fix Bug 1950 . the improved text filtering killed the safety logic. w…

…ow. we’ve been running blind since Dec 6.

cmake/matlab_clean.pl

@@ -8,23 +8,22 @@
 # This appears to be a necessary evil to get clean matlab output to ctest:  
 #  http://www.mail-archive.com/[email protected]/msg02175.html
 
+#use lib "./Text-Unidecode-0.04/lib";
+#use Text::Unidecode;
+
 my $cmd = 'matlab';
 foreach my $a(@ARGV) {
   $cmd .= " \"$a\"";
 }
 $cmd .= ' 2>&1 ';
 
-# use a pipe so as not to buffer output. in case matlab takes a vacation midstream
-open(my $command_stream, "-|", $cmd);
-while(<$command_stream>){
-    # le junque. on separate lines for readability
-    $_ =~ s/.*exclude an item from Time Machine.*\n//g;
-    $_ =~ s/[^\x{0009}\x{000a}\x{000d}\x{0020}-\x{D7FF}\x{E000}-\x{FFFD}]+//g;
-    $_ =~ s/\x1b\[\?1h//g;
-    $_ =~ s/\x1b=//g;
-    $_ =~ s/\x1b\[\?1l\x1b>//g;
-    print;
-}
+my $matlab_output = `$cmd`;
 my $retval = $? >> 8;
+
+$matlab_output =~ s/.*exclude an item from Time Machine.*\n//g;
+#$matlab_output = unidecode($matlab_output);  # didn't seem to help
+$matlab_output =~ s/[^\x{0009}\x{000a}\x{000d}\x{0020}-\x{D7FF}\x{E000}-\x{FFFD}]+//g;
+
+print($matlab_output);
 exit($retval);