[OffgridwithJD]

#1341, the twin-less suites: native_decode_gate_width.sh (#595) had no pytest twin. This PR adds one, every check by the shell's name, and seeds the suite into the ledger. With it, all four native_d* suites that lacked a twin have one: native_dml (#1454), native_dict_underfill (#1455), native_decode_gating (#1456) and this.

Stacked on #1456, at the top of my chain. This branch is #1456's head plus two commits.

What changes

• test/pytest/test_native_decode_gate_width.py, the twin: one test in a private schema. It holds the threshold per session; the shell sets it per database, because every helper opens a new connection. compare_to_bash.py grades it as covering all 35 names, and it joins COMPLETE.
• The two "not shipped" premises are now proven by SQLSTATE. They compared SHOW's output for the rejected probe settings with the empty string. A dead connection returns empty too, and the pytest layer refuses an empty expectation outright. Both twins now require SHOW to raise 42704 (undefined_object).
• Otherwise the shell needed no change. It pins the policy by moving the threshold across a fixed projection, including the >= boundary, and each width-gate mutation below reddens the arm it names.
• Ledger: native_decode_gate_width is seeded with 35 rows from five-major runs, and 3 of the 35 rows dated. suites_not_covered goes 214 → 213, and the census is 3720 + 35 = 3755 rows, 2723 + 32 = 2755 never and 997 + 3 = 1000 dated, matching the tsv. The gate gives rc=0 against origin/main.

Red first

Each mutation of the width test in columnar_reader.c was run one at a time on PG18, through both twins, with the .so md5 printed every time.

mutation	native_decode_gate_width.sh	test_native_decode_gate_width.py	the arms that went red
none (the head)	35/35	35/35	none
gating whatever the width (the comparison replaced by true)	33/35	fails	narrow, default … NOT gated: got [32] want [96]; wide, gate raised above width … stops gating: got [32] want [800]
> for >=	34/35	fails	wide, gate AT the width (>= boundary): got [800] want [32]
the payload count one too high	34/35	fails	wide, gate raised above width: got [32] want [800]

Never red (32): the twenty-one premises, plus the arms the width policy doesn't move: the gated counts, the gate-off controls and the correctness arms. The sibling native_decode_gating (#1456) dates those.

Runs

major	native_decode_gate_width.sh	test_native_decode_gate_width.py
15	35/35	35/35
16	35/35	35/35
17	35/35	35/35
18	35/35	35/35
19	35/35	35/35

Also measured: the guard half (438 tests), harness_selftest 1245/1245 and docs_style. cluster_tests goes 676 → 677 by collection, and TESTS.md gets §149.

🤖 Generated with Claude Code

[jdatcmd]

Approving cbaa5aa.

Measured on PG17: native_decode_gate_width.sh 35/35, and test_native_decode_gate_width.py 35 checks. The tsv at this head is 3755 / 2755 / 1000, matching its budget.

The boundary mutant, run myself: >= → > in the width test at columnar_reader.c:3041. Only the boundary arm goes red, in both twins: wide, gate AT the width (>= boundary): gates again -> qual column alone, got [800] want [32].

I took the other two mutants from your table. As with #1456, the stack's restate onto 0edb5ae covers this one.

[jdatcmd]

Re-approving 0cb4e37, the restate onto main db56168. I compared every commit's patch with the approved stack using git range-diff. The test and source commits are unchanged, and the ledger-seed commits changed only by re-anchoring on #1470. The census at this head matches its budget paragraph, and the counts step consistently. At #1467's head: the guards pass (438), 687 collected (matching the file), and harness_selftest is 1245/1245.