Validate the encoding of files marked UTF-8

tools/check-typo validates the encoding of any file with ocaml-typo=utf8
and will not proceed if the test fails.

tools/check-typo

@@ -81,9 +81,9 @@
 # You can ignore a rule by giving the option -<rule> on the command
 # line (before any file names).
 
-# Files which include the utf8 rule will have line-length computations take
-# UTF-8 sequences into account. As a special case, UTF-8 sequences are always
-# allowed in the copyright headers.
+# Files which include the utf8 rule will be validated using grep and line-length
+# computations will take UTF-8 sequences into account. As a special case, UTF-8
+# sequences are always allowed in the copyright headers.
 
 # First prevent i18n from messing up everything.
 export LC_ALL=C
@@ -163,6 +163,19 @@ IGNORE_DIRS="
       ocamldoc/*|*/ocamldoc/*) rules="long-line,$rules";;
     esac
 
+    case ,$svnrules, in
+      *,utf8,*)
+        # grep -a is used to force the file to be considered as text and -x
+        # requires the entire line to match. This specifically detects the
+        # presence of lines containing malformed UTF-8. It may be tested using
+        # https://www.cl.cam.ac.uk/~mgk25/ucs/examples/UTF-8-test.txt
+        if LC_ALL=en_US.UTF8 grep -qaxv '.*' "$f" ; then
+          echo File "$f" is not correctly encoded in UTF-8
+          exit 2
+        fi
+        ;;
+    esac
+
     (cat "$f" | tr -d '\r'; echo) \
     | awk -v rules="$rules" -v svnrules="$svnrules" -v file="$f" \
       '