You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit e11d9fc
Browse filesBrowse the repository at this point in the historyBrowse files
tests: futures and forex captures move to the private fixtures (US-SCOPE-1300, part 2) (#386)
tests: fourteen engine test files and a data header of futures and forex captures move to the private fixtures; one row joins the private rows (US-SCOPE-1300, part 2)
Owner ruling 2026-10-13: recorded market data of US futures, forex and spot metals is never public. Fifteen files leave the public tree for the private
fixtures repository (byte-exact copies, commit cb5e0c2638b8 on top of da6ad27; sha256 of every copy read back): tests/test_famag_close_survives.cpp and
_l4d, test_htf_weekly_lookahead.cpp, test_trail_close_restart_no_fold.cpp and _l4c, test_tv_money_precision.cpp and _l4b, test_tv_money_carried_pooc.cpp and
_l4b, test_tv_money_band.cpp and _l4b, test_market_admission_commission.cpp and _l4b, test_ta_sma_stochrsi_ties.cpp and test_ta_sma_stochrsi_ties_data.hpp.
Their tests pin floating-point residues of recorded captures (ulp directions of tick products, ten-digit money boundaries, crossover ties), so they are moved,
not rewritten: an invented series would need its own independently justified oracle.
Registration (the #375 way): seven of the rows were already private rows gated on a private data header (famag_close_survives_l4d, htf_weekly_lookahead,
trail_close_restart_no_fold_l4c, tv_money_precision_l4b, tv_money_carried_pooc_l4b, tv_money_band_l4b, market_admission_commission_l4b): their gate and their
recorded source-bound entries stay as they are, the overlay restores the .cpp beside the header. The six base files have no row (retired through
L3A_LEGACY_OWNER_TEST_SOURCES): their sources leave and the names stay in that list. test_ta_sma_stochrsi_ties was a public, source-free row: it leaves
TEST_SOURCES and registers like test_ta_window_round10 (the real test when the overlay supplies tests/test_ta_sma_stochrsi_ties.cpp, the pinned SKIP placeholder
otherwise). Pins: tests/private_fixture_rows.txt 374 -> 375 rows (and PRIVATE_FIXTURE_ROW_COUNT), tests/private_fixture_paths.txt +15 F lines (10,221 files,
12,925 paths). Floors: release private count 306 -> 307 (absent floor 604 -> 603), kernel private count 30 -> 31 (the row is source-free; absent floor
307 -> 306); the full totals 910 and 337 are unchanged.
Doc sentence reworded, named here for the doc-revert gate: "The full-run release and kernel floors are the table's own numbers above (604 and 307 rows
absent, 910 and 337 with the overlay present); full runs do not exclude a label." now says 603 and 306 for the absent floors; the two cells of the floor
table in docs/ci.md change the same way (604 -> 603, 307 -> 306).
Gates that carried the old numbers move with the floors: scripts/test_ci_verify.py (the literal pins (910,307)/(337,31) and absent floors 603/306, the kernel
absent scenarios with 31 pinned rows, the explicit --min-tests 306 case, the shrunken release registry 900/298 -> 602 which still fails the 603 floor, the
comments of the one-row-ran and websocket cases) and tests/private_fixture_reads.evidence (the CMake lines it cites moved by -1: 3251 -> 3250, :3271 -> :3270,
:3235 -> :3234, :3247 -> :3246, :3277-3278 -> :3276-3277); a comment in tests/CMakeLists.txt that cites the two gate suites follows (:2215/:2217 -> :2214/:2216).
None of our release-built assets contains these files (the prebuilt tarballs are cmake --install with PINEFORGE_BUILD_TESTS=OFF); only GitHub's automatic source
archive carries tests/. Git history is not rewritten (owner ruling 10-11). Not run locally (no suites on the Mac); the two-world proof runs on lab remote.
docs: two pages stop citing the two admission tests that moved to the private fixtures (US-SCOPE-1300, part 2)
The first release-and-preflight run of part 2 (lab job p382p2-absent: release and kernel passed) failed the doc-lint dead-path rule at four citations of
tests/test_tv_money_precision_l4b.cpp and tests/test_tv_money_band_l4b.cpp, which part 2 moved out of the public tree. The two pages now say what those tests are instead of
naming a path that no longer exists.
Sentences reworded, named here for the doc-revert gate:
* docs/pages/native-engine.md: "(the famr / famr3 tapes of `tests/test_tv_money_precision_l4b.cpp` and `tests/test_tv_money_band_l4b.cpp`)" now reads "(the famr / famr3
tapes of two tests that live in the private fixtures)".
* docs/design/native-feature-parity.md, row SZ10a: "pinned by 445 famr3 tapes (`tests/test_tv_money_band_l4b.cpp`) and 202 famr tapes
(`tests/test_tv_money_precision_l4b.cpp`)" now reads "pinned by 445 famr3 tapes and 202 famr tapes (two tests that live in the private fixtures)".
The tape counts and every other word of both sentences are unchanged. scripts/check_doc_lint.py reports 0 offenders on this tree.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
| SZ8 | Fractional reduce of the position / a cohort (`quantize_close_units` pine_adapter.cpp:15490-15518, `compute_exit_reservation` pine_adapter.cpp:4996-5125) |**no** — sizes are `ExplicitUnits` / `OwnerOpenedUnits` / `HostSized` only native_order.hpp:60-74 **Closed:**`Reduce{ScopeFraction{fraction, claim, basis}}` takes a fraction in `(0, 1]` of the scope the reduce is bound to — the book, one opening, or a cohort roster (`tests/test_native_sizing_bases.cpp`). | K | L3 | F:A7 O:A12 O:C4 S:S5| F's "partial" counts the units-reduce of OL2. R5-4 → L3. |
87
87
| SZ9 | Exit-quantity reservation against sibling exits (`compute_exit_reservation` pine_adapter.cpp:4755-5119, `reserved_exit_qty` pine_adapter.cpp:10894-11370) |**partial** — `PointBudget` native_order.hpp:379-382 and group `Reduce` native_order.hpp:220 give a generic reservation; no net-of-siblings claim for fraction sizes **Closed:**`ScopeClaim::NetOfSiblings` subtracts the units already claimed by live sibling reduces bound to the same scope, and `ScopeBasis::{AtMatch, AtAcceptance}` chooses when that scope is measured (`tests/test_native_sizing_bases.cpp`). | K+A | L3 | F:A14 S:S5 S:O10 O:A12| F proposed nothing. R5-4: generic half → L3; the Pine ledger → SZ11, SZ12. |
| SZ10a | *(quirk, measured retention)* The money band's affordable price, `sig10(sig10(E) / notional per unit of price)` against the sizing price — at placement (`entry`, `affordable_price` pine_adapter.cpp:8341), in the paired-reversal whole drop (`close`, `affordable_price` pine_adapter.cpp:9853), at the terms boundary (`resolve_terms`, `affordable_price` pine_adapter.cpp:15999), in the precommit money scopes (`validate_precommit`, `affordable_price` pine_adapter.cpp:16530, pine_adapter.cpp:16583) and in the carried-money check (`apply_reversal_gap_bracket_policy`, `affordable_price` pine_adapter.cpp:21438) | n/a — not a units conversion. Every money-to-units conversion in the adapter is the kernel's `native_sized_units` under the source's floors since R5 lane B-ADAPTER: the default at placement and at the fill (`default_sizing_units` pine_adapter.cpp:3093), the typed per-call quantity at placement, at the fill and in the probe (`typed_quantity_units` pine_adapter.cpp:3106, `typed_entry_units` pine_adapter.cpp:3124), and the pending-order fill-quantity probe (`probe_fill_qty` pine_adapter.cpp:26423) | A | — | AUDIT3-opus2 H4 (G2 R2, R3) | **Measured:** the quotient is TradingView's whole-drop PRICE, pinned by 445 famr3 tapes (`tests/test_tv_money_band_l4b.cpp`) and 202 famr tapes (`tests/test_tv_money_precision_l4b.cpp`); on a seven-digit tie (E 9998.99999996, 10 000 lots at tick(close) 0.9999) the rule admits while the kernel's conversion buys 9999.99999996 lots and binary64 gives E/Q 0.999899999996 (`tests/test_adapter_fill_qty_probe.cpp`, section 2). |
89
+
| SZ10a | *(quirk, measured retention)* The money band's affordable price, `sig10(sig10(E) / notional per unit of price)` against the sizing price — at placement (`entry`, `affordable_price` pine_adapter.cpp:8341), in the paired-reversal whole drop (`close`, `affordable_price` pine_adapter.cpp:9853), at the terms boundary (`resolve_terms`, `affordable_price` pine_adapter.cpp:15999), in the precommit money scopes (`validate_precommit`, `affordable_price` pine_adapter.cpp:16530, pine_adapter.cpp:16583) and in the carried-money check (`apply_reversal_gap_bracket_policy`, `affordable_price` pine_adapter.cpp:21438) | n/a — not a units conversion. Every money-to-units conversion in the adapter is the kernel's `native_sized_units` under the source's floors since R5 lane B-ADAPTER: the default at placement and at the fill (`default_sizing_units` pine_adapter.cpp:3093), the typed per-call quantity at placement, at the fill and in the probe (`typed_quantity_units` pine_adapter.cpp:3106, `typed_entry_units` pine_adapter.cpp:3124), and the pending-order fill-quantity probe (`probe_fill_qty` pine_adapter.cpp:26423) | A | — | AUDIT3-opus2 H4 (G2 R2, R3) | **Measured:** the quotient is TradingView's whole-drop PRICE, pinned by 445 famr3 tapes and 202 famr tapes (two tests that live in the private fixtures); on a seven-digit tie (E 9998.99999996, 10 000 lots at tick(close) 0.9999) the rule admits while the kernel's conversion buys 9999.99999996 lots and binary64 gives E/Q 0.999899999996 (`tests/test_adapter_fill_qty_probe.cpp`, section 2). |
90
90
| SZ11 |*(quirk)* Same-bar `strategy.close` callsite batching, two-call provenance, entry-id ledger (`enqueue_pooc_fifo_close` pine_adapter.cpp:8743-8982; ledger `close_batch_bar_` pine_adapter.hpp:2587-2593; the kernel's comment-only residue was deleted by R5 lanes E6 and E10) | n/a | A | — | S:O8 F:A15 O:A15|**Disagree:** S "generic → kernel" (its P2 `ReplaceByKey` / `Fifo`). R5-4: adapter. |
91
91
| SZ12 |*(quirk)* POOC reservation-growth market-entry population predicate (`select_reservation_growth_sources` src/compat/pine/reservation_expansion.cpp:9-34, called from the adapter, `select_reservation_growth_sources` pine_adapter.cpp:1263) | n/a | A | — | S:O10| (single-source) R5-4 confirms S's own split: concept → SZ9, predicate A. |
Copy file name to clipboardExpand all lines: scripts/ci_verify.py
+3-3Lines changed: 3 additions & 3 deletions
Original file line number
Diff line number
Diff line change
@@ -750,7 +750,7 @@
750
750
# at ci_verify.py:321 and ci_verify.py:595 do not move; the campaign's own
751
751
# gate is PRIVATE_FIXTURE_PROFILE_FLOORS below, the one place its numbers
752
752
# live. The class-fix round added the three remaining split parents (test_coof_close_slippage_tapes, test_cross_mechanism_tapes, test_margin_rules_forward_replay): each reads its fixture data at run time, the private-fixture removals take the whole directory it reads (the removed file beside distributed ones in PR 1, the remaining files in PR 1b), and the row had stayed unconditionally registered -- the fixture-less tree would have run it against absent data. Each now pins whole beside its *_private half on the same EXISTS guard (tests/CMakeLists.txt), exactly as test_source_buying_power_tapes already did at 324; a two-world registration sweep over tests/CMakeLists.txt and runner/CMakeLists.txt closes at `duplicate names: 0` in both worlds. test_callback_lifecycle_tapes joined at 123: its tapes.txt is a distributed list the private-fixture PR leaves in the tree, but 60 of the cases it names are removed, so tests/CMakeLists.txt registers the row as a placeholder while the listed cases are incomplete. One of the 324, test_run_json_outputs, is registered only by the release-shaped profiles (the source layer gates its block); the observed-skip lowering counts it only in the profiles it registers in, so the kernel profile is no longer lowered one row deeper than the rows it skips. Six more joined at 129 (the fifth proof-box round): test_native_oracle_day_key_l2 and test_native_oracle_reversal_same_bar_tx_l2 (their twin source #includes oracle/test_oracle_*.cpp), test_adapter_in_place_reads, test_adapter_runtime_ambient and test_adapter_settlement_carry (tests/pine_d2c_scenarios_fixture.hpp), and test_coof_forced_waypoint_tapes (tests/fixtures/coof_forced_waypoint/cases.inc) -- each target now exists only while its removed include does, so the compile that killed the fifth configure-through run cannot recur. The PR 1b batch moved the remaining 6,242 fixture/probe files (l50/pr1b-inventory.tsv) and took the count to 319: 134 rows in tests/CMakeLists.txt (each a row whose source .cpp, quoted #include or runtime fixture the removal takes, or one of five fixture-fed targets) and 53 in runner/CMakeLists.txt (the strategy-capabilities and confirmed-capabilities e2e rows), and 3 rows whose fixture data the sources read at run time through their fixture-dir defines (test_color_tapes, test_session_transition_clock_tapes, test_source_buying_power_tapes); PR 1's fix round then pinned five rows of its own that PR 1b's 319 does not carry -- test_carried_trail_open_tapes, pinned whole, and the *_private halves of test_coof_close_slippage_tapes, test_cross_mechanism_tapes, test_source_buying_power_tapes and test_margin_rules_forward_replay (each parent row keeps replaying the distributed part, the *_private row beside it runs the same binary with --private-only) -- and the union of the two pin sets, sorted and de-duplicated, is 324; the pinned file header still names the PR 1 manifest sha. The stream-security round took it to 361: tests/CMakeLists.txt registered the fifteen test_stream_security_<case> targets, from fixtures/stream_security/<case>/generated.cpp, inside an unguarded bare-form foreach -- the shape the box's run 1 died on at configure (add_executable could not find the removed source), and the shape the four class sweeps' line walker did not expand -- so each case now sits on its own if(EXISTS ...) and, absent, registers its two rows plus the four test_stream_security_control_* rows as pinned SKIP placeholders: 327 + 34 = 361. The kernel-python round (L6's S-73 two-mode round 4, class 4) took it to 364: three Python rows the kernel profile runs outside the placeholder loop now skip whole when their private input is gone instead of failing on read_text() -- test_native_cpp_abi (scripts/check_native_cpp_abi.py gates on tests/fixtures/native_cpp_abi/), test_pending_order_v1_full_prefix and test_pending_order_v1_compiler_prefix (the two check_pending_order_prefix* scripts gate on tests/fixtures/script_cpp_abi/basev8 and basev9) -- each printing the pinned literal and exiting 77 under SKIP_RETURN_CODE 77, each carrying LABELS private_fixtures so the pinned-vs-labelled self-test still closes at zero. The same round replaced the static lowering (floor minus PRIVATE_FIXTURE_ROW_COUNT, which read -36 for the kernel floor) with an observed-skip lowering, bounded then by a floor-minimum constant of 1; PR 1b's second pass (L5 01:43's ruling) replaced that bound with a per-profile margin. The fix round (R2-F1) replaced that whole machinery with PRIVATE_FIXTURE_PROFILE_FLOORS: the observed-skip lowering moved with the registry it was meant to check (release absent ran 603 against a floor of 602 after two of main's rows had already left the registry, and up to 99 kernel rows could vanish with the floor still met), and no margin repairs a floor that follows the run down. PR 1b's first pass, L5 01:37, took it back down to 347: it had also guarded the 17 legacy-owner base names (L3A_LEGACY_OWNER_TEST_SOURCES intersected with the private rows), which main never compiles -- only their _l4a/_l4b/_l4c/_l4d twins run -- and the fix dropped those guard blocks and moved this count and the pinned file together (364 -> 347), restoring main's registry. PR 1b's second pass (L5 01:43, class 3) took it to 351: the four receipt-gated rows in tests/CMakeLists.txt are now registered on an EXISTS guard over the seven ABI-provider manifests their providers gate on, so absent manifests make each the pinned SKIP placeholder (the literal, LABELS private_fixtures, SKIP_RETURN_CODE 77) and present manifests leave the registration exactly as it was -- check_abi_receipt_skips no longer counts them as failures whose input is simply absent, the preparer's own 77+literal is a skip rather than a failed preparation (ensure_prepared_provider accepts it), and the four skips are pinned accounting like every other private-fixture row's. The class-sweep round (the private-fixtures lane's iteration 2) took it to 354: scripts/check_private_fixture_reads.py found three rows (test_coof_marketable_stop_tapes, test_pooc_close_bar_fills_tapes, test_pooc_pending_stop_tapes) still registered by the TEST_SOURCES loop while every case each replays sits under a removed directory -- they ran on the fixture-less tree and aborted -- and each now pins whole on a removed file it reads, with the kernel-absent branch its source-bound TU needs.
0 commit comments