From ecfe8157bbde08cb8a07f4d9639e19af0fae95c8 Mon Sep 17 00:00:00 2001
From: Abdessamad Derraz <3028866+Abdess@users.noreply.github.com>
Date: Wed, 11 Nov 2026 10:27:53 +0100
Subject: [PATCH] fix: keep .mdx pages through the sweep

A body is written as .md and the decoration pass, after the sweep,
gives a page with components its .mdx name. The sweep did not count the
.mdx of the run before as produced: 402 pages were deleted and written
again on every run, so write_if_changed never kept their bytes.
---
 scripts/sitewrite.py     | 20 +++++++++++++++++---
 tests/test_site_write.py | 15 +++++++++++++++
 2 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/scripts/sitewrite.py b/scripts/sitewrite.py
index c90bf78de..6a56560b6 100644
--- a/scripts/sitewrite.py
+++ b/scripts/sitewrite.py
@@ -54,6 +54,21 @@ def write_decorated(path: str | Path, content: str) -> bool:
     _record(path)
     return _write_artifact(str(path), content)
 
+# A page body is written as .md and the decoration pass, which runs after
+# the sweep, gives it its final extension: .mdx when it holds components,
+# the other file removed. The .mdx a previous run left is that same page.
+_PAGE_SUFFIXES = (".md", ".mdx")
+
+
+def _was_produced(path: Path) -> bool:
+    """Whether this run produced *path*, a page under either extension."""
+    if os.path.realpath(str(path)) in _produced:
+        return True
+    return path.suffix in _PAGE_SUFFIXES and any(
+        os.path.realpath(str(path.with_suffix(suffix))) in _produced
+        for suffix in _PAGE_SUFFIXES if suffix != path.suffix)
+
+
 def _sweep_generated(docs: Path, directories: list[str]) -> int:
     """Delete files in the generated tree that this run did not produce."""
     removed = 0
@@ -62,7 +77,7 @@ def _sweep_generated(docs: Path, directories: list[str]) -> int:
         if not root.is_dir():
             continue
         for path in sorted(root.rglob("*")):
-            if path.is_file() and os.path.realpath(str(path)) not in _produced:
+            if path.is_file() and not _was_produced(path):
                 path.unlink()
                 removed += 1
         for path in sorted(root.rglob("*"), reverse=True):
@@ -81,8 +96,7 @@ def _sweep_top_pages(docs: Path) -> int:
     """
     removed = 0
     for path in sorted(docs.glob("*.md*")):
-        if (path.is_file() and path.suffix in (".md", ".mdx")
-                and os.path.realpath(str(path)) not in _produced):
+        if path.is_file() and path.suffix in _PAGE_SUFFIXES and not _was_produced(path):
             path.unlink()
             removed += 1
     return removed
diff --git a/tests/test_site_write.py b/tests/test_site_write.py
index e10709c1f..fc8cc2e5e 100644
--- a/tests/test_site_write.py
+++ b/tests/test_site_write.py
@@ -120,6 +120,21 @@ class Sweep(unittest.TestCase):
         self.assertTrue((self.docs / "emulators/kept.md").exists())
         self.assertTrue((self.docs / "emulators/nested/deep.md").exists())
 
+    def test_a_page_the_decoration_pass_made_mdx_is_kept(self):
+        """A body is written as .md, and the decoration pass, after the
+        sweep, renames a page with components to .mdx. The .mdx of the run
+        before was swept and written again on every run: 402 pages."""
+        (self.docs / "platforms/kept.mdx").write_text("x")
+        (self.docs / "platforms/gone.mdx").write_text("x")
+        self._produce("platforms/kept.md")
+        sitewrite._sweep_generated(self.docs, ["platforms"])
+        (self.docs / "top.mdx").write_text("x")
+        sitewrite._record(self.docs / "top.md")
+        self.assertEqual(sitewrite._sweep_top_pages(self.docs), 0)
+        self.assertTrue((self.docs / "platforms/kept.mdx").exists())
+        self.assertTrue((self.docs / "top.mdx").exists())
+        self.assertFalse((self.docs / "platforms/gone.mdx").exists())
+
     def test_a_directory_outside_the_list_is_never_touched(self):
         self._produce("emulators/kept.md", "emulators/nested/deep.md",
                       "emulators/stale.md")