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

CLISP 1995-12-08



There's a new release of CLISP ready.

You can get CLISP from:
 
   ma2s2.mathematik.uni-karlsruhe.de
     (in /pub/lisp/clisp/source and /pub/lisp/clisp/binaries)

   The binaries built so far are:
     i386 / LinuxELF
     MIPS / Irix 5.3
     m68k / NEXTSTEP 3.2
     Sun4m / Solaris 2.4
     Sun4c / SunOS 4.1.3

Due to bandwidth problems, the NEXTSTEP build has not yet been copied
to ma2s2.  The NEXTSTEP build and some other builds not available on ma2s2
are available from:
  
  http://sayre.sysc.pdx.edu:8001/clisp/binaries

However, note the files found on this US WWW site are not necessarily aligned
with official release dates.

Some free WWW browsers:
  Chimera (X Windows/non-Motif): ftp.cs.unlv.edu:/pub/chimera
  Lynx (terminal):  ftp2.cc.ukans.edu:/pub/lynx
  Mosaic (X Windows/Motif): ftp.ncsa.uiuc.edu/Web/Mosaic/Unix
  W3 (Emacs):  ftp.cs.indiana.edu:/pub/elisp/w3
  OmniWeb (NEXTSTEP): ftp.cs.orst.edu:/pub/next/binaries/wide-area-info

US users may prefer:  
  ftp.stat.ucla.edu:/pub/lisp/clisp

(but there will be some delay before the files are mirrored).

8 December 1995
================

User visible changes
--------------------

* Bug fix to prevent unbound slots from causing pathological 
  conflicts with the condition system.  Thanks to Bruno Haible
  and Heinrich Taube.

* When MAKE-PATHNAME :DIRECTORY is given a string, it is treated as 
  equivalent to (:ABSOLUTE string), as ANS dictates.  Thanks
  to Heinrich Taube for noticing this.

* Fixed a bug in the FFI where call-ins which returned structures or
  unions would not compile correctly.  Thanks to Michael Stoll.

* Added a <c-type>, C-PTR-NULL to the FFI.  C-PTR-NULL will convert
  NIL objects into NULL (so that type conflicts can be avoided).

* X3J13 vote <72> is implemented: all standard Common Lisp data objects 
  other than symbols and lists are self-evaluating. 
  Thanks to Pierpaolo Bernardi. 

* SOCKET-STREAMS are implemented.  See impnotes for details.

* Loop macro modified to allow sequential, list/vector-computed "FOR" values.

* Fixed a bug in the FFI which caused foreign double variables to be
  incorrectly converted.  Thanks to Bruno Haible and Ken Olum.

* Callbacks to code which don't respect CLISP global registers are possible. 
  Thanks to Bruno Haible and Ken Olum (Dec Alpha, OSF 2, FORTRAN).

* Fixed a double-float conversion routine used by the FFI. 
  0.0d0 wasn't dealt with properly.  Thanks to Bruno Haible and Ken Olum.

* Fixed a bug in the FFI which prevented multidimensional arrays from
  being used as :OUT arguments.  Thanks to Ken Olum.

* Fixed a bug in the floating point number printing routine:
  In rare cases, a number was printed with more digits than necessary
  (without, however, violating the print-read consistency). For example,
  2.1416s29 printed as 2.14159s29, and 3.002618f-31 printed as 3.0026179f-31.
  Thanks to Bruno Haible.

* Fixed a bug in ATAN: (ATAN y 0) with y < 0 returned pi/2 instead of -pi/2.
  Thanks to Bruno Haible.

* Fixed a bug in FORMAT which could (in very rare cases) have caused a
  wrong number to be printed by the FORMAT ~F, ~G, ~$, ~E directives.

Portability
-----------

* Runs on Windows NT.

* FreeBSD/NetBSD patches and compilation notes thanks to 
  Douglas Crosher (dtc@stan.xx.swin.oz.au).

* Updated OS/2 support thanks to Raymond Keoni Garcia
  (rkgarcia@students.wisc.edu).

* Remove Atari support.

* On systems where both the select system call and FIONREAD ioctl were
  indicated, take select in preference.  On Irix, where FIONREAD was
  only apparently available, CLISP went into a `function call not 
  implemented' OS error loop.  Symptom also could occur in Emacs (e.g. Ilisp).

Other modifications
-------------------

* Upgraded gettext support to version 0.10.  This provides locale name
  aliasing like with X Windows.

* Upgraded to autoconf-2.7.

* Specify bash for makefile.developer's makemake.
  Some shells may convert backslash sequences to control codes --
  the prebuilt dos, os2, etc. makefiles were previously corrupted.

* The top-level CLISP `configure' can be used in a more GNU-like
  fashion.  Specifically, there is a `--srcdir' option, with which
  `makemake' runs by default.