emrg: a release issue is finished by its artifact, not by a merge

scripts/check-issue-links.py

@@ -1399,7 +1399,25 @@ def _mention_detail(refs: Refs, number: int) -> str:
     return "; ".join(clauses)
 
 
-def _declined_detail(refs: Refs, number: int) -> str:
+#: A release version an issue's own text names, e.g. `Release v0.3.46`. It is read
+#: because it is the one fact an open issue carries about what finishes *it*: a PR's own
+#: title never reaches `judge_issues`, so "was the declaring PR a bump?" cannot be answered
+#: here at all, while "does this issue's completion name a published artifact?" can.
+_RELEASE_VERSION = re.compile(r"\bv\d+\.\d+\.\d+\b")
+
+
+def release_version(*texts: str | None) -> str | None:
+    """The release version this issue's own text names, or `None` when it names none."""
+    for text in texts:
+        if not text:
+            continue
+        found = _RELEASE_VERSION.search(text)
+        if found:
+            return found.group(0)
+    return None
+
+
+def _declined_detail(refs: Refs, number: int, release: str | None = None) -> str:
     """What to do about declarers that closed with no open one left.
 
     Two facts, one sentence each, because the remedies are opposite: a **merged**
@@ -1435,6 +1453,19 @@ def _declined_detail(refs: Refs, number: int) -> str:
                 "finish the work: do not close this issue on that merge, state here what "
                 "it still leaves"
             )
+        elif release is not None:
+            clauses.append(
+                f"{_numbers(refs.declared_merged)} declared `Closes #{number}` and {verb} "
+                f"merged, and what finishes this issue is the published `{release}`, not "
+                "that merge - a declaring pull request can move the version string while "
+                "nothing is released, so the merge is not the reading that settles it. "
+                f"`uv run --no-sync python3 scripts/check-release-published.py {release}` "
+                "is the reading: green across the platform matrix is what says the work "
+                "is done, and this issue may be "
+                "closed on that reading; anything else, or no answer at all, means it is "
+                "not done - do not close this issue on the merge, state here what it "
+                "still leaves"
+            )
         else:
             clauses.append(
                 f"{_numbers(refs.declared_merged)} declared `Closes #{number}` and {verb} "
@@ -1490,7 +1521,9 @@ def judge_issues(
                 )
                 continue
             if refs.declared_merged or refs.declared_abandoned:
-                detail = _declined_detail(refs, number)
+                detail = _declined_detail(
+                    refs, number, release_version(issue.get("title"), issue.get("body"))
+                )
             elif refs.mentioned_landed or refs.mentioned_abandoned or refs.mentioned_open:
                 detail = _mention_detail(refs, number)
             else:

tests/test_check_issue_links.py

@@ -2803,3 +2803,54 @@ def test_the_ledger_path_is_overridable(mod, monkeypatch):
     monkeypatch.setenv("EMRG_RANTS", "/tmp/from-env.jsonl")
     assert mod.rants_path() == Path("/tmp/from-env.jsonl")
     assert mod.rants_path("/tmp/wins.jsonl") == Path("/tmp/wins.jsonl")
+
+
+# --- an issue finished by a published artifact, not by a merge ---------------------
+#
+# A release issue is the case where "a declaring PR merged" and "the work is done" come
+# apart: the bump PR carries the closing keyword, so its merge closes the issue while
+# nothing has been tagged or published. The row then offered "the reading that says the
+# work is done" for work that had not begun. The fix reads the one fact the issue itself
+# carries - the release version it names - and sends the reader to that artifact's own
+# reading instead of inferring completion from the merge.
+
+
+def test_a_release_issue_is_not_told_to_close_on_its_bump_merge(mod) -> None:
+    detail = mod._declined_detail(mod.Refs(declared_merged={2782}), 2781, "v0.3.46")
+    assert "scripts/check-release-published.py v0.3.46" in detail
+    assert "either close this issue with the reading that says the work is done" not in detail
+
+
+def test_an_ordinary_issue_keeps_the_remedy_it_had(mod) -> None:
+    detail = mod._declined_detail(mod.Refs(declared_merged={2779}), 2778)
+    assert "either close this issue with the reading that says the work is done" in detail
+    assert "check-release-published" not in detail
+
+
+def test_the_artifact_is_what_changes_the_sentence_not_the_branch(mod) -> None:
+    refs = mod.Refs(declared_merged={2782})
+    withheld = mod._declined_detail(refs, 2781)
+    assert withheld == mod._declined_detail(refs, 2781, None)
+    assert withheld != mod._declined_detail(refs, 2781, "v0.3.46")
+
+
+def test_the_release_version_is_read_only_where_it_is_one(mod) -> None:
+    assert mod.release_version("Release v0.3.46") == "v0.3.46"
+    assert mod.release_version("an ordinary title") is None
+    assert mod.release_version(None, "") is None
+    assert mod.release_version("0.3.46") is None
+    assert mod.release_version("xv0.3.46y") is None
+
+
+def test_judge_issues_reads_the_version_off_the_issue_it_is_judging(mod) -> None:
+    issues = [
+        {"number": 2781, "title": "Release v0.3.46", "created_at": "2026-11-09T17:35:22Z"},
+        {"number": 2778, "title": "A change that needs no release", "created_at": "2026-11-09T15:49:24Z"},
+    ]
+    refs = {
+        2781: mod.Refs(declared_merged={2782}),
+        2778: mod.Refs(declared_merged={2779}),
+    }
+    rows = {row.number: row for row in mod.judge_issues(issues, refs, {})}
+    assert "check-release-published.py v0.3.46" in rows[2781].detail
+    assert "check-release-published" not in rows[2778].detail

tests/test_printed_commands_carry_the_python_runner.py

@@ -73,10 +73,11 @@
 what keeps the clause off prose: "update `VERSION_SOURCES` in `scripts/bump-version.py`"
 ends at the file name and is not a thing to run, while `scripts/check-vote-count.py <PR>` is.
 
-What the clause reads on this tree, measured 2026-10-24: **39** command-shaped mentions, in
-eight tools - `bump-version.py` 2, `cast-vote.py` 8, `check-merge-freshness.py` 8,
-`check-merge-landed.py` 1, `check-merge-sequence.py` 1, `check-release-published.py` 1,
-`check-vote-count.py` 1, `review-queue.py` 17. This number has gone down as well as up, and
+What the clause reads on this tree, measured 2026-11-10: **40** command-shaped mentions, in
+nine tools - `bump-version.py` 2, `cast-vote.py` 8, `check-issue-links.py` 1,
+`check-merge-freshness.py` 8, `check-merge-landed.py` 1, `check-merge-sequence.py` 1,
+`check-release-published.py` 1, `check-vote-count.py` 1, `review-queue.py` 17. This number
+has gone down as well as up, and
 it goes down the way it went up: a *printed remedy* is a printed command either way, so a
 remedy deleted together with the tool that printed it takes its mention with it.
 `review-queue.py` carried a second `abstain` branch - a head pushed below a window whose