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

Diff of /src/general-info/release-19f.txt

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.6.2.3 by rtoy, Fri Jun 27 17:24:14 2008 UTC revision 1.6.2.3.2.3 by rtoy, Wed Mar 18 15:37:29 2009 UTC
# Line 1  Line 1 
1  ========================== C M U C L  19 f =============================  ========================== C M U C L  19 f =============================
2    
 [--- WORK IN PROGRESS ---]  
   
3  The CMUCL project is pleased to announce the release of CMUCL 19f.  The CMUCL project is pleased to announce the release of CMUCL 19f.
4  This is a major release which contains numerous enhancements and  This is a major release which contains numerous enhancements and
5  bugfixes from the 19e release.  bugfixes from the 19e release.
# Line 21  New in this release: Line 19  New in this release:
19    
20    * Feature enhancements:    * Feature enhancements:
21      - ~R supports many more cardinal names.      - ~R supports many more cardinal names.
22        - Updated network support:
23          o Added BIND-INET-SOCKET to bind a socket to a local address.
24          o Added OPEN-NETWORK-STREAM to create a stream connected to a given host.
25          o Added ACCEPT-NETWORK-STREAM to create a stream connected to
26            the new network connection
27          o Updated CONNECT-TO-INET-SOCKET to allow binding the newly
28            created socket to a local address
29        - Added UNIX:UNIX-OPENPTY, an interface to the openpty C library
30          function.
31        - SSE2 support added for x86.
32          o CMUCL automatically detects whether sse2 is supported or not
33            and loads up the appropriate core file.
34          o New -fpu switch allows the user to specify explicitly which
35            core should be used.  The valid values are x87, sse2, or auto
36            (the default).
37          o Fasls compiled with sse2 support have the extension sse2f.
38            Otherwise, the normal x86f extension is used.  This allows the
39            user to do tests/experiments with both x87 and sse2 without
40            having to mess around with different directories and removing
41            fasls before building for a different FPU.
42          o If the chip supports sse2, but CMUCL can't find the sse2 core,
43            CMUCL will try to fall back to the x87 core.  (This only
44            happens if -fpu is auto.)
45        - Command line parsing now recognizes the option "--".  Everything
46          after "--" is not subject to CMUCL's command line parsing, and
47          everything after the "--" is placed in the new variable
48          EXT:*COMMAND-LINE-APPLICATION-ARGUMENTS*.
49    
50    * ANSI compliance fixes:    * ANSI compliance fixes:
51      - Fix bug in backquote printer.  If the variable is @foo, we want      - Fix bug in backquote printer.  If the variable is @foo, we want
52        to print ", @foo" not ",@foo".  Similarly, for .foo, we want to        to print ", @foo" not ",@foo".  Similarly, for .foo, we want to
# Line 35  New in this release: Line 60  New in this release:
60        merging the new-name with the truename of filespec.  Also, a        merging the new-name with the truename of filespec.  Also, a
61        logical pathname should be returned if new-name is a logical        logical pathname should be returned if new-name is a logical
62        pathname.        pathname.
63        - Character names need to be a capital letter followed by lower
64          case.  Needed to match what ~:C does.  (Found via ansi-tests).
65    
66    * Bugfixes:    * Bugfixes:
67      - Compiler can now derive the rank of an array, even if the array      - Compiler can now derive the rank of an array, even if the array
# Line 68  New in this release: Line 95  New in this release:
95        before.        before.
96      - The small bias in RANDOM for integer args up to 32 bits long      - The small bias in RANDOM for integer args up to 32 bits long
97        should now be gone.        should now be gone.
98        - Improved type derivation for LOGAND when one arg is bounded but
99          the other is not.
100        - Some issues with tracing on sparc and ppc have been fixed.  This
101          usually manifests itself with a segfault just after the function
102          result is printed.
103        - Fixed bug on sparc where C-c sometimes causes a segfault.  We
104          now handle the case where siginfo_t is NULL, which can also
105          happen on other architectures.
106        - The interpreter catches invalid EVAL-WHEN situations just like
107          the compiler, instead of silently ignoring them.
108        - FLOAT-PRECISION supports double-double floats.
109        - Tracing should now be working on Darwin/x86.  Previously,
110          certain cases would cause Lisp to segfault in bad ways where you
111          could not return to the repl.  Do not need to do encapsulation
112          by default anymore.
113        - The bounds for type (REAL lo hi) are computed better now.  The
114          REAL type is a union of SINGLE-FLOAT, DOUBLE-FLOAT, and
115          DOUBLE-DOUBLE-FLOAT.  The computed bounds for
116          DOUBLE-DOUBLE-FLOAT only had double-float accuracy if the bound
117          for REAL was a rational.
118        - The FLOAT type now requires that the bounds, if given, are
119          floats.  Previously, any real type would be accepted.  This
120          makes FLOAT behave like SINGLE-FLOAT and DOUBLE-FLOAT which
121          required the bounds to be a float of the appropriate type.
122    
123    * Trac Tickets:    * Trac Tickets:
124      - #16: Read-time hash-table issue      - #16: Read-time hash-table issue
125        Fixed.        Fixed.
126        - #17: LOOP NAMED NIL has no effect
127          Fixed.
128        - #18: Modular arith bug 1
129          Fixed
130        - #19: Modular arith bug 2
131          Fixed by not doing modular arith if the args are known to be
132          fixnums.
133        - #20: Modular arith bug?
134          Fixed via the fix for Trac #21.  The original workaround has
135          been removed.
136        - #24: Float contagion for expt
137          Float contagion is applied to the arguments before computing
138          expt.
139        - #21: Modular arith bug 3
140          Fixed by delaying the logand defoptimizer.
141        - #15: x86 double-float issue
142          Fixed when using SSE2 support.  We will not fix this for x87.
143        - #25: Compiler bug
144          Fixed.
145        - #26: slot-value type check
146          Fixed for some cases.  When used in methods, slot-value may not
147          do the type check.if the object is not a argument to the method.
148        - #29: make-condition doesn't accept class objects
149          Fixed.
150    
151    * Other changes:    * Other changes:
152      - IS1, IS2, IS3, and IS4 are recognized character names for the      - IS1, IS2, IS3, and IS4 are recognized character names for the
153        ASCII control codes US, RS, GS, FS, respectively.        ASCII control codes US, RS, GS, FS, respectively.
154      - Added OPEN-NETWORK-STREAM and ACCEPT-NETWORK-STREAM functions.      - Added OPEN-NETWORK-STREAM and ACCEPT-NETWORK-STREAM functions.
155        - When initializing a random state, try to read 627 words from
156          /dev/urandom to initialize the entire state vector with random
157          bits.  Previously, only one word was read.
158        - A seed of 0 is allowed in KERNEL:INIT-RANDOM-STATE.
159        - Updated User guide to include more examples of tracing.
160        - Enable gencgc page protection on x86/darwin.  This can speed up
161          GC a bit.  (Not measured.)
162        - Bignum truncate is significantly faster.  Some cl-bench
163          benchmarks are now almost twice as fast.
164        - The continuable error produced by raising an integer to a power
165          exceeding *intexp-maximum-exponent* is now a restart, giving the
166          user the option to continue and update the limit to the new
167          power.
168        - The Darwin/x86 port can run on Mac OS X 10.4 or later.
169    
170    * Improvements to the PCL implementation of CLOS:    * Improvements to the PCL implementation of CLOS:
171      - The compiler and interpreter should handle SLOT-VALUE the same      - The compiler and interpreter should handle SLOT-VALUE the same
# Line 86  New in this release: Line 175  New in this release:
175        slot-value-using-class have been fixed.        slot-value-using-class have been fixed.
176        Get-accessor-method-function was causing an error to be signaled        Get-accessor-method-function was causing an error to be signaled
177        incorrectly.        incorrectly.
178        - (setf (slot-value <obj> <slot>) <new>) will now signal an error
179          in some situations when the new value is not of the correct
180          declared type for slot.
181    
182    * Changes to building procedure:    * Changes to building procedure:
183        - For Linux, custom CFLAGS, CC, and LDFLAGS are supported.
184          Requested by Stelian Ionescu for Gentoo support.
185        - The FreeBSD config file is now named Config.x86_freebsd, which
186          is equivalent to Config.FreeBSD_gencgc, which is deprecated.
187        - The Linux config file is now named Config.x86_linux, which is
188          the equivalent to Config.linux_gencgc, which is deprecated.
189    
190      * Known issues:
191        - Executables (via save-lisp) on recent FreeBSD versions are
192          currently not working.  The executable is created, but the
193          resulting executable doesn't work correctly.
194    
195  This release is not binary compatible with code compiled using CMUCL  This release is not binary compatible with code compiled using CMUCL
196  19e; you will need to recompile FASL files.  19e; you will need to recompile FASL files.

Legend:
Removed from v.1.6.2.3  
changed lines
  Added in v.1.6.2.3.2.3

  ViewVC Help
Powered by ViewVC 1.1.5