Merge pull request #4950 from MarcZX100/codex/pr-overlap-compatible-merge-base

fix(tools): use the shared merge base for overlap checks

tools/pr_overlap.py

@@ -45,7 +45,24 @@ def open_prs(base):
     query = urlencode({"state": "open", "base": base, "per_page": 100})
     out = gh("api", "--paginate", f"repos/{REPO}/pulls?{query}")
     prs = {}
-    for pr in json.loads(out):
+    decoder = json.JSONDecoder()
+    offset = 0
+    page_count = 0
+    pull_requests = []
+    while offset < len(out):
+        while offset < len(out) and out[offset].isspace():
+            offset += 1
+        if offset == len(out):
+            break
+        page, offset = decoder.raw_decode(out, offset)
+        if not isinstance(page, list):
+            raise ValueError("gh api --paginate returned a non-array page")
+        page_count += 1
+        pull_requests.extend(page)
+    if page_count == 0:
+        raise ValueError("gh api --paginate returned no JSON pages")
+
+    for pr in pull_requests:
         prs[pr["number"]] = {
             "ref": f"refs/pr/{pr['number']}",
             "sha": pr["head"]["sha"],
@@ -127,7 +144,7 @@ def clashes(a, b, path):
 def conflicts(a, b):
     if not (a["tree"] and b["tree"] and a["files"] & b["files"]):
         return {}
-    files = merge("--merge-base=refs/pr/base", a["tree"], b["tree"])[1]
+    files = merge(a["tree"], b["tree"])[1]
     return {path: [] if path in a["added"] else clashes(a, b, path) for path in files}
 
 

tools/tests/test_pr_overlap.py

@@ -54,6 +54,28 @@ class OpenPullRequestsTests(unittest.TestCase):
             7: {"ref": "refs/pr/7", "sha": "head-sha", "author": "contributor", "depends": set()},
         })
 
+    def test_collects_concatenated_json_pages_from_paginated_gh_output(self):
+        pull_requests = [
+            {"number": number, "user": {"login": f"author{number}"},
+             "head": {"sha": f"sha{number}"}, "body": None}
+            for number in range(1, 4)
+        ]
+        pages = [pull_requests[:2], pull_requests[2:]]
+        output = "\n".join(json.dumps(page) for page in pages)
+
+        with patch.object(pr_overlap, "REPO", "owner/project", create=True), \
+                patch.object(pr_overlap, "gh", return_value=output):
+            found = pr_overlap.open_prs("main")
+
+        self.assertEqual(set(found), {1, 2, 3})
+        self.assertEqual(found[3]["sha"], "sha3")
+
+    def test_rejects_empty_paginated_output(self):
+        with patch.object(pr_overlap, "REPO", "owner/project", create=True), \
+                patch.object(pr_overlap, "gh", return_value="  \n"):
+            with self.assertRaisesRegex(ValueError, "returned no JSON pages"):
+                pr_overlap.open_prs("main")
+
 
 class FetchTests(unittest.TestCase):
     def test_fetches_each_head_commit_the_api_reported(self):
@@ -86,5 +108,14 @@ class FetchTests(unittest.TestCase):
         self.assertEqual(git.call_args_list[-1].args[-1], "+00000000000000000000000000000000000000f9:refs/pr/249")
 
 
+class ConflictTests(unittest.TestCase):
+    def test_uses_the_shared_base_parent_without_newer_merge_tree_option(self):
+        first = {"tree": "first-tree", "files": {"shared.cpp"}, "added": set()}
+        second = {"tree": "second-tree", "files": {"shared.cpp"}, "added": set()}
+        with patch.object(pr_overlap, "merge", return_value=("merged-tree", set())) as merge:
+            self.assertEqual(pr_overlap.conflicts(first, second), {})
+        merge.assert_called_once_with("first-tree", "second-tree")
+
+
 if __name__ == "__main__":
     unittest.main()