fix: strip links with bracketed text from meta

scripts/generate_site.py

@@ -2485,10 +2485,16 @@ def _browser_title(relative: Path, title: str) -> str:
     return title
 
 
+_LINK_TEXT = r"(?:`[^`]*`|\[[^\]\[]*\]|[^\]\[])"
+
+
 def _plain_markdown(text: str) -> str:
     text = re.sub(r"<[^>]+>", " ", text)
-    text = re.sub(r"!\[([^]]*)\]\([^)]*\)", r"\1", text)
-    text = re.sub(r"\[([^]]+)\]\([^)]*\)", r"\1", text)
+    # Link text may hold brackets: a code span or a No-Intro name such as
+    # `[BIOS] LaserActive PAC-N1 (Japan) (v1.02).bin` left the whole link,
+    # anchor included, in the Pioneer page's meta description.
+    text = re.sub(rf"!\[({_LINK_TEXT}*)\]\([^)]*\)", r"\1", text)
+    text = re.sub(rf"\[({_LINK_TEXT}+)\]\([^)]*\)", r"\1", text)
     # Code spans keep their text whole: validate_schemas.py read
     # validateschemas.py once every underscore was taken for emphasis.
     pieces = re.split(r"(`[^`]*`)", text)

tests/test_site_emulator_page.py

@@ -148,6 +148,15 @@ class WhatASearchSnippetSays(unittest.TestCase):
         self.assertTrue(plain.startswith(text[:-3]))
         self.assertIn(plain[len(text) - 3], " ,:")
 
+    def test_a_link_whose_text_holds_brackets_keeps_its_text_only(self):
+        """The Pioneer meta description published the whole link, anchor
+        included: the text `[BIOS] ...bin` stopped the pattern at its "]"."""
+        self.assertEqual(
+            gs._plain_markdown("**Start with** [`[BIOS] LaserActive PAC-N1 (Japan) (v1.02).bin`]"
+                               "(#sha1-6973e2593e66). ares reads it."),
+            "Start with [BIOS] LaserActive PAC-N1 (Japan) (v1.02).bin. ares reads it.")
+        self.assertEqual(gs._plain_markdown("[a [b] c](x) and [d](y)"), "a [b] c and d")
+
     def test_code_keeps_its_underscores(self):
         self.assertEqual(gs._plain_markdown("only `validate_schemas.py` and _this_"),
                          "only validate_schemas.py and this")