[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Release of new system



System 65.0 with microcode 739 is released to users.  This system contains
a large number of changes which users need to know about.  It is consistent
with the soon-to-be-released new edition of the Chine Nual.

This message is very long, and so begins with a summary.

FILLARRAY and LISTARRAY are Maclisp compatible.
The BREAK and ABORT keys behave differently.
The catch tag SI:TOP-LEVEL has been renamed.
The file LMIO;CHSTBL and the variables it defined are gone.
INTERN with a symbol as argument is a little different.
LOAD-PATCHES is global now and takes different arguments.
The functions <, >, <=, and >= have been changed.
Notifications and errors inside the window system work better.
STRING-SEARCH-SET, STRING-TRIM, etc. have been improved.
(VALUES) works now.
New subprimitives for hacking closures.
Who line changes.
Internal lambdas named differently.
The arguments to READ, TYI, etc. have been changed slightly.
----

FILLARRAY and LISTARRAY are now fully Maclisp compatible; they will
accept arrays of any dimensionality.

The way the ABORT and BREAK keys work has been changed around; the Control
and Meta shifts modify the action of these keys.  Without Control, the key
takes effect when it is read.  Some programs (especially the editor) will
handle this case themselves.  With Control, the key takes effect
immediately, except in "super image" programs such as SUPDUP; in that case
the key must be prefixed with TERMINAL, and the Control shift is
irrelevant.  Without Meta, the BREAK key calls the BREAK function and the
ABORT key aborts to the previous command level.  With Meta, BREAK goes to
the error handler and ABORT resets the process.  This is designed so that
the "double bucky" combination is convenient.  If you have any trouble
remembering this, as usual type TERMINAL HELP and an explanation will be
displayed.
	Old:		New:			Does:
	break		break			deferred break
			control-break		immediate break
			meta-break		deferred error-handler
	terminal-break	control-meta-break	immediate error-handler
	abort		abort			deferred abort one level
			control-abort		immediate abort one level
			meta-abort		deferred abort all levels
	control-abort	control-meta-abort	immediate abort all levels

By popular demand, c-Z will act like ABORT in BREAK and the error handler
again.  Anyone with opinions as to whether or not the CALL key should be
changed also to be consistent with this scheme should mail them to
BUG-LISPM.

The following incompatibilities associated with this may be visible to some
user programs:  The catch tag SI:TOP-LEVEL has been renamed to
SYS:COMMAND-LEVEL.  The :ABORT, :BREAK, and :FORCE-BREAK messages have been
flushed.  The variable TV:KBD-IGNORE-ABORT has been replaced by the
variable TV:KBD-INTERCEPTED-CHARACTERS, which the user can bind to prevent
ABORT and BREAK without Control from getting their default handling.  Good
user-interface dictates that in this case the user program should do
something consistent with the default behavior.

The Chaosnet host table and associated data structures have been
reorganized as follows.  The file LMIO;CHSTBL and the variables
CHAOS:KNOWN-NAME-ALIST, CHAOS:FINGER-ALIST, and
CHAOS:CONSOLE-LOCATION-ALIST have been replaced by the files LMIO;HSTTBL
(generated automatically) and LMIO;LMLOCS and the variable
CHAOS:MACHINE-LOCATION-ALIST.

Calling INTERN (or INTERN-LOCAL) with a symbol as the first arg never
creates a new symbol.  Instead, if the symbol is not found in the specified
package, it is interned there as well as in its original package.  This
does not affect the package cell of the symbol.
Use (INTERN (GET-PNAME sym) pkg) to create a symbol that looks the
same as sym but is in pkg.

LOAD-PATCHES is now on the global package; you don't have to say "SI:" any
more (though it will work if you do).  It now takes keyword arguments.  The
keywords :SELECTIVE and :NOSELECTIVE control whether or not you get asked
about whether to load patches.  The default is :SELECTIVE, meaning that you
get asked.  You can use the :SYSTEMS keyword, followed by a list of system
names to specify which systems to load the patches of.  This is what the
argument to LOAD-PATCHES used to mean.  The default is to load patches for
all systems.  (This change to the arguments is incompatible, but it seems
quite likely that nobody uses the old argument anyway.)

The functions > and < have been upgraded to be exactly the same as GREATERP
and LESSP; that is, they will now accept more than two arguments.  The
functions >=, <=, , and  have been reimplemented to work correctly; the
old implementatation (as macros) did not have the property of evaluating
all arguments exactly once.  As always, all these functions work on all
types of numbers.

:BATCH mode works in MAKE-SYSTEM, saving the compiler warnings in a file
and otherwise conspiring not to want human intervention during compilation.

Notifications (those unsolicited, beeping messages enclosed in square
brackets) have been extensively revised.  They should behave much better
now.  As part of this the interface between the error handler and the
window system has been greatly improved; there should be much less
mysterious behavior.  The function TV:PRINT-NOTIFICATIONS will reprint all
the notifications received since the machine was last cold-booted.  If you
use any of the following, you are obsolete and need to convert (probably no
user programs use any of these):
Deleted messages:  :NOTIFY, :NOTIFY-STREAM, :NOTIFY-OUTPUT, :NOTE-INPUT-WAIT
Deleted variables: TV:BACKGROUND-STREAM-BELL-COUNT
Deleted functions: TV:GET-NOTIFICATION-STREAM, TV:NOTIFY-USER, TV:MAKE-SELF-INTERESTING
If you need help converting, don't hesitate to ask; help will be forthcoming.

The string-manipulation functions that deal in sets of characters, such as
STRING-SEARCH-SET and STRING-TRIM, now accept strings as well as lists for
representing sets of characters.  (string-trim "ab" "bbfooaa") => "foo".

The restriction against returning zero values has been lifted.  This is
only visible when the receiver of the values is a MULTIPLE-VALUE-LIST,
because of the way the language handles calls in which the callee returns
fewer values than the caller asked for.  Try typing (VALUES) at a Lisp
Listener.

New microcoded functions SI:%BINDING-INSTANCES, SI:%USING-BINDING-INSTANCES, and
SI:%INTERNAL-VALUE-CELL exist.  
  (SI:%BINDING-INSTANCES <list-of-symbols>)  "captures" the current bindings
     of a list of variables.  This operation is very similar that
     provided by CLOSURE.  In fact, CLOSURE could have been defined by
     (DEFUN CLOSURE (VARS FUN) (%MAKE-POINTER DTP-CLOSURE 
					      (CONS FUN (SI:%BINDING-INSTANCES VARS))))
     The value of SI:%BINDING-INSTANCES is a list of locatives of even length,
     which alternately point to the internal value cell of a variable and the
     paired external value cell.
  (SI:%USING-BINDING-INSTANCES <instance-list>) "installs" an instance list such as
     created by SI:%BINDING-INSTANCES.  In a manner similiar to the BIND function
     (which should be called %BIND), these bindings are "added" to the current
     binding frame, and remain in effect until it is unwound.  For this reason,
     you probably dont want the "user" to use SI:%USING-BINDING-INSTANCES directly,
     but a simple interface function which uses it.
     SI:%USING-BINDING-INSTANCES checks for redundant bindings and ignores them.
     (A binding is redundant if the symbol is already bound to the desired external
     value cell).  This check avoids excessive growth of the special pdl in some cases
     and is also made by the microcode which "enters" closures, entities, and instances.
  (SI:%INTERNAL-VALUE-CELL <symbol>) returns the current internal value cell of
     <SYMBOL>.  ONE-Q-FORWARD forwarding pointers are followed, but 
     EXTERNAL-VALUE-CELL-POINTERS are not.
  In addition, certain internal changes were made to allow EXTERNAL-VALUE-CELL-POINTER
  datatypes to appear in the PDL-BUFFER.  These should not be visible to the user.

The who line has been reimplemented.  Each field in the who line is now a
window, and new fields can be added by adding windows in a simple fashion.
Along with this feature, the who line can change size, and a new mixin
flavor has been defined to permit changing the size of the "screen", so
that windows scale correctly.  A new default who line line has been added,
called the mouse documentation line, which gives a one line description of
what the mouse can do at any time.  This description is usually supplied by
the window.  We have documented many system facilities, and will continue
to improve the documentation as time permits.  The menu item list format
has been added to in an upwards compatible fashion to permit users to
document their menus.  If people are interested in general, I'll send out a
general note on how to use the documentation facility, and on who line
fields (--HIC).  The who line documentation can be turned on and off with
the function (TV:WHO-LINE-DOCUMENTATION &optional (ON-P T)).  An argument
of T (or no arg), turns it on, and an argument of NIL turns it off.  Note
that the screen is now one line smaller by default.  This will probably
break some code that uses numbers to specify the size of windows (tough, we
warned you not to do that).  The documentation window is in the variable
TV:WHO-LINE-DOCUMENTATION-WINDOW, and if you are bummed out by the reverse
video, you can send it a :SET-REVERSE-VIDEO-P NIL message.

The function TIME:NWATCH-ON is now TV:NWATCH-ON.  Some people have this in their
init files.  It is unnecessary to call this function anyway, since it has been
the default for a long time.

There is a new function-spec, (:INTERNAL function-spec number), which is
used for internal lambdas in compiled programs--"(FUNCTION (LAMBDA (...".
This replaces the old "function-INTERNAL-G6985" symbols.

The functions which take optional stream and eof-option arguments
(READ, TYI, etc.) now interpret these arguments in a more consistent fashion.
Refer to the on-line copy of the manual for details, or ask someone.
The change is not completely upward-compatible but should result in
much more reasonable behavior.