/[cmucl]/src/general-info/release-19b.txt
ViewVC logotype

Contents of /src/general-info/release-19b.txt

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.28.2.4 - (show annotations)
Tue Jul 26 15:49:13 2005 UTC (8 years, 8 months ago) by rtoy
Branch: release-19b-branch
CVS Tags: release-19b-pre2
Changes since 1.28.2.3: +5 -0 lines
File MIME type: text/plain
Update.
1 ========================== C M U C L 19 b =============================
2
3 The CMUCL project is pleased to announce the release of CMUCL 19b.
4 This is a major release which contains numerous enhancements and
5 bugfixes from the 19a release.
6
7 CMUCL is a free, high performance implementation of the Common Lisp
8 programming language which runs on most major Unix platforms. It
9 mainly conforms to the ANSI Common Lisp standard. CMUCL provides a
10 sophisticated native code compiler; a powerful foreign function
11 interface; an implementation of CLOS, the Common Lisp Object System,
12 which includes multimethods and a metaobject protocol; a source-level
13 debugger and code profiler; and an Emacs-like editor implemented in
14 Common Lisp. CMUCL is maintained by a team of volunteers collaborating
15 over the Internet, and is mostly in the public domain.
16
17 New in this release:
18
19 * Newly supported platforms:
20
21 - CMUCL has been ported to Mac OS X. The port includes support for
22 callbacks from foreign code, and linkage-table support (which
23 makes it possible to dump images that contain references to
24 foreign code).
25 - the NetBSD/x86 platform is supported, with generational
26 conservative garbage collection (the :gencgc *feature*),
27 linkage-table support, and the Motif interface.
28
29
30 * Feature enhancements:
31
32 - Added a much faster and much less consy bignum printer and a
33 faster bignum reader.
34 - Extended REQUIRE to support *MODULE-PROVIDER-FUNCTIONS* from
35 SBCL.
36 - SIMPLE-STREAM fixes:
37 o READ-SEQUENCE and WRITE-SEQUENCE for
38 signed-byte/unsigned-byte 8 streams.
39 o Memory-mapped file streams.
40 - Large file support for Solaris, so CMUCL can handle files
41 larger than 2GB.
42 - A much faster GCD implementation.
43 - A faster Karatsuba multiplier when the number of words of the
44 bignums is not a power of 2. (Break-even is still around 1500
45 bits, though.)
46 - New command-line switch -debug-lisp-search to show how lisp is
47 searching for its core file.
48 - DEBUG-RETURN implemented which allows you to return a value
49 from function. Only enabled with DEBUG 3.
50 - Lazy sxhash on symbols implemented for ppc.
51 - Pretty-printer for LOOP.
52 - Added some more pprinters for various forms.
53 - Loop analysis code added to the compiler. It's incorporated
54 into the build and can be enabled by setting *loop-analyze* to
55 T, but the compiler doesn't use the loop information for
56 anything yet.
57 - Errors from socket functions signal a SOCKET-ERROR instead of
58 a generic ERROR condition.
59 - Efficiency notes are no longer printed if inhibit-warnings >=
60 2.5.
61 - Simple support for datagram networking added: INET-RECVFROM,
62 INET-SENDTO, INET-SHUTDOWN.
63 - MAKE-INDENTING-STREAM and INDENTING-STREAM-P are exported from
64 the EXTENSIONS package instead of being internal.
65 - TRACE will work now for the known-return convention, which
66 used to require encapsulation so that only the first call was
67 traced. Useful for tracing recursive functions.
68 - Bivalent streams are supported. Thus, streams with
69 element-type (unsigned-byte 8) can be used for character
70 input/output as well as binary input/output. Supply the
71 option :CLASS 'EXT:BINARY-TEXT-STREAM for OPEN to get bivalent
72 streams. EXT:READ-VECTOR and EXT:WRITE-VECTOR also
73 implemented. This is somewhat experimental right now, and may
74 change slightly in a future version.
75 - Loading fasls compiled for a different version of CMUCL will
76 now signal an EXT:INVALID-FASL error instead a generic error.
77
78 * Numerous ANSI compliance fixes:
79
80 - FILE-AUTHOR should signal errors appropriately.
81 - CMUCL was computing the wrong value for some special functions
82 when the argument was on the branch cut for the function.
83 - MULTIPLE-VALUE-BIND accepts &REST and friends in the var list.
84 - When constructing logical pathnames, all components will be
85 automatically upcased. Needed for print/read consistency, and
86 fixes an issue with ASDF and logical pathnames.
87 - Numerous fixes to the float printer, including a new float
88 printer.
89 - Justification (~< ~> FORMAT directives) works better.
90 - ~V with the value of NIL in format strings is treated to mean
91 as if it weren't given at all, as specified by ANSI CL.
92 - Characters print consistently with ~:C and ~S.
93 - Numerous bug fixes for pretty-printing and format strings.
94 - Pathnames are now printed with #P instead of #p.
95 - The types COMPLEX and (COMPLEX REAL) are the same now.
96 - UPGRADED-COMPLEX-PART-TYPE returns RATIONAL for integer types
97 now. (CMUCL doesn't really have a (COMPLEX INTEGER) type).
98 - When a pathname can't be printed readably, signal a
99 PRINT-NOT-READABLE condition.
100 - Printing arrays readably was sometimes wrong.
101 - FOR-AS-ON-LIST works with dotted lists.
102 - DEFSTRUCT predicates for list structures don't produce errors
103 anymore on dotted or short lists.
104 - UNBOUND-SLOT condition no longer takes the :slot keyword. It
105 uses the ANSI-specified :name keyword. This is not backward
106 compatible!
107 - MAKE-SYMBOL takes STRINGs, not just SIMPLE-STRINGs.
108 - PPRINT-POP did not return NIL when the list arg to
109 PPRINT-LOGICAL-BLOCK was NIL.
110 - Signal appropriate error conditions instead of a plain error
111 condition in many situations.
112 - TYPECASE was treating the type T as an otherwise clause even
113 though it was not the last clause. It should be just a
114 regular clause.
115 - ASIN, ACOS, ACOSH, and ATANH were returning the wrong values
116 when the arg was a real number on the branch cut.
117 - Missed some places where loop variables were duplicated, which
118 should signal a program-error.
119 - LOOP INITIALLY clauses occur after variable initialization for
120 FOR, WITH, and AS clauses, as required by ANSI CL.
121 - Format bugs in ~{ and ~@[X~], as noted by ansi-tests, have
122 been fixed.
123 - Evaluation order of REMF, INCF, and DECF.
124 - When *READ-SUPPRESS* is T, READ, READ-PRESERVING-WHITESPACE,
125 READ-DELIMITED-LIST, and READ-FROM-STRING now return NIL, as
126 specified by ANSI CL. (Previously, CMUCL would return a list
127 of NILs.)
128 - Pathname components of type :UNSPECIFIC are no longer printed in
129 namestrings.
130 - Dumping of symbols in fasls now works according to ANSI.
131 - READ-FROM-STRING was sometimes not returning the correct
132 second value when the string was a displaced string.
133 - The invalid trait for constituent characters is handled
134 better.
135 - (formatter "~V[...~]") now returns a correct second value.
136 - MAKE-DISPATCH-MACRO-CHARACTER is supposed to return T.
137 - APROPOS and APROPOS-LIST no longer take 3 arguments. ANSI CL
138 says they only take 2.
139 - A reader error is signaled when bad ratios (like 1/0) are read
140 or when a float can't fit into the specified float type.
141 - A form like #(1 . 2) now signals a reader error.
142 - NIL as a keyword arg is no longer silently accepted, unless
143 NIL was actually a specified keyword arg.
144 - Some cases of reader errors being signaled when
145 *READ-SUPPRESS* was T no longer occur.
146 - The datum slot for check-type contains the place-value now
147 instead of the place.
148 - The expected-type and datum should now be consistent in
149 type-errors (datum is not of the expected-type).
150 Similar situations in other errors are also corrected.
151
152 * Numerous bugfixes:
153
154 - FFLOOR and friends preserve the sign of the arg, so -0.0 is
155 returned for negative args.
156 - EXPT sometimes returned NaN instead of 1 when the power was 0.
157 - Weak pointers no longer have O(n^2) GC behavior.
158 - X11 authorization changes to solve problems seen with servers
159 that don't listen to TCP connection requests.
160 - UNIX-READ fix
161 - The compiler would give an error on deriving the result type
162 of RANDOM when the arg to RANDOM was a union type.
163 - "~*~" was not being parsed to mean a wildcard version.
164 - UNIX-GETPWUID implemented for Darwin. Needed for FILE-AUTHOR.
165 - The compiler no longer gives errors when deriving the result type of
166 COERCE when converting a integer to a float that would not fit
167 in a float.
168 - Fix bug in reading from streams with element-type SIGNED-BYTE
169 for sizes greater than 32: Positive numbers were incorrectly
170 being returned as negative.
171 - REMOVE-DUPLICATES was not honoring the :START option.
172 - PPRINT-LOGICAL-BLOCK sometimes didn't close the block properly
173 because PPRINT-EXIT-IF-LIST-EXHAUSTED caused an early exit.
174 - Macro expansion was not looking in the local environment
175 properly for local definitions.
176 - Circular weak pointers can now be printed when *PRINT-CIRCLE*
177 is T.
178 - "foo.lisp.*" is interpreted to have version :wild, just like
179 "foo.lisp.~*~".
180 - DIRECTORY will report all versions of a file if the pathname
181 has version :wild.
182 - Fixed bug with CLX on MacOS X where the WM-SIZE-HINTS were
183 larger than CLX expected. Replaced card16 with card32.
184 - Callbacks on sparc that returned double-floats were sometimes
185 causing a sigbus error.
186 - Type specifiers like (float low high) where low and high were
187 outside the range of single-float numbers no longer cause
188 compiler errors.
189 - Fix source-form reporting in the disassembler.
190 - Arrays in fasl files had non-NIL fill-pointer. Make them NIL.
191 - FILE-POSITION was incorrectly set for simple-streams after a
192 READ-VECTOR.
193 - Printing 1d23 no longer causes an error.
194 - Work around the problem of (concatenate 'fixnum '(1)) causing
195 deletion of all following code.
196 - A directory with a component of :WILD was not correctly
197 handled.
198 - CREATE-INET-LISTENER was missing a call to htonl.
199 - (coerce #c(1 2) '(complex float)) works now.
200 - Type derivation for logior and logxor was broken.
201 - Fixed bug where (setf (aref f 0) (aref d 0)) doesn't produce a
202 warning and causes code including the function return to be
203 deleted, where f is a single-float array and d is a
204 double-float array.
205 - COUNT for bit arrays was sometimes failing to include the
206 count from the last 32-bits of the array.
207 - The default heap size was causing CMUCL to fail on Solaris
208 10. Change the default so this doesn't happen. (The real
209 cause is unknown.)
210 - In some cases, the compiler was spending large amounts of time
211 deriving union of a bunch of disjoint integer intervals. If
212 the union is long enough, we give up and just return the
213 smallest new interval that contains all of the disjoint
214 intervals. (This needs a better fix.)
215 - The :WHEREIN option to TRACE handles methods now.
216 - The maximum dynamic space size was wrong on Linux.
217 - Type inferencing for OPEN was broken when :IF-DOES-NOT-EXIST
218 was given and was not a compile-time constant.
219 - ~G output always had a zero for the last digit when
220 exponential form was used. Fix this so the last digit has the
221 correct digit.
222 - Output to (make-two-way-stream (make-concatenated-stream)
223 (make-broadcast-stream)) silently discards output instead of
224 producing an error.
225 - STRING<, STRING<=, STRING>, STRING>= no longer produce
226 incorrect results.
227 - COPY-SEQ sometimes produced erroneous results because the
228 compiler was confused by the derived result type of COPY-SEQ.
229 - Many (all?) compiler failures about a STRANGE-TEMPLATE-FAILURE
230 should be fixed.
231 - Type derivation for LOGAND is fixed. (Affects LOGAND and
232 LOGBITP.)
233 - Type derivation for SCALE-FLOAT is fixed. Should also work
234 better on x86. (But there are still bugs in x86's handling of
235 single-floats.)
236 - generic-site.lisp won't give an error when updating
237 *HERALD-ITEMS* when the -quiet flag is used.
238 - The disassembler handles closures better.
239 - The result array for bit-* operations was sometimes too short
240 if the args had fill pointers.
241 - Weak hash tables no longer cause GC to go into an infinite loop.
242 - (PEEK-CHAR nil s nil foo) doesn't remove the character
243 that was read if FOO happens to match the character that
244 PEEK-CHAR returns.
245 - PROFILE sometimes crashes when an internal data structure was
246 set to a non-boolean value.
247
248 * Other changes:
249
250 - Infinite loops in map and friends caused by specifying
251 sequence types that are too hairy for CMUCL should no longer
252 occur. But CMUCL will now signal an error, which may not be
253 compliant either.
254 - Compiler is a little smarter in deriving the result type of
255 MAP.
256 - MAKE-PATHNAME will accept a :HOST arg of NIL, and will use NIL
257 as the host slot instead of filling in the host slot from the
258 defaults.
259 - CMUCL/sparc should now run on Solaris 10. This port no longer
260 uses the reserved register %g7 anymore. The heap size is now
261 rounded to a multiple of 8 MB. Without this, CMUCL eventually
262 gets a GC error on Solaris 10 when using gencgc. There appear
263 to be no problems with Cheney GC on Solaris 10.
264 - The config script uses REQUIRE instead of LOAD to load up
265 CMUCL's system libraries so there are no package-lock errors.
266 - REQUIRE no longer disables package-locks when loading
267 requested file. But package-locks are disabled for CMUCL's
268 own system libraries like gray-streams, clx, etc.
269 - CMUCL should fail gracefully on x86 if something is already
270 mapped at addresses CMUCL wants for itself. Previously, CMUCL
271 would just use it.
272 - Due to some bugs in modular arithmetic, generating either
273 wrong code or much slower code, you can now disable modular
274 arithmetic by setting c::*enable-modular-arithmetic* to NIL.
275 This is a workaround pending a better solution.
276
277 * Improvements to the PCL implementation of CLOS:
278
279 - AMOP-style custom slot allocation supported.
280 - The value of a slot that is specified as shared in the old
281 class and as local in the new class is now retained.
282 - CMUCL supports using the class name for the :metaclass
283 argument.
284
285 * Changes to rebuilding procedure:
286
287 - Added a new script to run all of the subscripts needed to
288 build CMUCL. Won't handle all cases, but will handle the most
289 common builds.
290
291
292
293 This release is not binary compatible with code compiled using CMUCL
294 19a; you will need to recompile FASL files.
295
296 See <URL:http://www.cons.org/cmucl/> for download information,
297 guidelines on reporting bugs, and mailing list details.
298
299
300 We hope you enjoy using this release of CMUCL!
301

  ViewVC Help
Powered by ViewVC 1.1.5