[OffgridwithJD]

Part of #1339: the last two test_ttl_expire findings assigned to me, and one mislabel. Each fix makes the two twins ask the same question. Each is shown red under a mutation that the pre-fix twin survives, both twins run on PG17 against main (95b4112).

Findings

finding	mutation	main	this branch
pytest :511 [twin_mismatch]: "expire on a date column retires at least one group" discarded expire's return, called it again, and decided from the row count. The shell twin reads the return value.	expire returns 0 * retired (it still retires)	pytest date test green (shell red)	pytest "got 'RETIRED NOTHING (0)' want 'retired'", both twins red
shell :136 [twin_mismatch]: the no-retention arm grepped for ERROR|no retention|ttl, so any failure passed: a missing function, a non-owner, an internal error. The pytest twin requires 55000.	the refusal raised as ERRCODE_INTERNAL_ERROR	shell 67/67 (pytest red)	shell "got [XX000] want [55000]", both twins red
:659 [mislabelled]: premise: compact_rewrite retired a group checks product code, so it is an arm (#1236)	compact_rewrite rewrites nothing	red in both twins, as a premise	the same red, now as an arm

The shell twin now uses the suite's existing ttl_state helper, moved up from the negative-interval section so this arm can use it. The pytest date arm reads the first call's return value and drops the second call.

Evidence

• Clean, PG 15–19: ttl_expire 67/67 and test_ttl_expire.py 9 passed on each.
• Ledger:
  • "a table with no declared retention…" is dated by the XX000 mutation;
  • "compact_rewrite retired a group" is renamed from its premise row, with its history, and dated by the rewrite-0 mutation (the wording reused from test: #1339 findings in collation_change, tier_partition and column_extent, in both twins #1351's column_extent row);
  • rows stay at 2325, and the census goes 1802 → 1800.
• Gate: --against origin/main, including test: no premise the ledger has seen go red, and a gate rule against it (#1236) #1347's dated-premise rule, gives rc=0, the census agrees, and the ceiling stays 246 → 246.
• Database-free set: 429 passed. No tests were added or removed.

Overlap

With #1353 (same two files): git auto-merges the test files. Only the ledger and budget conflict, so whichever lands second restates. #1350, #1351 and #1354 also move the budget.

🤖 Generated with Claude Code

[jdatcmd]

Approving the content at 8edecce. On PG18 I reproduced the shell claim: with the no-retention refusal raised as ERRCODE_INTERNAL_ERROR, main's shell twin passes 67/67, and this head fails with a table with no declared retention is an error, not a silent success: got [XX000] want [55000]. The date arm now reads expire's return, as the shell does, and moving ttl_state up so both twins ask the same question is right.

Order: main is now e848f8f (#1354 merged) at 2327 / 1803, so this needs a rebase like the others. Since it shares test_ttl_expire and ttl_expire.sh with #1353, I'd land it right after #1353: #1350 → #1351 → #1353 → #1355 → my #1356. Rebase it when #1353 is in, and I'll check the ledger and merge.

[OffgridwithJD]

@jdatcmd, rebased onto 25118c0 (#1353 merged). The new head is 3d87595, replacing 8edecce. Ledger: 2343 / 1796, as you expected.

The merge.

• Test files: ttl_expire.sh and test_ttl_expire.py auto-merged with test: ttl_expire, the #1339 findings in both twins: a timestamp column, a zone that always differs, accurate premises #1353's changes.
• Ledger: check_ledger.tsv conflicted where test: ttl_expire, the #1339 findings in both twins: a timestamp column, a zone that always differs, accurate premises #1353 had renamed the neighbouring premises. I kept main's block with my one rename, then dropped my old ledger commit and regenerated the rows from fresh runs on the merged files.

Evidence, re-run on the merged tree (main 25118c0 against this head):

• Clean: ttl_expire 70/70 and test_ttl_expire.py 10 passed on each of PG 15–19.
• PG17, each mutant against main and this head:
  • expire returns 0 * retired: main's pytest date test stays green; here it is red ("RETIRED NOTHING (0)").
  • The no-retention refusal raised as XX000: main's shell is 70/70; here it is red ("got [XX000] want [55000]").
  • compact_rewrite returns 0: red in both twins, on main as a premise and here as an arm.

Checks.

• Gate: --against origin/main gives rc=0, the census agrees at 1796, and the ceiling stays 246 → 246.
• Database-free set: 429 passed.

[OffgridwithJD]

@jdatcmd, CI is green on 3d87595 (15/15). CI runs the suites on 17 and 18 only; my re-run on this merged head covers 15–19 (ttl_expire 70/70, test_ttl_expire.py 10 passed). Ready for your approval and merge. After it, #1361 is the oldest of my open ones that touches no file #1356 does, as far as I can tell. Restating is fine in any order.

[jdatcmd]

Re-approving 3d87595.

• Ledger: a recount gives 2343 / 1796, which agrees with the budget. No duplicate keys, no dated premise: row.
• Patch: the ttl_expire changes are identical to 8edecce, the head I reviewed and reproduced (41 changed lines, no difference).
• Counts, on this head: database-free set 429 passed; cluster collection matches cluster_tests.

Merging.