Withdraw samorzad_gov_pl: operator footer, not the authors' grant
Withdraws samorzad_gov_pl from main, so it is not in v0.2.6 (slayerlabs/polish-dynaword-v2#49, decision of 2026-10-17).
Why. The CC BY-SA 4.0 licence comes from the platform operator's (KPRM) shared footer, but the texts are written by 247 municipalities, counties, schools and other institutions. The operator's agreements with them are unpublished, no grant from them was found, and KPRM has not answered the question sent on 2026-10-06. Our assessment is that the footer alone does not license the institutions' own texts. Dropping slices cannot fix this, because it applies to the whole source.
What changes
data/samorzad_gov_pl/is removed: 72,328 docs / 39,890,244 tokens on main (72,771 / 41,154,506 in the v0.2.5 tag, which stays as published). Its.gitattributesLFS line goes with it.src/sources.pykeeps the entry (release: "0.2.5"), with a "Withdrawn from main on 2026-10-17"legal_note.src/clean_samorzad_gov_pl.pyand its rule tests stay for re-admission (v2#103). The release-artifact test is replaced bytest_withdrawn_from_main.artifacts/source_findings.md: new section "samorzad_gov_pl — withdrawn (2026-10-17)".
Not in this PR. Per AGENTS.md ("Keep the PR to your source"), the README row, the header totals and CHANGELOG go in the release commit (v2#80). make_docs skips a source without stats, so a regenerated README drops it.
Tests. Full suite vs main at c7fb112: 99 failed / 854 passed / 18 errors, against 101 / 856 / 18. The two failures that go away are test_release_artifact_matches_stats_and_has_institution_authors and test_added_dates::test_shard_carries_the_registry_date[samorzad_gov_pl]. Three passing per-directory parametrisations go with the directory (test_card_states_the_registry_date, test_stats_state_the_registry_date, test_cited_paths), and test_withdrawn_from_main passes. No other test changes outcome.
🤖 Generated with Claude Code
refs/pr/123 ref