Changes in 6.8.3:
* The arguments to ghc-pkg now understand lists and wildcards, and
there is a new command ghc-pkg find-module.
* The filename completion in ghci is now more intelligent.
* We now look for GHCi config files in these locations:
./.ghci
$HOME/.ghc/ghci.conf
$HOME/.ghci
* You can now give multiple -e options to GHC.
* You can now use the -prof and -threaded flags together, although
you cannot also use +RTS -N2.
* SCC names are no longer allowed to contain spaces, as some of the
profiling tools break if they do.
* Various changes have been made to GHC's internals, so there are some
differences in the API exposed by the ghc package. Most notably,
checkAndLoadModule has now been implemented.
- Fixed infinite loop when a character is output to a port
whose character encoding does not support the character (e.g. (display
(integer->char 1000)) when character encoding is ISO-8859-1).
- Added procedures to extract the threads and thread
groups contained in a thread group (thread-group->thread-list,
thread-group->thread-vector, thread-group->thread-group-list,
thread-group->thread-group-vector).
- Renamed open-process' show-window: setting to the more
reasonable show-console: since it controls the visibility of the
console.
- Added readtable-max-unescaped-char and
readtable-max-unescaped-char-set which control the external syntax of
characters in symbol, string and character objects written with the
write and pretty-print procedures.
- Added tcp-server-socket-info to get the IP address and
port-number of a tcp-server-port (useful when the server port was
created with a kernel assigned port-number, i.e. setting port-number:
0).
- Fixed incorrect calls to ___P macro in lib/*.c, and added
lib/check___P script to automatically detect such calls (script
contributed by Derek Peschel).
- Added a "terminate" interrupt which terminates the Gambit
process cleanly (all the exit jobs are executed before exiting). On
Unix this interrupt is raised by the SIGTERM signal so a "kill 1234"
where 1234 is the pid of the Gambit process will terminate the process.
On Windows the interrupt occurs when the console window is closed or
the system is shutting down (however currently the system hangs while
executing the exit jobs).
- Fixed handling of incomplete characters at end-of-file,
which are now handled like illegal characters (error signaled unless
char-encoding-errors setting is set to #f).
- Fixed recently introduced bug in handling of end-of-line
encoding.
- Added escaping of non-graphical characters in symbols:
(string->symbol (string (integer->char 255))) gives |\377| .
- Protect gcc extension _builtin_expect with GCC version >= 3
because old gcc versions (up to 2.95.3 it seems) did not have that
extension.
- Added char-encoding-errors: setting to open-file to
enable/disable character encoding error checking, and -:fr and -:fR
runtime options to select the default setting for file I/O (on/off).
- Removed non-ASCII characters from lib/_num.scm .
- Fixed error introduced with end-of-line encoding of the history
file.
- Changed end-of-line encoding from lf to cr-lf when reading
source code. This makes the load procedure and the compiler more lenient
when reading source code with non-Unix-style end-of-line encoding. This
means that source code with a string containing a CR/LF sequence will
construct a string with a single #\newline character (previously both$
#\return and #\newline were put in the string).
- Fixed compiler crash when compiling files containing the forms
#123, #, ##, etc (file location information was not properly attached to
expression)
- Improve error messages produced by the C compiler when it
compiles a file generated by a different version of the Gambit compiler.
Removing patch-ad: Issue fixed upstream.
User visible changes:
* CLISP built natively on 64-bit platforms (i.e., with 64-bit pointers)
now has :WORD-SIZE=64 in *FEATURES*.
See <http://clisp.cons.org/impnotes/features.html> for details.
* Module syscalls now offers OS:ERRNO and OS:STRERROR (for the sake of
FFI modules).
See <http://clisp.cons.org/impnotes/syscalls.html#errno> for details.
* Modules MIT-CLX and NEW-CLX export a new macro XLIB:WITH-OPEN-DISPLAY.
* Module netica has been upgraded to the Netica C API version 3.25 (from 2.15).
See <http://clisp.cons.org/impnotes/netica.html> for details.
* Module libsvm has been upgraded to the upstream version 2.86.
See <http://clisp.cons.org/impnotes/libsvm.html> for details.
* Bug fixes:
+ Work around the absence of tgamma() on solaris.
+ Avoid a rare segfault on SIGHUP.
+ Improve module portability to systems with non-GNU make.
+ Fix GRAY:STREAM-READ-SEQUENCE and GRAY:STREAM-WRITE-SEQUENCE.
+ Fix the remaining bugs in special bindings in evaluated code on
TYPECODES (64-bit) platforms.
+ Fix SOCKET:SOCKET-CONNECT with timeout to a dead port.
Note that the tcl bindings are not (yet) built. Next time...
pkgsrc changes not trivial but should be invisible.
Changelog:
Changes since 1.0 release;
=========================
Added new (experimental) builtin function groups:
"DBM" -- "ndbm" keyed database access
"COM" -- COM/OLE Automation on Win32
"STCL" -- SNOBOL/Tcl interface (optional)
New library "dynamic.sno" to help compile/link
dynamicly loadable extensions.
Added HTTP.POST function to http.sno
Handle environments where pointers can have high order bit set
Linux 2.6 kernels in Fedora Core, SuSE 10 distributions
Avoid broken "install" utilities
autoconf compatibility: Honor CC, CFLAGS, CPPFLAGS,
LDFLAGS, DESTDIR environment variables.
Fix compilation on pre-panther OSX
Fix second argument to DEFINE(): was not case folded!
Fix backtracking over unevaluated expressions
Source &FILE and &LINE shown in TRACE output
Experimental: GC stats & tracing (-g on command line, >RACE keyword)
Completed support for "memory I/O" when CSNOBOL4 used as a library!
Added "PREDICATE" and "SYSPRED" return types in wrapper.sno
Changes since 0.99.44 release;
=============================
Added snolib(3) man page for SNOBOL4 library routines
Treat all bytes with 8th bit set as ALPHA on ASCII systems
allows UTF-8 encoded characters, and national
character sets to used in labels and identifiers!
command line:
-S option to enlarge interpreter stack
configure:
added --mandir & --bindir
IPv6 Support (both client and server)
bi-directional "pipe" I/O on systems with sockets
"privileged" port binding for UDP
Lots of cleanup!
DJGPP (DOS gcc) fixes
default to -O3 with gcc
add C prototypes for all functions
(function declarations still "K&R" style)
MINGW (Win32 gcc) fixes
Fix LOAD() on Darwin (MacOS X)
Fix timing script on Debian
Ported to HP-UX/IA64
VMS: LOAD() support (not tested)
Compilation on VMS7.3
support magic pathnames "/dev/null" and "/dev/tty"
Win32: native (low overhead) support for character console I/O
support magic pathnames "/dev/null" and "/dev/tty"
fixed UDP I/O
Fix for parallel make (GNU make, BSD make)
Compilation under MS Visual Studio .NET 2003
Make struct descr packable (BITFIELDS_SAME_TYPE)
Check for oversize integer constants (ERANGE from strtol())
Changes since 0.99.4 release;
============================
Command line:
-d (dynamic region size) and -M (pattern match stack size)
now take values in descriptors
-v shows CSNOBOL4 version
Increased default dynamic space to 64K descriptors
Fix faulty lexical comparison function optimzations for null string
Performance improvments;
New string hash function
(reduced both compilation and run time of genc.sno by 5%)
Optimized most common case (CONTIN actions) in
"stream" operation used for lexical analysis.
(reduced compilation time of genc.sno by 8%)
Merged all scanner (pattern matching) functions into
single C function (eliminates mutual recursion, stack
overflows)
genc runtime reduced 15%
atn.sno runtime reduced by 63%!! (2.7x faster)
worst case (4096 char string) reduced 69% (4.8x faster)
snocone self-compile runtime reduced 5%
Made pattern building functions (lib/pat.c) inlinable
(8% reduction in genc.sno compile time)
Eliminate space padding on lines read by compiler
(reduced genc.sno compile time by 65%)
Raised default C compiler optimization levels
Default &MAXLNGTH to largest possible string
Documentation:
README file describes available includes
Include PDF of manual page
Implement fatal "Output error" for output, ENDFILE() errors
Configuration script re-write;
Eliminate static config files & autoconf script
configure generates config.m4 and config.h
(with all defines) for use by loadable functions
New configuration options;
**** see "INSTALL" file for more information ****
--add-define=
Add a #define to config.h (visible to external functions)
--double
EXPERIMENTAL option to use C "double" for SNOBOL4 REAL data type
--fast
Architecture specific optimizations on native compilers
--longlong
EXPERIMENTAL!!
try to use C "long long" (64-bit integer) for INTEGER datatype
when longs/pointers are only 32-bits (ILP32 data model).
--lp64
Request use of 64-bit long/pointer data model
if available on this architecture, and not the default.
--no-opt
Compile without optimization
New SNOBOL library files (see README);
compatibility libraries:
fence.sno, logic.sno, not.sno
spitbol.sno, snobol4+.sno
HOST() function codes:
host.sno
URL (http:, ftp:, file:) access functions
Use large files (64-bit offsets) where available. Allows access
and creation of files larger than 2GB
New functions;
FUNCTION() -- predicate to test if a function is defined
** EXPERIMENTAL **
LABEL() -- predicate to test if a label is defined
idea from Steve Duff's version of Macro SPITBOL
ORD() -- inverse of CHAR()
SERV_LISTEN() -- act as TCP server
SSET() -- scaled SET() allow large files when INTEGER is 32bits
VDIFFER() -- returns first arg if two arguments DIFFER()
idea from Steve Duff's version of Macro SPITBOL
Extended HOST() functions (see host.sno) for system diagnosis
HOST() function will accept REAL or integer encoded STRING args
TIME() function now returns REAL datatype allows both better
resolution (for small values) and larger range. Using
32-bit floats, runtimes larger than four and a half
hours will lose resolution.
Statistics reports compiler and execution times as REAL numbers
Average statement runtime reported in microseconds (us).
Allow a REAL any place an INTEGER is required (from SNOBOL4+)
Contexts include TABLE(), ITEM(), array indices,
INPUT(), OUTPUT(), SET(), keyword values, CHAR(),
RPAD(), LPAD(), FIELD(), COLLECT(), DUMP(), DUPL(),
OPSYN(), and SUBSTR()
load.h updates;
all RETxxx macros safe in any context
RETSTR() takes single argument, allows NULL pointer
RETSTR2() has old RETSTR() functionality
RETINT() and RETREAL() always set return data type.
All include files needed to build loadable functions
installed in SNOLIB directory
Include BSD tsort program; allow builds from scratch
on platforms with GNU tsort (Linux, Cygwin)
New ports;
Ported to Darwin (MacOS X)
Ported to Cygwin (http://cygwin.com), a free Unix emulation
environment for Windows.
Attempt at native Win32 port
Updated Borland C port
Updated VMS port; pipe open support, attempt at tty support
Also update the tk dep to no longer demand tk83. It does not seem to
work all that well with tk84, but it works about the same as it does
with tk83. And depend explicitly on tcl; that was missing.
Primary changes since 1.2.1 seem to be GPLv2 -> GPLv3 and
whitespace/reindenting, but there seems to be a bit more in there
too. There's also now a 12 meg kdevelop blob in the distfile, yay.
Changelog:
GNU Sather 1.2.3 - Jul 7, 2007 - Michael R. Taylor
* Made INT 32-bit even on 64-bit systems
* Changed license to (GPLv3/LGPLv3) or later
There's no changelog for 1.2.2, dunno why.
Changes since previous (2.44.1) release:
User visible changes
--------------------
* Experimental Just-In-Time Compilation of byte-compiled closures is now
done using GNU lightning (this is a configure-time option).
Thanks to Yann Dauphin <yann-nicolas.dauphin@polymtl.ca>.
* New command-line option -lp adds directories to *LOAD-PATHS*.
See <http://clisp.cons.org/impnotes/clisp.html#opt-load-paths> for details.
* New function FFI:OPEN-FOREIGN-LIBRARY allows pre-opening of shared libraries.
See <http://clisp.cons.org/impnotes/dffi.html#dffi-open-lib> for details.
* New macro EXT:COMPILE-TIME-VALUE allows computing values at file compilation.
See <http://clisp.podval.org/impnotes/macros3.html#compile-time-value>
for details.
* New function FFI:FOREIGN-POINTER-INFO allows some introspection.
See <http://clisp.cons.org/impnotes/dffi.html#fptr-info> for details.
* Versioned library symbols are now accessible via the :VERSION argument of
DEF-CALL-OUT and DEF-C-VAR.
Thanks to Kaz Kylheku <kkylheku@gmail.com>.
See <http://clisp.cons.org/impnotes/dffi.html#def-call-out> for details.
* New functions GRAY:STREAM-READ-SEQUENCE and GRAY:STREAM-WRITE-SEQUENCE have
been added for portability reasons.
Suggested by Anton Vodonosov <avodonosov@yandex.ru>.
See <http://clisp.cons.org/impnotes/gray.html#st-rd-seq> for details.
* New user variable CUSTOM:*SUPPRESS-SIMILAR-CONSTANT-REDEFINITION-WARNING*
controls whether the redefinition warning is issues when the new
constant value is visually similar to the old one.
See <http://clisp.cons.org/impnotes/defconstant.html#defconstant-similar>
for details.
* REPL commands can now accept arguments.
See <http://clisp.cons.org/impnotes/repl.html> for details.
* Updated the postgresql module to PostgreSQL 8.3.
See <http://clisp.cons.org/impnotes/postgresql.html> for details.
* Module syscalls now interfaces to <stdio.h> (for the sake of FFI modules).
See <http://clisp.cons.org/impnotes/syscalls.html#stdio> for details.
There are others additions there also, but they are too numerous to
be enumerated here.
* Bug fixes:
+ Fix handling of quoted objects by READ-PRESERVING-WHITESPACE. [ 1890854 ]
+ Fix rectangle count in NEW-CLX XLIB:SET-GCONTEXT-CLIP-MASK. [ 1918017 ]
+ Fix argument handling in NEW-CLX XLIB:QUERY-COLORS. [ 1931101 ]
+ Fix compilation on systems not supporting returning void. [ 1924506 ]
+ Fix TANH floating point overflow for large floats. [ 1683394 ]
+ Avoid extra aggressive bignum overflow reporting in READ. [ 1928735 ]
+ Improved floating point number formatting. [ 1790496, 1928759 ]
+ COMPILE no longer discards MACRO doc strings. [ 1936255 ]
+ Improved accuracy of LOG on complex numbers. [ 1934968 ]
+ Fix COERCE for compound float result-types. [ 1942246 ]
+ Fix $http_proxy parsing. [ 1959436 ]
+ Fix LISTEN on buffered streams when the last character was CRLF. [ 1961475 ]
+ Cross-compilation process has been restored to its former glory,
Changes:
* [05 Jun 08] Changed all '#!/bin/bash" in shells scripts into '#!/bin/sh'
and changed all $(...) into `...` because there are compatibilities
problems according to Unix versions.
* [31 Mar 08] Fixed bug: "let f ~_x () = ()" generated syntax error.
* [31 Mar 08] Fixed bug: "x $ y" generated syntax error (normal syntax).
* [02 Jan 08] Added compatibility with OCaml 3.10.3.
This is security fix:
http://www.ruby-lang.org/en/news/2008/06/20/arbitrary-code-execution-vulnerabilities
Fri Jun 20 18:25:18 2008 Nobuyoshi Nakada <nobu@ruby-lang.org>
* string.c (rb_str_buf_append): should infect.
Fri Jun 20 16:33:09 2008 Nobuyoshi Nakada <nobu@ruby-lang.org>
* array.c (rb_ary_store, rb_ary_splice): not depend on unspecified
behavior at integer overflow.
* string.c (str_buf_cat): ditto.
Wed Jun 18 22:24:46 2008 URABE Shyouhei <shyouhei@ruby-lang.org>
* array.c (ary_new, rb_ary_initialize, rb_ary_store,
rb_ary_aplice, rb_ary_times): integer overflows should be
checked. based on patches from Drew Yao <ayao at apple.com>
fixed CVE-2008-2726
* string.c (rb_str_buf_append): fixed unsafe use of alloca,
which led memory corruption. based on a patch from Drew Yao
<ayao at apple.com> fixed CVE-2008-2726
* sprintf.c (rb_str_format): backported from trunk.
* intern.h: ditto.
Tue Jun 17 15:09:46 2008 Nobuyoshi Nakada <nobu@ruby-lang.org>
* file.c (file_expand_path): no need to expand root path which has no
short file name. [ruby-dev:35095]
Sun Jun 15 19:27:40 2008 Akinori MUSHA <knu@iDaemons.org>
* configure.in: Fix $LOAD_PATH. Properly expand vendor_ruby
directories; submitted by Takahiro Kambe <taca at
back-street.net> in [ruby-dev:35099].
* Update RUBY18_VERSION to 1.8.7 and RUBY18_PATCHLEVEL to 17.
* Use vendor_ruby instead for site_ruby.
* Introduce macros for relative path and use them instead of old absolete
path.
RUBY_LIB lib/ruby/${RUBY_VER_DIR}
RUBY_ARCHLIB ${RUBY_LIB}/${RUBY_ARCH}
RUBY_SITELIB_BASE lib/ruby/site_ruby
RUBY_SITELIB ${RUBY_SITELIB_BASE}/${RUBY_VER_DIR}
RUBY_SITEARCHLIB ${RUBY_SITELIB}/${RUBY_ARCH}
RUBY_VENDORLIB_BASE lib/ruby/vendor_ruby
RUBY_VENDORLIB ${RUBY_VENDORLIB_BASE}/${RUBY_VER_DIR}
RUBY_VENDORARCHLIB ${RUBY_VENDORLIB}/${RUBY_ARCH}
RUBY_DOC share/doc/${RUBY_NAME}
RUBY_EG share/examples/${RUBY_NAME}
Old these macros are removed after 2008Q2 branch.
RUBY_LIBDIR
RUBY_ARCHLIBDIR
RUBY_SITELIBDIR
RUBY_SITEARCHLIBDIR
RUBY_VENDORLIBDIR
RUBY_VENDORARCHLIBDIR
RUBY_DOCDIR
RUBY_EXAMPLESDIR
* update PRINT_PLIST_AWK macro to reality and move some of them from
ruby/modules.mk to ruby/rubyversion.mk.
changes:
-new module: `(srfi srfi-88)'
-New `postfix' read option, for SRFI-88 keyword syntax
-Some I/O primitives have been inlined, which improves I/O performance
-New object-based traps infrastructure
-New support for working on Guile code from within Emacs
-bugfixes
pkgsrc note: added a patch which affects builds without thread
support only (fixes crash on termination)
approved by gdt
build the newer version. This should allow the package to be
included in bulk builds, and eliminate the troubles with other
Common Lisp systems being fragile WRT building this system.
Update to 1.0.16
