

Handled now. The converter derives 35 enum declarations, up from 18, and 34 of 34 comparable ones are identical to the hand recovery, 0 differing. The one unmatched is my cross-check script pairing http.t27 with server/http.t27 by basename -- the same trap as #2673, in my own tooling again.

union: and cases: carry a nested map per case rather than bare names, which the flat field grammar cannot express. Left open deliberately rather than half-done.

Measured across all of it: ast-check valid 296, errors 588, 0 specs worse, parse gate holds at 105 events / 0 lost.

Original file list

• specs/lsp/language.t27
• specs/lsp/protocol.t27
• .github/workflows/phi-loop-ci.yml

The instruments

Every one of these reads and prints; none of them writes. Run them from the repository root, with <spec> replaced by the file in ## Boundary:

.t27 is not Rust and not Zig, and a bee fluent in either writes one by
accident. Measured 2026-09-20: a bee filled eight bodies in
specs/file/watcher.t27 with return Ok(());, Err(FileError::WatcherNotFound)
and for i in 0..watchers.length. The parse ratchet refused the whole file --
Unexpected token in expression: RParen, line 123 -- and a spec that does not
parse generates nothing, so every test it already carried stopped running.

At the top level the parser accepts exactly these eight forms, with an optional
pub, and no others:

const   var   fn   enum   struct   test   invariant   bench

There is no trait, no impl, no type X = ..., no generics, no macro. What
a bee reaches for from another language, and what happens:

• Rust's result sugar (Ok(()), Err(E::V), ?) is not a construct here.
• Rust macros (println!, format!, vec!) are not constructs here.
• A range loop (for i in 0..n) is Rust and Zig, not this language.
• Zig builtins (catch unreachable, @intCast, @constCast) are not constructs here.
• Cross-module reuse (use other::fn;) parses and then generates a comment and an unqualified call, so the Zig fails with use of undeclared identifier (use a::b; generates no import, so no spec can reuse a function from another spec #4298). It does not work yet.

t27c parse <spec> answers in one line whether what you wrote is the language.
Run it before you report, every time.

• t27c spec-status <spec> - the compiler's one-word verdict: IMPLEMENTED, PARTIAL, UNWRITTEN, NOPARSE, NOFN. Exit code is 0 whatever it says, so read the word.
• t27c symbols <spec> - every name the file declares, with its kind. Answers "does this already exist here?" before you add it.
• t27c outline <spec> - per function: its locals, what it calls, what it returns. The contract you are implementing against.
• t27c coverage <spec> - which functions have a test and which do not. The issue asks for tests; this is how you check you wrote them.
• t27c lint <spec> - style and shape warnings. It OVER-REPORTS has no test or invariant: measured 2026-09-20 over 60 specs it printed 677 of those where coverage found 190 untested functions, disagreeing on 59 of the 60 - it warns about bit_to_trit_pair in specs/base/ternary_encoding.t27, which test bit_to_trit_pair_zero calls on line 249. Read it as a hint; coverage is the answer.
• t27c typecheck <spec> - types, before generation. Prints Typecheck OK (0 errors, 0 warnings) or the errors.
• t27c test-report <spec> - builds this spec and runs its own tests, which is what the oracle does. BLOCKED means the generated Zig does not compile, with the error beside it.
• python3 tools/dupe_scan.py --name <function> - where that function already lives, if it does. 576 of 4021 bodies here are byte-identical copies.

The rest of the toolbelt is in docs/BEE_TOOLBELT.md.

Queen-runnable restatement (2026-10-05)

Item 1 never reached master: commit 0b4e93d (the length fixes) is not an ancestor of master, and master's specs/lsp/language.t27 is a 14-line stub with one TODO and no tests. specs/lsp/protocol.t27 already carries METHOD_COMPLETION as a slice (#4786). Items 2 and 3 are workflow YAML and the Rust converter, out of scope under the only-t27 rule.

Boundary

• specs/lsp/language.t27
• .trinity/seals/lsp_Specs.Lsp.Language.json
• .trinity/seals/lsp-language.json
• .trinity/seals/lsp_lsp-language.json

User Scenarios & Testing

• Given master, where specs/lsp/language.t27 is a stub and t27c test-report prints tests 0, when the language spec is restored, then it declares the language ids (with LANG_ZIG = "zig", three bytes) and its tests run with FAIL 0.

Requirements

• FR-001: specs/lsp/language.t27 MUST declare the language id constants (LANG_T27, LANG_ZIG, LANG_PYTHON, ...) and helpers such as language_id_from_extension and is_t27_extension, as in 0b4e93d, written in constructs t27c compiles, with tests asserting concrete values (for example the length of LANG_ZIG is 3).
• FR-002: The change MUST NOT add or modify non-t27 hand-written code (specs/policy/own_language.t27).
• FR-003: Remove the TODO: Implement stub body.
• FR-004: After the edit run t27c seal specs/lsp/language.t27 --save (the seal JSON is t27c output).

Success Criteria

• grep -c 'LANG_ZIG' specs/lsp/language.t27 prints at least 1
• grep -c 'TODO: Implement' specs/lsp/language.t27 prints 0
• t27c test-report specs/lsp/language.t27 2>&1 | grep -cE '^ (FAIL +0|pass +[1-9][0-9]*)$' prints 2