Port gHashTag/trios:crates/trios-trainer/src/data.rs (Rust, 7 functions) to specs/port/trios/crates/trios-trainer/src/da

[gHashTag]

Closes #7342

Written by a bee on queen-7342 and published by tools/queen/publish.py. The branch itself is the bee's; the second commit is the coordination entry every pull request must add, which a bee has no way to know about.

1 file changed, 88 insertions(+)

🤖 Generated with Claude Code

[t27-bees]

Reviewer bee verdict for head 270db28c1b03ebedc504ee2d9de108fdc2a7021d (tools/bees/reviewer.py, zai glm-4.7-flash, 3 turns, 63 s).

BEE-VERDICT: REQUEST_CHANGES
summary: Violates boundary (added docs/now/), has placeholder tests and tests calling undefined functions
criterion: "test -f specs/port/trios/crates/trios-trainer/src/data.t27 && echo present" -- met -- evidence: brief.md:698
criterion: "grep -cE '^\s*(pub )?fn (FineWebDataset_load|...)' prints 7" -- met -- evidence: brief.md:699
criterion: "t27c gen specs/...data.t27 && grep -c 'not yet implemented' prints 0" -- met -- evidence: brief.md:700
criterion: "t27c gen specs/...data.t27 | wc -l prints at least 13" -- met -- evidence: brief.md:701
criterion: "t27c spec-status prints not NOPARSE" -- met -- evidence: brief.md:702
criterion: "grep -cE '^test' prints at least 7" -- met -- evidence: brief.md:703
criterion: "t27c test-report prints no BLOCKED" -- met -- evidence: brief.md:704
criterion: "File outside boundary" -- unmet -- evidence: issue:31-33 (only this file); pr.diff:1-2 (added docs/now file)
criterion: "Test functions assert behavior from original" -- unmet -- evidence: data.t27:93-98 (assert(true)), 100-105 (assert(true)), 107-111 (assert(true))
criterion: "No test calls undefined functions" -- unmet -- evidence: issue:51 (and no test may call one); data.t27:95-97, 103-104, 109 (tests call undefined functions)
discounted-check: Corpus ratchet (expected-failure ledger) -- This failure occurs on master with the same pattern (brief.md:530), so it does not count against this head
discounted-check: GitGuardian Security Checks -- Not run on master for last 6 commits, insufficient data to compare (brief.md:694)
discounted-check: pr-dashboard -- This failure is infrastructure/network ("unexpected end of JSON input"), not caused by this PR's code (brief.md:636)
blocking-check: duplicate-bodies -- The logs show three new duplicate bodies were created: bytes64_eq, nibble (brief.md:579-582), and these were not in any ledger (brief.md:578)
blocking-check: t27b-native-ratchet -- The ratchet verdict is red (brief.md:684-685), caused by the new spec file violating the "only this file" rule

Evidence

Review of Pull Request #7539

Analysis

1. Does the change do what issue #7342 asks?

NO. The PR violates multiple requirements from the issue:

Violation 1: File outside the boundary

• The issue explicitly states "Only this file" and "Write this one file and nothing else"
• The PR added docs/now/2026-10-07-published-port-ghashtag-trios-crates-trainer-src-data-rust-7-.md
• This file is NOT in the boundary (specs/port/trios/crates/trios-trainer/src/data.t27)
• The brief.md even notes: "This entry is written by the publisher, not by the bee. A pull request must add exactly one docs/now/ entry and a bee has no way to know that: its brief names a boundary file and acceptance criteria, and docs/now/ is neither."

Violation 2: Tests that don't assert anything from the original

• test get_slice_function (line 93-98): just assert(true) - doesn't assert what the original does
• test sample_sequence_undefined (line 100-105): just assert(true) - the function returns undefined but the original sample_sequence returns actual values based on rng_state
• test get_eval_batch_undefined (line 107-111): just assert(true) - same issue

Violation 3: Functions not ported correctly

• The issue states: "Port the DECISION inside it as a pure helper and test the helper; the function itself keeps its name with an undefined; body that no test calls"
• However, 3 of the 7 test functions call FineWebDataset_get_slice, FineWebDataset_sample_sequence, and FineWebDataset_get_eval_batch directly
• These functions return undefined (lines 54, 61, 67 in data.t27) but the tests call them - violating the requirement that "no test may call one"

The runner's criteria show all passed (facts from brief.md lines 698-704), but these facts are misleading because:

1. They only check file existence and function names
2. They don't verify the tests actually assert behavior from the original
3. They don't catch the violation of "no test may call undefined functions"

2. Red checks

BLOCKING CHECK: duplicate-bodies

• Why it counts: The failure occurred because the new spec file introduces three new duplicate bodies in the generated code
• The logs show: "a body written under bytes64_eq is now copied [2] time(s) and was in no ledger"
• This failure is directly caused by adding this new spec file

BLOCKING CHECK: t27b-native-ratchet

• Why it counts: The ratchet verdict is red
• This failure is caused by the new spec file violating the rule that only the boundary file should change

DISCOUNTED CHECK: Corpus ratchet (expected-failure ledger)

• Why it does NOT count: On master, the same corpus ratchet check failed with "A type name may not gain a second definition" - this is the same expected failure pattern
• brief.md line 530 confirms master 2749341 did not accept a discount for this check

DISCOUNTED CHECK: pr-dashboard

• Why it does NOT count: The failure is "unexpected end of JSON input" from GitHub API - this is an infrastructure/network issue, not caused by this PR's code
• brief.md line 636 confirms master 2749341 is before the PR's merge base, so we can't verify the check status on master

DISCOUNTED CHECK: GitGuardian Security Checks

• Why it does NOT count: Not run on master for the last 6 commits - insufficient data to compare

Evidence

Violation of boundary file rule:

• Issue #7342 line 29: "Boundary: specs/port/trios/crates/trios-trainer/src/data.t27"
• Issue #7342 line 31-33: "Only this file... Anything you need to try out goes under /tmp, never into the repository"
• PR diff shows 2 files changed (brief.md lines 466-469)

Tests that don't assert original behavior:

• data.t27 lines 93-98: test get_slice_function with assert(true)
• data.t27 lines 100-105: test sample_sequence_undefined with assert(true)
• data.t27 lines 107-111: test get_eval_batch_undefined with assert(true)

Tests calling undefined functions:

• data.t27 lines 95-97: test calls FineWebDataset_get_slice which has undefined body
• data.t27 lines 103-104: test calls FineWebDataset_sample_sequence which has undefined body
• data.t27 line 109: test calls FineWebDataset_get_eval_batch which has undefined body

Issue requirement (line 51): "and no test may call one"
Issue requirement (line 74): "a test that asserts true ports nothing"

duplicate-bodies failure logs:

• brief.md lines 579-582: "a body written under bytes64_eq is now copied [2] time(s) and was in no ledger"

t27b-native-ratchet failure:

• brief.md lines 684-685: "ratchet verdict red, exit 1"