Multi byte characters should not be treated as part of an operator

compiler/lexer.nim

@@ -25,7 +25,7 @@ const
   SymChars*: set[char] = {'a'..'z', 'A'..'Z', '0'..'9', '\x80'..'\xFF'}
   SymStartChars*: set[char] = {'a'..'z', 'A'..'Z', '\x80'..'\xFF'}
   OpChars*: set[char] = {'+', '-', '*', '/', '\\', '<', '>', '!', '?', '^', '.',
-    '|', '=', '%', '&', '$', '@', '~', ':', '\x80'..'\xFF'}
+    '|', '=', '%', '&', '$', '@', '~', ':'}
 
 # don't forget to update the 'highlite' module if these charsets should change
 

lib/packages/docutils/highlite.nim

@@ -130,7 +130,7 @@ proc nimNumber(g: var GeneralTokenizer, position: int): int =
 
 const
   OpChars  = {'+', '-', '*', '/', '\\', '<', '>', '!', '?', '^', '.',
-              '|', '=', '%', '&', '$', '@', '~', ':', '\x80'..'\xFF'}
+              '|', '=', '%', '&', '$', '@', '~', ':'}
 
 proc nimNextToken(g: var GeneralTokenizer) =
   const

tests/parser/tunicodeidents.nim

@@ -0,0 +1,11 @@
+discard """
+    action: run
+"""
+
+# #7884
+
+type Obj = object
+    ö: int
+
+let o = Obj(ö: 1)
+doAssert o.ö == 1