@comment -*- Mode:Text; Package:User; Readtable:CL -*-

@subsection Common-LISP Function

Evaluating @l[(common-lisp t t)]
now causes all new Lisp Listeners and Zwei
Buffers without @l[:readtable] attributes to use the CommonLISP readtable by default.

@subsection READ

@l[SET-MACRO-CHARACTER] now signals a legible error when passed certain
kinds of invalid character arguments.  Invalid arguments include
any characters with special bits (font, control, meta, etc.).

@l[UNREAD-CHAR] now returns NIL, in conformity with CommonLISP.

The error strings and proceed types for various reader end-of-file conditions have been fixed:

@enumerate
@item When the LISP Listener encounters end-of-file immediately after a package prefix,
it now displays the warning ``EOF after package prefix''.

@item @l[READ-END-OF-FILE] errors now have proceed options appropriate to the type
of error encountered:

@itemize @bullet
@item
unterminated strings generate a proceed option which offers to return the
null string;
@item
unterminated symbols generate a proceed option which offers to return NIL;
@item
unterminated lists generate a proceed option which offers to "close off all unfinished lists."
@end(itemize)

@item Both the CommonLISP and ZetaLisp versions of READ-FROM-STRING now
report end-of-file correctly.

@end(enumerate)

Problems with displacing a list into a back-quoted array have been solved.
The following now works correctly:

@lisp
(setq x '(a b c))
`#(a ,@@x b)

#(A A B C B)
@end(lisp)

@subsection Symbols

@l[EXPORT] no longer reports name conflicts with external symbols which have
been previously specified as shadowed.

@subsection Numbers

Negative bignums are now printable.

Complex numbers 
with zero imaginary parts are no longer coerced into real numbers; functions
which relied on this behavior will now signal errors.

@subsection Output

The pretty-printer now displays the @l[function] special form using
the @l[#'] abbreviation.
Hence, @l[(pprint '(function car))] will now display @l[#'CAR].

@subsection Hash Tables

Hash tables of size 1 may now be created using @l[make-hash-table],
as may hash tables with integer @l[:rehash-threshold] and unspecified
@l[:actual-size].

HASH-TABLE is now an alias for BASIC-HASH-TABLE; that is to say,
the following will now return T:
@lisp
(typep (make-hash-table) 'hash-table)
T
@end(lisp)

Hash tables can be created using the @l[EQL] test predicate as  the @l[:TEST] keyword value to @l[MAKE-HASH-TABLE].
These are distinguishable from hash tables using the @l(EQ) predicate.

Passing an invalid @l(:TEST) keyword value to @l(MAKE-HASH-TABLE)
now signals a legible error. For example:
@lisp
(make-hash-table :test "foo")

>>ERROR: Test "foo" is not valid: should be one of =, STRING-EQUAL, STRING=, EQUALP, EQUAL, EQL, EQ

@end(lisp)

@subsection Compiler

The compiler no longer mishandles the following case of local variable binding:

@lisp
(defun foo1 (x)
  (let ((x x)) 
    x))
@end lisp

When the compiler optimized PUSHNEW, it caused an error.
The optimized code was applying the test function to
its arguments in the wrong order.
This problem occurred only with test functions for which
argument order is significant.
For example, the following now works correctly:

@lisp
(let ((x '(1 2 3))) (pushnew 4 x :test '<))

(4 1 2 3)
@end(lisp)

Under certain circumstances, the compiler mis-optimized @l(=), @l(<), and @l(>) into non-existent functions;
they are now optimized into the correct functions.

@subsection String Input Streams

@l[MAKE-STRING-INPUT-STREAM] now creates a stream which supports
the following standard stream methods:

@itemize @bullet
@item :CLOSE
@item :TYI
@item :ANY-TYI
@item :TYIPEEK
@item :UNTYI
@item :POINTER
@item :GET-STRING-INDEX
@item :SET-POINTER
@item :WHICH-OPERATIONS
@end(itemize)

The generated stream calls the stream-default's handlers
to deal with any operations other than those listed above.

@subsection Sequences

The @l(SETF) method for @l(CAAAR) has been implemented correctly.

@l[mismatch] now behaves as advertised when the :FROM-END keyword is
specified and the second sequence argument is shorter than the first.

All modifying @l[setf] methods for @l[bit] and @l[sbit] now handle 
multi-dimensional arrays as well as one-dimensional arrays.  

All functions with the prefix @l[bit-] (such as @l[bit-not], @l[bit-xor],
etc.) now operate only on arrays with similar rank and similar dimension.

@l[ADJUST-ARRAY] now no longer requires the @l[:NEW-DIMENSIONS] to be a list
for the case of one-dimensional arrays.

@l[REMOVE], @l[SUBSTITUTE], @l[DELETE], and other functions which take sequences as
arguments now respect and maintain the active lengths of array arguments with
fill-pointers.

The @l(SETF) method for the keyword argument @l[:fill-pointer] to @l[make-array]
now performs a check with
@l[array-has-leader-p] before assigning the fill-pointer to the front of the
array. 

Error messages generated by @l[MAKE-ARRAY] now print the number of dimensions,
array-rank-limit, and dimension lists of the array in decimal.

@l(ADJUST-ARRAY) now makes the following CommonLisp-compatible error checks:
@enumerate
@item an array cannot be displaced to itself;
@item an array cannot be displaced to an array which is indirected (to any level)
to the first array;
@item an array cannot be displaced to another array of a different type.
@l(MAKE-ARRAY) can still displace across array types.
This fails, however, if you try to displace an ART-Q array to an ART-STRING array.
@end(enumerate)

If the :displaced-index-offset is specified without a :displaced-to value,
@l[MAKE-ARRAY] uses the current indirect array as the array to which to
displace the original array.  If there is no current indirect array, an error
is generated.  An error is also generated if more than one of
@l[:initial-element], @l[:initial-contents], and @l[:displaced-to] is
specified.

@subsection Miscellaneous Improvements

Alternate macro definitions are now provided for both @l[PROG1] and @l[PROG2],
in conformity with the CommonLISP specification.
The @l[defun] (function)
definitions remain the default definitions for both @l[PROG1] and @l[PROG2].

@l[PSETQ] now returns NIL, in conformity with CommonLISP.
Also, @l[PSETQ] signals the correct error when given an odd number of arguments.

The following Common LISP functions now signal an error if both 
a @l[:test] and a @l[:test-not] keyword is specified:

@lisp
ADJOIN
ASSOC
COUNT
DELETE
DELETE-DUPLICATES
FIND
INTERSECTION
MEMBER
MISMATCH
NINTERSECTION
NSET-DIFFERENCE
NSET-EXCLUSIVE-OR
NSUBLIS
NSUBST
NSUBSTITUTE
POSITION
RASSOC
REMOVE
REMOVE-DUPLICATES
SEARCH
SET-DIFFERENCE
SET-EXCLUSIVE-OR
SUBLIS
SUBSETP
SUBST
SUBSTITUTE
UNION
@end lisp

For structures of @l[(:type vector)], the generated type predicate
no longer causes an error when invoked.