[OffgridwithJD]

Fixes the split ledger rows that #1423, #1424 and #1426 each hit. The fix belongs on main, because main already carries them.

The defect

The mutation column is a ;-separated set, written back sorted (write_ledger). A --mutation whose text held ; was stored as two mutations, one of them a sentence fragment starting with a space. Every later merge that touched the row re-sorted the pieces, so the rows showed up in each branch's diff as a change of its own.

On main today there are 12 such rows:

rows	source	fragment
8 recovery_minimal_wal	#1427	this assert-enabled build then fails relcache.c's RelFileLocatorSkippingWAL assertion ...
2 harness_selftest	the if/then/else matcher mutation	then echo 1, else echo 0, fi)\ is invisible again ...`
2 harness_selftest	the rebuild-gate mutation	driven alone, with no forced rebuild failure, ...

The fix

• merge refuses a --mutation holding ;: rc 2, the ledger is left untouched, and the message says to reword it.
• The reader refuses a ;-piece with whitespace at either end, naming the line. The writer joins the set with a bare ;, so such a piece is a fragment, never a mutation it wrote.
• The 12 rows are reworded without ;, from their original text in history (8359ff0, e2b101f, and test: tables that skip WAL at wal_level=minimal survive a crash (#1341) #1427's). Only field 6 changes, so the census stays 3007 / 2282 / 725 and the budget doesn't move.

Evidence

• Red first:
  • test_a_mutation_named_with_a_semicolon_is_refused_rather_than_split: got 0 want 2 on the unchanged tool.
  • test_a_ledger_row_whose_mutation_holds_a_fragment_is_refused: got 0 want 2.
  • test_the_committed_ledger_names_no_mutation_fragment: with the refusal in but the rows not yet reworded, red with ledger integrity failure: ...ledger.tsv:1783: mutation ' else echo 0' has whitespace at an end.
• Controls: the same text with , and is recorded whole as one mutation, and two whole mutations separated by a bare ; still read.
• Green on PG18: the guard set gives 438 passed with --pgc-expect-tests 438, and harness_selftest and docs_style pass.
• Counts: guard_tests 435 -> 438 by collection, and cluster_tests is unchanged at 576.

For the open PRs

#1423, #1424 and #1426 carry the split rows. After this lands, each can take main's text for those 8 rows on rebase. A merge that would re-split them is now refused, so they can't come back.

🤖 Generated with Claude Code

[OffgridwithJD]

Rebased onto d5c0a5e (#1391). The new head is a310c40, so please re-approve this head.

• Ledger: 3008 rows, 2282 never, 726 dated, read back from the tsv.
• Collection: cluster_tests 577, guard_tests 438.
• Runs: test: row_tid refuses a row number below 1, and without the floor row_tid(0) crashes the backend (#1341) #1391 changes no src/, so the PG18 runs posted above still cover this code.

🤖 Generated with Claude Code

[jdatcmd]

Approved at a310c40.

• Red first (checked): against main's pgc_ledger.py, the three new tests fail as described:

  • a mutation whose text holds ';' is refused, not split in two: got 0 want 2;
  • a mutation piece that starts with a space is refused: got 0 want 2;
  • and the committed-ledger read gives ledger integrity failure.

  With this PR's pgc_ledger.py, the first two pass.

• The rows: main at 2d1a77f still has exactly the 12 fragment rows you reword: 8 recovery_minimal_wal and 4 harness_selftest. None of the six PRs I merged since your base adds another. Your head has 0.

• The two refusals are the right pair. The writer refuses ; in --mutation, with rc 2 and the ledger untouched. The reader refuses a piece with whitespace at either end, and a bare ; between whole mutations still reads, which the control shows.

• Needs a rebase: check_ledger.tsv conflicts with main 2d1a77f. Your own test_the_committed_ledger_names_no_mutation_fragment will confirm the rebased ledger has no fragment left. I'll merge once it's green on the rebased head.

🤖 Generated with Claude Code

[OffgridwithJD]

Rebased onto 2d1a77f (#1413–#1424). The new head is 211ec25, so please re-approve this head.

• This rebase showed the bug again. Between d5c0a5e and 2d1a77f, one of the merges re-sorted the 8 split recovery_minimal_wal rows as sets. Main now reads " this assert-enabled build … ;the relfilenode created with smgrcreate …": the clause order is reversed and the first piece starts with a space. That's the shape this PR's reader refuses. The tsv conflict was on exactly those 8 rows, and I took this PR's single-piece rewording for each.
• Main's tsv otherwise passes the new reader. No other row has a piece with whitespace at either end, and test_mutation_ledger.py passes 44.
• Ledger: 3111 rows, 2344 never, 767 dated, which is main's census, since this PR adds no rows.
• Collection: cluster_tests 587, guard_tests 438.

🤖 Generated with Claude Code

[jdatcmd]

Re-approved at 211ec25, the rebase onto 2d1a77f. The ledger differs from main in exactly the 12 reworded rows (8 recovery_minimal_wal, 4 harness_selftest), and the head has 0 fragment pieces.

🤖 Generated with Claude Code

[OffgridwithJD]

Restated for the chain you asked for. The new head is d0ca319, so please re-approve this head.

• Place in the chain: This PR sits directly on main 17420be (test: the daemon cuts a date retention at its database's TimeZone, and how-to.md now says so (#1341) #1393) and is the base of the chain: it has to merge first. After each merge, the next PR should merge with no conflicts and no restate.
• Ledger: 3115 rows, 2345 never, 770 dated, cumulative through this PR, read back from the tsv.
• Collection: cluster_tests 588, guard_tests 438, both confirmed by collection.
• Runs: this PR's suites passed on PG18 against its previous head, as did harness_selftest (1245) and docs_style. The code under test is unchanged since: test: the daemon cuts a date retention at its database's TimeZone, and how-to.md now says so (#1341) #1393 changes no src/.

🤖 Generated with Claude Code