feat: complete Chinese tutorial text and layout validation

tools/script_text.py

@@ -20,6 +20,15 @@ FORMAT = re.compile(r'%%|%[-+ #0]*[\d*]*(?:\.[\d*]+)?[hlL]?[diouxXeEfgGcs%]')
 CONTROLS = re.compile(r'[\a\b\f\n\r\t\v]')
 
 
+def format_tokens(text):
+    # "speed 0% is ..." is prose, not the printf conversion "% i".
+    # Only exclude a numeric percent whose apparent conversion cuts a word.
+    return [m.group() for m in FORMAT.finditer(text)
+            if not (m.start() > 0 and text[m.start() - 1].isdigit()
+                    and m.end() < len(text) and text[m.end()].isascii()
+                    and text[m.end()].isalpha() and ' ' in m.group())]
+
+
 @dataclass(frozen=True)
 class Literal:
     start: int
@@ -121,6 +130,17 @@ def extract_display_literals(source: bytes, encoding: str = 'latin1') -> list[Li
             if len(arg) == 1 and arg[0].kind == 'string':
                 lit = arg[0]
                 found.append(Literal(lit.start, lit.end, lit.value, token.value, number))
+            elif (len(arg) == 3 and [t.kind for t in arg] == ['string', 'name', 'string']
+                  and arg[1].value == 'F'
+                  and arg[0].value.startswith('\fTo use a charged spell select it')
+                  and arg[0].value.endswith('pressing the ')
+                  and arg[2].value.startswith(' KEY marked under it')):
+                # Stock English Tutorial.dsc has unescaped quotes around F in
+                # this specific display argument. Keep the complete text span;
+                # do not generalize to expressions or other malformed strings.
+                found.append(Literal(arg[0].start, arg[2].end,
+                                     arg[0].value + '"F"' + arg[2].value,
+                                     token.value, number))
     return sorted(found, key=lambda literal: literal.start)
 
 
@@ -150,7 +170,7 @@ def patch_display_calls(source: bytes, translations: dict[str, str], *, allow_al
             translated_controls = ['ALIGN' if c in '\a\b\f\v' else c for c in translated_controls]
         if original_controls != translated_controls:
             raise ValueError('control sequence changed: %r' % literal.text)
-        if FORMAT.findall(literal.text) != FORMAT.findall(translated):
+        if format_tokens(literal.text) != format_tokens(translated):
             raise ValueError('format sequence changed: %r' % literal.text)
         replacements.append((literal.start, literal.end, encode_literal(translated)))
     result = source