self-host: gen-c drops the value of a global [_]T{v} ** N repeat #6050

Profile (read first)

You are working as Gamma (Compiler Core) on epic #5980: rewrite t27c in t27 itself.
The self-hosted core (specs/compiler/core/t27core.t27, PR #6041) must write exactly the
bytes t27c gen-c writes, so every gen-c defect is a defect the core has to refuse instead of
learn. Fixing it here is what lets the core grow. Rules for this task:

• Change the Rust emitter only inside the Boundary below. gen/ is generated: never hand-edit it.
• Build with cargo build --release -p t27c and use ./target/release/t27c from the repository
  root, not a t27c on PATH.
• Write the regression test first and watch it fail on the unfixed compiler. Then fix, and
  revert the fix once to watch the test go red again. Say in the PR that you did both.
• Any other spec whose generated C changes must be listed in the PR by path, with one line on
  why the new bytes are right. A silent corpus change is a failed task.
• The PR body says Closes # plus this issue's number and Part of #5980. ASCII only in source files.

Context

A global array initialised with [_]u8{5} ** 4 is emitted as /* repeat: { 5 } ** 4 */ {0}: the repeated value is lost and the array is all zeros. #6038 fixed the typed [T; N] local and return forms; this is the global [N]T form, still wrong on master. The self-hosted core refuses any repeat value other than 0 (E_ARRAY).

Reproduce

cargo build --release -p t27c
printf 'module m;\nconst N: i64 = 4;\nvar r: [N]u8 = [_]u8{7} ** N;\nvar a: [4]u8 = [_]u8{5} ** 4;\n' > /tmp/rep.t27
./target/release/t27c gen-c /tmp/rep.t27

Today the relevant output is:

static uint8_t r[N] = /* repeat: { 7 } ** N */ {0};
static uint8_t a[4] = /* repeat: { 5 } ** 4 */ {0};

Original checks (kept for history; the Queen does not run these)

• 1. ./target/release/t27c gen-c /tmp/rep.t27 > /tmp/rep.c && printf '#include "rep.c"\nint main(void){return (a[0]==5 && a[3]==5 && r[0]==7 && r[3]==7)?0:1;}\n' > /tmp/repm.c && cc -w -o /tmp/repm /tmp/repm.c && /tmp/repm; echo $? prints 0 (today 1)
• 2. the same program builds and passes with gcc too if it is installed; say in the PR which compilers you ran
• 3. cargo test --release -p t27c --test genc_global_repeat passes

Original file list

• bootstrap/src/compiler.rs
• bootstrap/tests/genc_global_repeat.rs

Original scenarios

• Given /tmp/rep.t27 written by the Reproduce block,
  when ./target/release/t27c gen-c /tmp/rep.t27 runs and its output is compiled with cc,
  then every element of a is 5 and every element of r is 7, for a literal count and for a const count.

Original requirements

• FR-001: the repeated value MUST reach every element; {0} is right only when the value is 0.
• FR-002: do not change the [T; N] paths fix(c): initialize typed repeat arrays inside their value wrapper #6038 just fixed.
• FR-900: cargo test --release -p t27c --test genc_global_repeat MUST pass, and MUST fail with the fix reverted.
• FR-901: if bootstrap/tests/core_selfhost.rs exists on your base, cargo test --release -p t27c --test core_selfhost MUST still pass.

Queen-runnable restatement (2026-10-05)

This block restates the issue so the Queen's runner reads one boundary and one list of checks; the text above is kept for history and its own file list and checks are superseded by the sections below.

The bee branch queen-6050 carries bootstrap/tests/genc_global_repeat.rs (63 lines of hand-written Rust test), which cannot land under only-t27. On master t27c gen-c still emits static uint8_t a[4] = /* repeat: { 5 } ** 4 */ {0}; for a global array repeat, so the bug is live. The t27 part is a rule spec in the shape of specs/compiler/zig_test_shadowing.t27: it pins the construct with tests that run today through the Zig backend, and its comments state what the C backend must emit, so the Rust fix (owner-approved, separately) has an executable target. Globals must not be named a or r: the generated Zig reports a parameter shadowing them.

Boundary

• specs/compiler/c_global_repeat.t27

User Scenarios & Testing

• Given a module-level array initialised with [_]u8{5} ** 4 and with [_]u8{7} ** N (N a const), when the rule spec reads the elements back, then every element equals the repeated value, and the spec's comments name the C output each global must get ({5, 5, 5, 5}, not {0}).

Requirements

• FR-001: specs/compiler/c_global_repeat.t27 MUST declare a global array initialised by a literal-count repeat and one initialised by a const-count repeat, and test first and last element of each.
• FR-002: Its // comments MUST cite self-host: gen-c drops the value of a global [_]T{v} ** N repeat #6050 and state the C initialiser the backend must emit for each global.
• FR-003: The change MUST NOT add or modify non-t27 hand-written code (specs/policy/own_language.t27).

Success Criteria

• test -f specs/compiler/c_global_repeat.t27 && echo present prints present (today the file does not exist)
• grep -cE '\{[0-9]+\} \*\* ' specs/compiler/c_global_repeat.t27 prints at least 2 - both repeat forms are present
• grep -c '6050' specs/compiler/c_global_repeat.t27 prints at least 1
• t27c test-report specs/compiler/c_global_repeat.t27 2>&1 | grep -cE '^ FAIL +0$' prints 1 - the generated code compiles and every test passes
• t27c test-report specs/compiler/c_global_repeat.t27 2>&1 | grep -E '^ tests ' | tr -dc 0-9 prints at least 2 - the count of tests that ran