fix(ci): parse gh api --paginate output as a single JSON document

str.splitlines() also splits on Unicode line separators (U+2028, U+2029)
which are valid inside JSON string values without escaping. When an open
GitHub issue title or body contained one of these characters, the
duplicate-issue check script broke the gh output mid-string and crashed
with 'Unterminated string ...' before the auto-close path could run.

gh api --paginate for an array endpoint already concatenates pages into
a single JSON array, so just parse the whole stdout once. Add a
regression test covering U+2028/U+2029 and the existing PR-filter path.

Co-authored-by: Krrish Dholakia <krrish-berri-2@users.noreply.github.com>

.github/scripts/close_duplicate_issues.py

@@ -49,20 +49,13 @@ def fetch_open_issues(repo: str | None) -> list[dict]:
         endpoint = "repos/{owner}/{repo}/issues?state=open&per_page=100&sort=created&direction=asc"
     cmd = ["api", "--paginate", endpoint]
 
-    raw = gh(*cmd)
-    # gh --paginate concatenates JSON arrays, so we may get multiple arrays
-    issues = []
-    for line in raw.strip().splitlines():
-        line = line.strip()
-        if not line:
-            continue
-        parsed = json.loads(line)
-        if isinstance(parsed, list):
-            issues.extend(parsed)
-        else:
-            issues.append(parsed)
+    raw = gh(*cmd).strip()
+    if not raw:
+        return []
+
+    parsed = json.loads(raw)
+    issues = parsed if isinstance(parsed, list) else [parsed]
 
-    # Filter out pull requests (they also appear in the issues endpoint)
     return [i for i in issues if "pull_request" not in i]
 
 

.github/scripts/test_close_duplicate_issues.py

@@ -0,0 +1,89 @@
+"""Tests for the duplicate-issue close script.
+
+Run with: python -m pytest .github/scripts/test_close_duplicate_issues.py
+"""
+
+import json
+import os
+import sys
+import types
+from unittest.mock import patch
+
+import pytest
+
+sys.path.insert(0, os.path.dirname(__file__))
+
+import close_duplicate_issues as mod
+
+
+def _make_completed_process(stdout: str):
+    return types.SimpleNamespace(stdout=stdout, returncode=0, stderr="")
+
+
+def test_fetch_open_issues_handles_unicode_line_separator():
+    """Regression: U+2028 / U+2029 inside JSON strings must not break parsing.
+
+    Python's str.splitlines() treats these characters as line breaks, but they
+    are valid inside JSON string values without escaping, so any line-based
+    parsing of gh's compact JSON output would split mid-string.
+    """
+    issues = [
+        {
+            "number": 1,
+            "title": "title with line\u2028separator",
+            "body": "paragraph\u2029separator and line\u2028break",
+        },
+        {"number": 2, "title": "plain", "body": "plain"},
+    ]
+    raw = json.dumps(issues, ensure_ascii=False)
+    assert "\u2028" in raw and "\u2029" in raw
+
+    with patch.object(mod.subprocess, "run", return_value=_make_completed_process(raw)):
+        result = mod.fetch_open_issues("owner/repo")
+
+    assert [i["number"] for i in result] == [1, 2]
+    assert result[0]["title"] == "title with line\u2028separator"
+
+
+def test_fetch_open_issues_filters_pull_requests():
+    payload = [
+        {"number": 1, "title": "real issue"},
+        {"number": 2, "title": "a PR", "pull_request": {"url": "..."}},
+    ]
+    raw = json.dumps(payload)
+
+    with patch.object(mod.subprocess, "run", return_value=_make_completed_process(raw)):
+        result = mod.fetch_open_issues("owner/repo")
+
+    assert [i["number"] for i in result] == [1]
+
+
+def test_fetch_open_issues_empty_response():
+    with patch.object(mod.subprocess, "run", return_value=_make_completed_process("")):
+        assert mod.fetch_open_issues("owner/repo") == []
+
+    with patch.object(
+        mod.subprocess, "run", return_value=_make_completed_process("[]")
+    ):
+        assert mod.fetch_open_issues("owner/repo") == []
+
+
+def test_find_duplicate_normalizes_prefixes_and_case():
+    target = {"number": 10, "title": "[Bug] Crash on startup"}
+    older = [
+        {"number": 1, "title": "feature request: add streaming"},
+        {"number": 2, "title": "crash on startup"},
+    ]
+    dup = mod.find_duplicate(target, older, threshold=0.85)
+    assert dup is not None
+    assert dup["number"] == 2
+
+
+def test_find_duplicate_returns_none_below_threshold():
+    target = {"number": 10, "title": "Streaming hangs on Anthropic"}
+    older = [{"number": 1, "title": "Completely unrelated topic"}]
+    assert mod.find_duplicate(target, older, threshold=0.85) is None
+
+
+if __name__ == "__main__":
+    raise SystemExit(pytest.main([__file__, "-v"]))