From 617fdb838c9939846e1062a54ba46f36b4315593 Mon Sep 17 00:00:00 2001 From: Dmitrii Vasilev Date: Mon, 5 Oct 2026 03:31:39 +0700 Subject: [PATCH 1/2] fix(gates): green doc-refs and orphan-artefacts on main after #510 #510 (opened 2026-08-08, merged 2026-10-04) added the GF-T paper documents from before the GF-T -> TEF -> TNF renames. Main has failed two gates since 57de4a1f5, and every branch cut from it fails gates-can-fail because those two gates are untestable on a red tree. doc-refs, 7 new dangling references, 9 unpublished exclusions, 2 stale baseline entries: - 3 references in the inventory named files in our other repositories by bare name, and 2 of the bare names matched an unrelated file in a sibling (PLAN.md in tri-net, main.tex in t27). Qualified with the repository the inventory row already names: gHashTag/goldenfloat-preprint (private), gHashTag/arith2027-goldenfloat, gHashTag/paper3-methodology. - 7 cross-repo exclusions are genuine (tri-net fpga/gft/*.v, t27 docs/arxiv-submission/trinity-gf16.tex) and are now published, written by --update-lists against materialised siblings. - 6 references name the GF-T oracles and a build directory that no tree has. The documents are records of the August state; they go in the baseline, as the same names from sibling documents already do, and RENAME_MAP says what they became. - 2 baseline entries no longer reproduce because #510 added the files they named; removed. orphan-artefacts: the two competition artefacts in research/arxiv_gft/ have no generator in any gHashTag repository that code search can see. Listed in the baseline rather than covered by a generator that was not the one used. Co-Authored-By: Claude Opus 5.5 --- research/GF_INVENTORY_2026-08-08.md | 8 ++++---- research/RENAME_MAP.md | 10 ++++++++++ tools/doc_refs_baseline.txt | 8 ++++++-- tools/doc_refs_crossrepo.txt | 7 +++++++ tools/doc_refs_upstream.txt | 4 ++++ tools/orphan_artefacts_baseline.txt | 2 ++ 6 files changed, 33 insertions(+), 6 deletions(-) diff --git a/research/GF_INVENTORY_2026-08-08.md b/research/GF_INVENTORY_2026-08-08.md index c43fedec9f..d976c20231 100644 --- a/research/GF_INVENTORY_2026-08-08.md +++ b/research/GF_INVENTORY_2026-08-08.md @@ -24,9 +24,9 @@ GF-T — это не «GF на тернаре», это отдельный фо | Репозиторий | Что это | Состояние | |---|---|---| -| **`goldenfloat-preprint`** 🔒 | **Канонический источник препринта GoldenFloat** — `gf_preprint_v19.tex` + PDF + CHANGELOG | v19. Правки GF вносить СЮДА | -| **`arith2027-goldenfloat`** | Заготовка под **ARITH 2027**, 8 страниц IEEE CS | пока только `PLAN.md` + README — не написана | -| **`paper3-methodology`** | «An 84-Format Numeric Catalog with Bit-Exact Conformance Vectors» — `main.tex` + два PDF (v3, v4) | опубликована как arXiv:2606.09686 | +| **`goldenfloat-preprint`** 🔒 | **Канонический источник препринта GoldenFloat** — `gHashTag/goldenfloat-preprint/gf_preprint_v19.tex` + PDF + CHANGELOG | v19. Правки GF вносить СЮДА | +| **`arith2027-goldenfloat`** | Заготовка под **ARITH 2027**, 8 страниц IEEE CS | пока только `gHashTag/arith2027-goldenfloat/PLAN.md` + README — не написана | +| **`paper3-methodology`** | «An 84-Format Numeric Catalog with Bit-Exact Conformance Vectors» — `gHashTag/paper3-methodology/main.tex` + два PDF (v3, v4) | опубликована как arXiv:2606.09686 | | **`trinity-papers-ru`** | Русские версии статей (для ВАК?) | TeX | | **`phi-paper`** 🔒 | Pellis–Vasilev–Olsen, методологическая | TeX | | **`trinity-clara`** | Пакет подачи DARPA CLARA | TeX | @@ -115,7 +115,7 @@ GF-T — это не «GF на тернаре», это отдельный фо ## Куда что вносить -- Правки **GF** (бинарная лестница) → `goldenfloat-preprint`, файл `gf_preprint_v19.tex` +- Правки **GF** (бинарная лестница) → `goldenfloat-preprint`, файл `gHashTag/goldenfloat-preprint/gf_preprint_v19.tex` - Правки **GF-T** → новая статья `trinity-fpga/research/arxiv_gft/gft_paper.tex` - Материал под конференцию → `arith2027-goldenfloat` (пустая заготовка, дедлайн ARITH 2027) - Оракулы и векторы → `trinity-fpga/conformance/` diff --git a/research/RENAME_MAP.md b/research/RENAME_MAP.md index 993d6ef19d..6c60638f3d 100644 --- a/research/RENAME_MAP.md +++ b/research/RENAME_MAP.md @@ -24,6 +24,16 @@ throughout, in prose and in filenames. | `tef_mul_wp.v`, `gft_add_w` -- absent from every tree | the TEF-era module names; the surviving RTL is under `fpga/tnet/` and `fpga/phiscale/` | | `main_ru.tex`, `paper1-goldenfloat/main_ru.tex` -- absent from every tree | retired -- the Russian-language paper line was not carried forward | | `ARXIV_UPLOAD_EN_v2.md`, `ARXIV_GFT16_v2.md` | superseded by `research/arxiv_tnf/` | +| `trinity-fpga/build/gft_mul8/gft_mul.v` -- a build directory, never tracked | the module of that name is `fpga/gft/gft_mul.v` in `tri-net`; its width-correct replacement here is `fpga/gft/gft_mul_w.v` | + +Two of those names, `ARXIV_GFT16_v2.md` and `gft_paper.tex`, are back in the +tree as records. #510, opened 2026-08-08 and +merged 2026-10-04, added `research/ARXIV_GFT16_v2.md` and `research/arxiv_gft/` +(the GF-T paper, its figures and its build notes) together with the `fpga/gft/` +RTL. Those documents predate both renames and name the GF-T oracles in the first +row, which this tree no longer has, so the accuracy numbers in the GF-T paper +cannot be re-derived from this tree. The current paper is still +`research/arxiv_tnf/`. ## The prefix hazard diff --git a/tools/doc_refs_baseline.txt b/tools/doc_refs_baseline.txt index d5624f42bf..af3044c670 100644 --- a/tools/doc_refs_baseline.txt +++ b/tools/doc_refs_baseline.txt @@ -193,9 +193,10 @@ docs/v3-squeeze/S-17-razor-ff.md: names `tools/check_no_star.sh`, which does not docs/v3-squeeze/S-19-dual-mac-pe.md: names `check_no_star.sh`, which does not exist docs/v3-squeeze/S-20-multi-vt-cells.md: names `check_no_star.sh`, which does not exist research/ARXIV_BODY_FIXES_READY_TO_PASTE.md: names `main_ru.tex`, which does not exist -research/ARXIV_GFT16_SNIPPET.md: names `ARXIV_GFT16_v2.md`, which does not exist research/ARXIV_GFT16_SNIPPET.md: names `conformance/gft16_ref.py`, which does not exist research/ARXIV_GFT16_SNIPPET.md: names `gft16_ref.py`, which does not exist +research/ARXIV_GFT16_v2.md: names `trinity-fpga/build/gft_mul8/gft_mul.v`, which does not exist +research/ARXIV_GFT16_v2.md: names `trinity-fpga/conformance/gft16_ref.py`, which does not exist research/ARXIV_V2_CORRECTION_PACKAGE.md: names `ARXIV_UPLOAD_EN_v2.md`, which does not exist research/ARXIV_V2_CORRECTION_PACKAGE.md: names `main_ru.tex`, which does not exist research/ARXIV_V2_CORRECTION_PACKAGE.md: names `paper1-goldenfloat/main_ru.tex`, which does not exist @@ -203,6 +204,9 @@ research/CITED_NUMBERS_2026-08-20.md: names `gf_paper/src/main.tex`, which does research/CITED_NUMBERS_2026-08-20.md: names `takum_src/extracted/hunhold-beating_posits_at_their_own_game.tex`, which does not exist research/DIV_SQRT_HONESTY.md: names `gf_div_ref_tb.v`, which does not exist research/DIV_SQRT_HONESTY.md: names `verify_div_rtl.py`, which does not exist +research/GFT_DEFECTS_2026-08-08.md: names `gft_ref.py`, which does not exist +research/GF_INVENTORY_2026-08-08.md: names `gft16_ref.py`, which does not exist +research/GF_INVENTORY_2026-08-08.md: names `gft_ref.py`, which does not exist research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `conformance/gft16_ref.py`, which does not exist research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `conformance/gft_ref.py`, which does not exist research/LITERATURE_SCAN_2024_2026.md: names `synth.sh`, which does not exist @@ -217,10 +221,10 @@ research/SUBMISSION_CHECKLIST.md: names `main_ru.tex`, which does not exist research/SUBMISSION_CHECKLIST.md: names `trinity-papers-ru/paper1-goldenfloat/main_ru.tex`, which does not exist research/SUBMISSION_PACKAGE_2026-08-05.md: names `conformance/gft16_ref.py`, which does not exist research/SUBMISSION_PACKAGE_2026-08-05.md: names `paper3-methodology/main.tex`, which does not exist -research/THREE_PAPER_UPDATE_2026-08-08.md: names `gft_paper.tex`, which does not exist research/THREE_PAPER_UPDATE_2026-08-08.md: names `gft_ref.py`, which does not exist research/WITNESS_RECORDS_READY_TO_PASTE.md: names `fpga/openxc7-synth/bcd_decode.v`, which does not exist research/XC7A200T_GF16_DATAPOINT_2026-08-05.md: names `trinity-papers-ru/paper1-goldenfloat/main_ru.tex`, which does not exist +research/arxiv_gft/README.md: names `conformance/gft16_ref.py`, which does not exist research/arxiv_submission/submission_checklist.md: names `_sub.json`, which does not exist research/arxiv_submission/submission_checklist.md: names `docs/migration-map.md`, which does not exist research/arxiv_v3_catalog_update.md: names `_add.json`, which does not exist diff --git a/tools/doc_refs_crossrepo.txt b/tools/doc_refs_crossrepo.txt index 3a71c3a8b0..9f803e1bad 100644 --- a/tools/doc_refs_crossrepo.txt +++ b/tools/doc_refs_crossrepo.txt @@ -5,5 +5,12 @@ docs/REFEREE-PAGE.md: names `gft_smul.t27` -> resolves in t27 docs/docs/tri27/index.md: names `hello.t27` -> resolves in tri-net docs/handoff/HANDOFF_AGENT.md: names `docs/NOW.md` -> resolves in t27 docs/research/gf16_family_phi_layout.md: names `vectors.json` -> resolves in zig-golden-float +research/GFT_DEFECTS_2026-08-08.md: names `gft_add64.v` -> resolves in tri-net +research/GFT_DEFECTS_2026-08-08.md: names `gft_mul.v` -> resolves in tri-net +research/GFT_DEFECTS_2026-08-08.md: names `gft_mul32.v` -> resolves in tri-net +research/GFT_DEFECTS_2026-08-08.md: names `trinity-gf16.tex` -> resolves in t27 +research/GF_INVENTORY_2026-08-08.md: names `gft_add64.v` -> resolves in tri-net +research/GF_INVENTORY_2026-08-08.md: names `gft_mul.v` -> resolves in tri-net research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `tri_compute_bitnet.t27` -> resolves in tri-net +research/arxiv_gft/README.md: names `gft_mul32.v` -> resolves in tri-net research/frontier/DOCUMENT_TRACEABILITY_2026-08-10.md: names `cases.yaml` -> resolves in claim-audit-lab diff --git a/tools/doc_refs_upstream.txt b/tools/doc_refs_upstream.txt index b091ae2974..2561f34bce 100644 --- a/tools/doc_refs_upstream.txt +++ b/tools/doc_refs_upstream.txt @@ -41,6 +41,10 @@ research/ERRATUM_arXiv_2606.09686_catalog_count.md: names `gHashTag/t27/specs/nu research/ERRATUM_arXiv_2606.09686_catalog_count.md: names `gHashTag/t27/specs/numeric/formats_catalog.t27` research/ERRATUM_arXiv_2606.09686_catalog_count.md: names `gHashTag/t27/specs/numeric/formats_catalog.t27` research/GFT16_BEATS_TEKUM16_2026-08-05.md: names `gHashTag/t27/specs/numeric/gft16.t27` +research/GF_INVENTORY_2026-08-08.md: names `gHashTag/arith2027-goldenfloat/PLAN.md` +research/GF_INVENTORY_2026-08-08.md: names `gHashTag/goldenfloat-preprint/gf_preprint_v19.tex` +research/GF_INVENTORY_2026-08-08.md: names `gHashTag/goldenfloat-preprint/gf_preprint_v19.tex` +research/GF_INVENTORY_2026-08-08.md: names `gHashTag/paper3-methodology/main.tex` research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `gHashTag/t27/specs/numeric/gft16.t27` research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `gHashTag/t27/specs/numeric/gft32.t27` research/GF_T_GOLD_STANDARD_LADDER_2026-08-05.md: names `gHashTag/t27/specs/numeric/gft4.t27` diff --git a/tools/orphan_artefacts_baseline.txt b/tools/orphan_artefacts_baseline.txt index ff3ebfcf5b..c3697e4b16 100644 --- a/tools/orphan_artefacts_baseline.txt +++ b/tools/orphan_artefacts_baseline.txt @@ -2,5 +2,7 @@ research/8BIT_LUT_COMPARISON.json research/GPU_BENCHMARK_RESULTS.json research/H100_FORMAT_COMPARISON.json research/H100_FORMAT_COMPARISON_OFFICIAL_BPB.json +research/arxiv_gft/competition.json +research/arxiv_gft/gft_competition.pdf research/block/kurtosis_all.json research/unified_fineweb_benchmark.json From dbe7885ba20954c6e8d5ab937906c9f7b2a7be84 Mon Sep 17 00:00:00 2001 From: dmitrii-f-t27 Date: Sat, 10 Oct 2026 21:32:44 -0300 Subject: [PATCH 2/2] docs(agents): agents.json is a format name, not a path -- doc-refs after gHashTag/t27#8054 AGENTS.md and CLAUDE.md list, in one sentence, the file formats outside tools expect (`llms.txt`, agents.json, `ai.txt`, A2A agent cards, ...). The name agents.json was in backticks, which check_doc_refs.py reads as a path. It "resolved" only because gHashTag/t27 had an unrelated .trinity/seals/agents.json with the same basename, so it sat in tools/doc_refs_crossrepo.txt as a cross-repo reference. gHashTag/t27#8054 deleted that seal on 2026-10-09 and doc-refs went red on every PR here. The sentence never referred to that file. The markup now matches the meaning, and the two stale cross-repo rows are removed. No checker code and no baseline entry change. Closes #908 --- AGENTS.md | 2 +- CLAUDE.md | 2 +- tools/doc_refs_crossrepo.txt | 2 -- 3 files changed, 2 insertions(+), 4 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index da81a86330..105e95ed75 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -346,7 +346,7 @@ Owner's rule, 2026-09-20: stop writing in other people's languages, we have our own. This bites on any file whose only reason to exist is that an outside tool -expects that shape: `llms.txt`, `agents.json`, `ai.txt`, `.well-known/*.json`, +expects that shape: `llms.txt`, agents.json, `ai.txt`, `.well-known/*.json`, A2A agent cards, `ai-plugin` manifests, OpenAPI stubs, JSON-LD blocks, a README that restates a spec. The reflex is to write four of them in four foreign formats, and the reflex is wrong: a project whose claim is "here is a language diff --git a/CLAUDE.md b/CLAUDE.md index a2eefeb3f5..c3af866802 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -301,7 +301,7 @@ Owner's rule, 2026-09-20: stop writing in other people's languages, we have our own. This bites on any file whose only reason to exist is that an outside tool -expects that shape: `llms.txt`, `agents.json`, `ai.txt`, `.well-known/*.json`, +expects that shape: `llms.txt`, agents.json, `ai.txt`, `.well-known/*.json`, A2A agent cards, `ai-plugin` manifests, OpenAPI stubs, JSON-LD blocks, a README that restates a spec. The reflex is to write four of them in four foreign formats, and the reflex is wrong: a project whose claim is "here is a language diff --git a/tools/doc_refs_crossrepo.txt b/tools/doc_refs_crossrepo.txt index 9f803e1bad..32b907547d 100644 --- a/tools/doc_refs_crossrepo.txt +++ b/tools/doc_refs_crossrepo.txt @@ -1,5 +1,3 @@ -AGENTS.md: names `agents.json` -> resolves in t27 -CLAUDE.md: names `agents.json` -> resolves in t27 docs/REFEREE-PAGE.md: names `gft_sadd.t27` -> resolves in t27 docs/REFEREE-PAGE.md: names `gft_smul.t27` -> resolves in t27 docs/docs/tri27/index.md: names `hello.t27` -> resolves in tri-net