===================================================================
RCS file: /cvs/mandoc/TODO,v
retrieving revision 1.302
retrieving revision 1.307
diff -u -p -r1.302 -r1.307
--- mandoc/TODO	2020/04/26 21:41:07	1.302
+++ mandoc/TODO	2020/10/26 13:30:46	1.307
@@ -1,6 +1,6 @@
 ************************************************************************
 * Official mandoc TODO.
-* $Id: TODO,v 1.302 2020/04/26 21:41:07 schwarze Exp $
+* $Id: TODO,v 1.307 2020/10/26 13:30:46 schwarze Exp $
 ************************************************************************
 
 Many issues are annotated for difficulty as follows:
@@ -33,6 +33,21 @@ Obviously, as the issues have not been solved yet, the
 are mere guesses, and some may be wrong.
 
 ************************************************************************
+* assertion failures
+************************************************************************
+
+- .if n .ce in the middle of .TS data
+  afl case f1/id:000103,sig:06,src:009024+009105,op:splice,rep:2 (jes@)
+  While roff_parseln() prevents .ce and similar requests in the middle
+  of a tbl, the guard is no longer effective when the .ce is wrapped
+  in a roff block, for example a conditional.  The resulting assertion
+  has never been seen in any real-world manual page.
+  This is too dangerous to fix before release because it requires
+  reorganizing the very delicate internals of roff_parseln(),
+  which risks causing more severe bugs.
+  loc * exist *** algo *** size * imp *
+
+************************************************************************
 * missing features
 ************************************************************************
 
@@ -83,6 +98,25 @@ are mere guesses, and some may be wrong.
   Jan Stary 20 Apr 2019 20:16:54 +0200
   loc *  exist ***  algo ***  size **  imp *
 
+- mandoc replaces all ASCII control characters except tab and line feed
+  with '?' during input.  It would be better to replace them with
+  Unicode escapes in preconv_encode() or somewhere in the vicinity,
+  such that the already existing better replacement strings show
+  up in the output.  Emulating groff is not desirable: groff replaces
+  0x00, 0x0b, and 0x0d to 0x1f with the empty string (bad because
+  that's easy to overlook for the document author), 0x01 with '.'
+  (very confusing), and passes through 0x02 to 0x08, 0x0c, and 0x7f
+  raw (bad because that is insecure output).  Remember that 0x07 may
+  need special handling because it is sometimes used for certain
+  delimiters, so it may need handling *after* roff.c rather than before.
+  reminded by John Gardner 16 Jun 2020 14:26:28 +1000
+  Actually, more ASCII control characters than just 0x07 may need
+  later handling because they can for example be used in macro names.
+  So they may need handling after roff(7) processing.
+  pointed out by John Gardner 23 Jun 2020 18:28:08 +1000
+  more info from John Gardner 29 Jun 2020 19:54:04 +1000
+  loc **  exist **  algo **  size **  imp *
+
 --- missing mdoc features ----------------------------------------------
 
 - .Sh and .Ss should be parsed and partially callable, see groff_mdoc(7)
@@ -487,6 +521,10 @@ are mere guesses, and some may be wrong.
 ************************************************************************
 * warning issues
 ************************************************************************
+
+- shorten/simplify error messages for usage errors
+  To: deraadt@ 25 Oct 2020 23:37:01 +0100
+  loc **  exist *  algo *  size **  imp ***
 
 - warn about duplicate .Sh/.Ss heads
   gre(4): Rename duplicate sections 20 Apr 2018 15:27:33 +0200