fix(core): ui questionmark

tests/test_import_flow.py

@@ -160,7 +160,7 @@ def test_audio_tags_written_on_import(client):
     }]})
     assert resp.json()["results"][0]["ok"]
 
-    dest = root / "Ulf Blanck" / "Die drei ？？？ Kids Folge 085"  # "?" becomes fullwidth
+    dest = root / "Ulf Blanck" / "Die drei ??? Kids Folge 085"
     files = sorted(dest.iterdir())
     tags = ID3(files[0])
     assert str(tags["TALB"]) == "Die drei ??? Kids Folge 085"

tests/test_naming.py

@@ -9,9 +9,9 @@ def req(**kw):
     return SimpleNamespace(**base)
 
 
-def test_sanitize_replaces_forbidden_chars():
-    assert sanitize("Die drei ???") == "Die drei ？？？"
-    assert sanitize("a/b") == "a⧸b"
+def test_sanitize_keeps_special_chars_but_strips_separators():
+    assert sanitize("Die drei ???") == "Die drei ???"
+    assert sanitize("a/b\\c") == "abc"
     assert "\x01" not in sanitize("a\x01b")
 
 

wordarr/naming.py

@@ -1,16 +1,12 @@
 import re
 
-# characters not allowed in (Windows/SMB-safe) file names are replaced with
-# fullwidth lookalikes so names like "Die drei ???" stay readable
-_REPLACEMENTS = str.maketrans({
-    "<": "＜", ">": "＞", ":": "：", '"': "＂", "/": "⧸",
-    "\\": "⧹", "|": "｜", "?": "？", "*": "＊",
-})
-_CONTROL = re.compile(r"[\x00-\x1f]")
+# only path separators and control chars are truly forbidden on Linux; keep
+# everything else literal so names like "Die drei ???" survive as-is
+_FORBIDDEN = re.compile(r"[/\\\x00-\x1f]")
 
 
 def sanitize(part: str, max_len: int = 120) -> str:
-    part = _CONTROL.sub("", part.translate(_REPLACEMENTS)).strip(" .")
+    part = _FORBIDDEN.sub("", part).strip(" .")
     part = re.sub(r"\s+", " ", part)
     return part[:max_len].strip(" .") or "Unknown"