I hate to sound ungrateful, but....

Somewhere between version 18 and version 19 someone has hacked on the
VMS sections of the code.  And it looks like they didn't finish.  Most
of the problems I've found are minor, however, I've run into some that
are leaving me puzzled.

[ They were quite unfinished... --- Richard Levitte ]
================
In sysdep.c:

There is a small piece of code in sysdep.c that refers to undeclared
variables.  It is not commented, and it is not in 18.58.  I have no idea
what it is supposed to be doing, so for now I've commented it out.  This
gets sysdep.c to compile, with a few warning messages.  Here's the
function with the code:

kbd_input_ast ()
{
  register int c = -1;
  int old_errno = errno;
  extern EMACS_TIME *input_available_clear_time;

  if (waiting_for_ast)
    SYS$SETEF (input_ef);
  waiting_for_ast = 0;
  input_count++;
#ifdef ASTDEBUG
  if (input_count == 25)
    exit (1);
  printf ("Ast # %d,", input_count);
  printf (" iosb = %x, %x, %x, %x",
	  input_iosb.offset, input_iosb.status, input_iosb.termlen,
	  input_iosb.term);
#endif
  if (input_iosb.offset)
    {
      c = input_buffer;
#ifdef ASTDEBUG
      printf (", char = 0%o", c);
#endif
    }
#ifdef ASTDEBUG
  printf ("\n");
  fflush (stdout);
  sleep (1);
#endif
  if (! stop_input)
    queue_kbd_input ();
/* I don't know what this is doing!  The variables buf, cbuf and i are
   not declared.  This is new from version 18, what does it do?
  if (c >= 0)
    {
      struct input_event e;
      e.kind = ascii_keystroke;
      XSET (buf[i].code, Lisp_Int, cbuf[i]);
      e.frame = selected_frame;
      kbd_buffer_store_event (&e);
    }
*/
  if (input_available_clear_time)
    EMACS_SET_SECS_USECS (*input_available_clear_time, 0, 0);
  errno = old_errno;
}

So the question is, who put it in and why?

[ Quite simple, really... Someone copied the code from
  read_avail_input () in keyboard.c, and started making appropriate
  changes... and left it half done... It is fixed now --- Richard Levitte ]

[ kbd_input_ast has been further changed, to be able to cope with X events
  and other stuff --- Richard Levitte ]

================
In regex.c:

I can't figure out what VAXC hates about the macros EXTRACT_NUMBER and
EXTRACT_NUMBER_AND_INCR.  Here is a typical error

    Compiling REGEX...
		      EXTRACT_NUMBER_AND_INCR (j, p);
    %CC-W-REPLACED, Replaced reserved word "char" with ")".
		    Listing line number 5925.
		    At line number 2658 in GNU:[EMACS19.SRC]REGEX.C;1.

		      EXTRACT_NUMBER_AND_INCR (j, p);
    %CC-E-SYNTAXERROR, Found ")" when expecting
		    one of { "," ";" }.
		    Listing line number 5925.
		    At line number 2658 in GNU:[EMACS19.SRC]REGEX.C;1.

		      EXTRACT_NUMBER_AND_INCR (j, p);

And here is the actual source line(s)

	case jump:
        case jump_past_alt:
	case dummy_failure_jump:
          EXTRACT_NUMBER_AND_INCR (j, p);
	  p += j;	
	  if (j > 0)
	    continue;
            
The macro definition looks fine to me, and the expansion is okay.  This
looks to me like a bug in VAX C.  The stuff below is from the listing
produced by the compiler (formatted pretty for readability).

    EXTRACT_NUMBER_AND_INCR (j, p); 
    1	do { 
	    EXTRACT_NUMBER (j, p);
	    (p) += 2;
	} while (0);
    2   do {
	    do {
		(j) = *(p) & 0377;
		(j) += SIGN_EXTEND_CHAR (*((p) + 1)) << 8;
	    } while (0);
	    (p) += 2;
	} while (0);
    3   do {
	    do {
		(j) = *(p) & 0377; 
		(j) += ((signed char) (*((p) + 1))) << 8; 
	    } while (0);
	    (p) += 2;
	} while (0);

It seems to be at this third level of expansion that the compiler barfs.
It replaces the (valid) reserved word `char' with a ')' which is
horribly wrong.  It screws up the cast.

[ The reason for this behavior is that VAX C doesn't know about signed.
  It is now fixed. --- Richard Levitte ]

================
In vmsproc.c

The function call_process_cleanup is missing.  For the time being, I've
lifted the version from process.c.  It will probably cause Emacs to
crash the first time it is called, but will serve for now.

The macros DCL_PROMPT, INTERACTIVE, RUNNING, EXITTED, and CHANGED are
not defined.  I defined them in vmsproc.h as "$" for DCL_PROMPT, with
the rest as integers, which is probably wrong.

Vprocess_alist is referenced, but it is not at all clear to me that this
is what should be being used here.  I don't understand the ideas behind
how this was to be run in the first place....

[ Neither did I at first... I've hacked this part of Emacs a lot, and
  it works now, but there might be more to do. --- Richard Levitte ]