.ax Release Notes
.in rnochaini
.f.j
.p
The new version of RUNOFF has a number of new features and enhancements.
This list is an attempt to enumerate all of these changes, but it may
not be complete.  The RUNOFF manual should be consulted for a complete
description of the new version of RUNOFF.
.LIST
.LM-4
.le
A new method of dispatching commands has been adopted which resulted
in a very substantial increase in speed at a small investment in memory.
The old method would search the entire command table for each command,
whereas the new method uses the first letter in the command to begin
searching at the appropriate location in the table.  This increased
the memory requirements by about 30. words, but reduced the CPU
requirements by 20-30%%, depending on the file being processed.
.le
On operating systems which support it, the new version of RUNOFF will
use multi-buffering for the input and output files. This, along with
the above CPU time reductions, reduces the total running time for
many RUNOFF applications by a total of 35-50%%, with a usual reduction
of around 40%%, depending on the file being processed and the disks
being used for the file I/O.
.le
The buffer usage for all of RUNOFF's internal buffers has been
re-arranged to be standard for all of RUNOFF's internal buffers.  This
has allowed the modification of RUNOFF to be much easier than dealing
with the haphazard methods used before the modification.
.le
RUNOFF now has a character table which contains character attributes
for all of the ASCII characters.  These attributes include whether
the character is an upper or lower case alphabetic, numeric, etc.
This replaces a number of special case code sequences for locating upper or
lower case characters, numerics, and others;  there are still a few
places which look for special cases (not contained in the character
table)  but these are rare.
.le
RUNOFF's usage of dynamic memory has been standardised.  The old method
had two stacks growing toward each other, one for the index and the
other for the footnotes.  This made it difficult to use dynamic memory
for other purposes, and also made it virtually impossible to use the
extend task directive under RSX-11M (and soon to be supported in IAS),
especially since pointers in the index and footnote buffers would have
to be modified and the buffers moved in order to properly use the new
areas allocated.  The new method has all dynamic memory allocations
go through a common subroutine, which maintains a heap of available
memory.  The degradation in time is usually quite minimal, less than
1%% for most applications, especially since memory is not usually
allocated and deallocated in ways which fragment the heap.
This has allowed RUNOFF to use dynamic memory for numerous applications
where it is appropriate, and also to properly use the extend task
directive for RSX-11M.
.le
The infamous self-modifying code in RUNOFF has all been removed.
This has allowed RUNOFF to be broken up into read-only and
read-write sections.  For systems which support sharable read-only
code, this means that only one copy of the read-only portion (and
it is substantial) need be in memory at one time, even if multiple
copies of RUNOFF are running.
.le
The number of error messages for RUNOFF has been increased and the
messages themselves have been made easier to understand.  Numerous
conditions which were once passed by without comment (such as truncating
index items which were too long to be printed) are now remarked
upon to the user so that corrective action can be taken if desired.
Because this increase in the number of error messages would imply
a large increase in task image size, the error messages have been
broken off into a file, and only the message required is read into
memory. This does not use dynamic memory for the buffer, in order
to be able to report dynamic memory exhaustion without getting into
an infinite error loop.
.le
Much of the internals of RUNOFF have been cleaned up and documented
properly for the first time.  Also, related subroutines have been
broken up into more modules to allow easier modification.
.le
The references to %$MUL and %$DIV have been conditionalised to be used
only on non-EIS machines, dependent on the definition of R%$%$EIS in
the prefix file.  This allows a moderately faster and smaller version
to be built on an EIS machine. R%$%$EIS is also used in a few places
to conditionalise the use of the ASH instruction.  Currently, there
are a few uses of the SOB instruction which should be supported on
all machines other than the PDP-11/20 and PDP-11/10.
.le
It is now possible to request output for specific ranges of chapters
as well as for ranges of pages.  In addition, the default base of the
numbers in the switches for RUNOFF has been changed from octal to
decimal.
.le
RUNOFF now supports superscripts and subscripts for some types of
output devices.  The DIABLO terminal (and a few others which are
compatible with it)  is supported in this mode;  in connection with
this, two new meta-characters have been introduced:  %{ (begin
superscripts) and %} (begin subscripts).  These work like the
underlining meta-characters: just %{ only superscripts the next
character, while %^%{ puts the terminal into a "latched-superscript"
mode.  The subscript character works similarly.
.le
The bugs in change bars have been fixed.
.le
RUNOFF supports include files.  These files can be included in the
text at any point via a .include command, and can be used to reference
standard text or standard RUNOFF commands.  Included files can be
nested (that is, they can include other files), and the only limit to
the depth of nesting is the amount of dynamic memory available.
.le
RUNOFF now supports multiple input files on the command line, and will
support multiple command lines.  Multiple command lines are requested
by having the last significant character of the command line be a "-".
.le
RUNOFF now provides the ability to change meta-characters from their
default settings to a setting more compatible with the user's needs
with the .change command.
.le
Variables and expressions are supported wherever numbers were supported
before.  A new command, .setn, is provided to set numeric variables. An
unusual quirk appears in relative numbers:  a + or - at the beginning
of an expression means to add or subtract the expression from the old
value of the item being set (as in the older versions of RUNOFF), but
a + or a - at the beginning of a term within an expression has the
normal arithmetic meaning of not negating or negating the term,
respectively.  Thus, in order to specify a negative value in an
expression as opposed to a relative value, it would be necessary to
write, for example, (-3) as opposed to -3.
Dynamic memory is used to hold the symbol table.
.le
Page numbering can now be done at the bottom center of each page
instead of in the upper right hand corner.
.le
The default mode for the page number header has been changed from
UPPER to MIXED (i. e., from "PAGE" to "Page").
.le
A .date command can be used to stamp each page with the date of the
creation of the RUNOFF file (or any other short string).
This can also be applied to the Table of Contents file.
.le
A table of contents output file is now supported.  This places all of
the chapter headings into the table of contents with the page number
where the chapter begins.  Along with this, it is possible to specify
the table of contents date (similar to the .date command above), and
to swap the default output file to the table of contents file (say to
have a title page or a prefix to the table of contents listing or a
postfix to the table of contents).  Pages in the table of contents
are numbered with Roman numerals.
.le
Alphabetic list elements are now supported.
.le
Roman numeral list elements are now supported.
.le
List elements with a bullet (o) denoting the beginning of the item
are now supported.
.le
Numeric list elements with numbers greater than 99 are now supported.
.le
List elements with a user-supplied string as the list element identifier
are now supported.
.le
List elements which have no indentation for the first line and
indentation for all lines thereafter in the list element are now
supported.
.le
The .ap command now takes the same arguments as the .p command.
.le
It is now possible for the RUNOFF file to pass a message to the
user (such as "This file last edited on mm/dd/yy") which is printed
out on the user's terminal rather than on the output file.
.LM+4
.end list
.p
The new version of RUNOFF has a few restrictions not found in all of
the previous versions of RUNOFF.  This is an attempt to list those
restrictions.
.list
.LM-4
.le
The new RUNOFF (as noted above) has a few references to the SOB
instruction.  This instruction is supported on all machines (including
non-EIS PDP-11/40's, LSI-11's, and others) with the exception of the
PDP-11/20 and the PDP-11/10.  If it is necessary to use RUNOFF on one
of these machines, it would be easy to put a macro defining the SOB
instruction in the prefix file.
.le
The new version of RUNOFF is somewhat more strict about where
certain switches are placed.  In particular, the /WA switch
must now be specified only on the output file (not on the
input file or the table of contents file).  This was necessary
because the addition of the extra file made some of the possible
combinations ambiguous.
.lm+4
.end list