jjgarcia
3dd7d345d1
By keeping more bytes from the hash key in a package hash table we spare more calls to string_eq().
2004-06-13 08:54:14 +00:00
jjgarcia
8659e4e4e7
Fixed and extended the implementation of RUN-PROGRAM.
2004-06-13 08:53:36 +00:00
jjgarcia
428c31794f
Preliminary version of RUN-PROCESS
2004-06-11 20:12:33 +00:00
jjgarcia
d47c45ee6c
MinGW32 does not have asinh/acosh/atanh
2004-06-11 15:10:45 +00:00
jjgarcia
32ae214c26
MingW does not like 'ERROR' as goto label.
2004-06-11 14:00:21 +00:00
jjgarcia
5fa8787f6e
Both in CLISP and CMUCL, #\Linefeed is printed as #\Newline.
2004-06-08 14:46:10 +00:00
jjgarcia
47893c22d5
*** empty log message ***
2004-06-08 14:43:53 +00:00
jjgarcia
0bd3f15ca8
*** empty log message ***
2004-06-08 07:52:09 +00:00
jjgarcia
24b91659b2
C-{INT,LONG,UINT,ULONG}-{MIN,MAX} could exceed the limits of a fixnum and thus have to be created at boot time.
2004-06-08 07:14:18 +00:00
jjgarcia
24688d57f8
Teach the compiler how to produce C functions with a fixed number of arguments when the corresponding lisp function does not have any &optional, &rest, &key. Teach it also how to produce slimmer code for MULTIPLE-VALUE-SETQ/BIND. The type of a C1FORM can now be a (VALUES ...) expression, the first value being obtained with C1FORM-PRIMARY-TYPE.
2004-05-17 07:32:52 +00:00
jjgarcia
67eb87d545
write_ch = writec_stream when there is no pretty printer
2004-05-12 19:23:36 +00:00
jjgarcia
468c5e0396
*** empty log message ***
2004-05-12 14:55:04 +00:00
jjgarcia
db9bc0763a
Recast some function pointers as (void*) to solve some problems with the C++ compiler.
2004-05-12 08:48:02 +00:00
jjgarcia
3246405dcc
Imported the PRETTY-PRINTER from CMUCL and integrated it with the previously commited format.lsp. The old pretty printer remains merged in and is used by default.
2004-05-12 08:07:53 +00:00
jjgarcia
4e3189eddd
Big changes in the way functions are compiled, unifying the code that handles DEFUN, DEFMACRO and LAMBDA, and fixing an important bug in the optimizer for tail-recursive calls.
2004-05-05 08:38:07 +00:00
jjgarcia
959180c2bd
Protect the initialization routines of the lisp files with the prefix ECL, so that they do not collide with files supplied by the user
2004-04-27 08:30:30 +00:00
jjgarcia
20e3561ae0
Label with no statement. GCC complains.
2004-04-26 11:22:16 +00:00
jjgarcia
9eddd9fb08
All non-graphic characters below (code-char 128) have now character names from the ASCII standard
2004-04-21 13:42:25 +00:00
jjgarcia
4b480b2cec
Some opcodes set the output in the VALUES stack, and cannot use REG0.
2004-04-08 14:02:21 +00:00
jjgarcia
213ef3de8e
MULTIPLE-VALUE-SETQ now uses (SETF (VALUES ...)) when some of the places to be set is not a variable, but a generalized place.
2004-04-07 16:39:31 +00:00
jjgarcia
c503e0fc95
The different EPSILONs are now computed at compilation time, because the values provided by the C library are not all that precise.
2004-04-07 15:29:21 +00:00
jjgarcia
daa8064584
+ Streams in ECL can now have element types CHARACTER, (UNSIGNED-BYTE 8)
...
and (SIGNED-BYTE 8). LISTEN and PEEK-CHAR had to be modified so that
they work with CONCATENATED-STREAMS. OPEN now signals errors of type
FILE-ERROR when a file cannot be opened.
+ Several special functions (ASIN, ACOS, ASINH, ...) have been reimplemented
using more stable formulas found in CMUCL, plus the functions from the C
mathematical library. They also serve as examples of how to use FFI:C-INLINE.
+ Several errors regarding packages have been rewritten as correctable errors,
and now the carry clearer messages.
+ In functions dealing with sequences, :COUNT can now be a bignum.
+ In the bytecodes compiler, MULTIPLE-VALUE-PROG1 did not save the values.
2004-04-05 07:21:54 +00:00
jjgarcia
62d43f4892
Preliminary implementation of the UFFI.
2004-03-29 11:13:17 +00:00
jjgarcia
403511c488
The portable CLX library has been incorporated.
2004-03-25 09:02:27 +00:00
jjgarcia
72dc89f61b
Fixed generation of function prototypes, and other glitches so that the lisp code can be compiled with a C++ compiler. The core library is still compiled with a C compiler.
2004-03-03 09:18:20 +00:00
jjgarcia
21e7d900a7
Implemented a safer routine to retrieve the pointer of a foreign data structure.
2004-03-01 08:42:33 +00:00
jjgarcia
0d9781668f
*** empty log message ***
2004-02-29 16:45:03 +00:00
jjgarcia
860dbd69ec
Fixes due to type sizes, updates of the old garbage collector and the terrible bug with hash tables merged into main trunk
2004-02-29 15:47:01 +00:00
jjgarcia
71f921ed39
Remove some warnings from GCC
2004-01-24 17:00:24 +00:00
jjgarcia
feb3467d3d
Remove unused header
2004-01-20 08:54:01 +00:00
jjgarcia
0d863a2a66
Most functions that operate with the native filesystem must complain if they get get pathnames with wild components
2004-01-16 15:02:51 +00:00
jjgarcia
e8b1a4b4c5
LOAD-LOGICAL-PATHNAME-TRANSLATIONS now implemented (uses sys:host.translation)
...
WILD-PATHNAME-P implemented.
PROBE-FILE and TRUENAME now complain when passed a pathname with wild components.
Fixed a bug in GENTEMP.
Some missing symbols are now defined (*PRINT-MARGIN*, etc), even though not used.
Destructuring lambda lists now identify NIL or '() as an empty list.
Fixed INTEGER-LENGTH and LOGCOUNT (they now work with bignums).
SYMBOL-MACROLET now complains about redefinition of global variables.
ENSURE-DIRECTORIES-EXIST now outputs the two requird values and produces some informative messages if requested.
2004-01-16 10:31:04 +00:00
jjgarcia
a5af1f450c
Implemented LOAD-LOGICAL-PATHNAME-TRANSLATIONS. TRUENAME now merges the pathname with the current working directory.
2004-01-15 19:16:21 +00:00
jjgarcia
710e6e2888
Fix the way SUBLIS and NSUBLIS handle the key and test arguments.
2004-01-15 18:34:47 +00:00
jjgarcia
a6f856e73b
When merging pathnames that have versions, there is a complicated rule for defaulting
2004-01-15 14:19:30 +00:00
jjgarcia
63e1c1789f
Implemented WILD-PATHNAME-P
2004-01-15 13:30:46 +00:00
jjgarcia
90174a13e1
Fixed typo
2004-01-12 13:50:40 +00:00
jjgarcia
19d3e6ace5
Missing header for memcpy()
2004-01-12 13:49:37 +00:00
jjgarcia
13055827c8
Simplify the checks.
2004-01-12 10:54:31 +00:00
jjgarcia
7802543f2d
Small optimization
2004-01-09 13:46:37 +00:00
jjgarcia
738b711ea2
Fixed typo
2004-01-09 13:10:33 +00:00
jjgarcia
02ed9d6468
Fixed several typos.
2004-01-09 12:30:39 +00:00
jjgarcia
1884126793
Fixed typo.
2004-01-08 17:15:39 +00:00
jjgarcia
0eae764528
Do not set *DEFAULT-PATHNAME-DEFAULTS* because some people expect it to be empty.
2004-01-08 11:23:50 +00:00
jjgarcia
90a0382ae8
+ Fixes in the routines that deal with pathnames: it is now possible to
...
use *DEFAULT-PATHNAME-DEFAULTS* as current working directory, because it
is set to the initial directory and because it can be updated with SI::CHDIR.
+ Pathname versions are now properly handled and printed. At least for logical
pathnames, because physical namestrings cannot have version numbers.
+ An error is signaled when one tries to print _readably_ a pathname that does
not have a namestring such that parsing this namestring produces the same
pathname. This happens, for instance, with pathnames that do not have hosts,
or that have version numbers and are not logical.
+ Fixes in the CLOS subsystem: ADD-METHOD and REMOVE-METHOD now update the
GENERIC-FUNCTION slot of method objects, and output the right values.
It is now possible to use two nested SYMBOL-MACROLET that define the same
symbols.
2004-01-08 11:03:05 +00:00
jjgarcia
20564352dd
ASH was bitten by a C feature: that numbers cannot be right shifted (>>) more than the number of bits they have. Otherwise the result is unspecified, and in intel bogus (due to the use of SAR).
2003-12-19 16:35:42 +00:00
jjgarcia
f99b39e820
Fixed typos.
2003-12-19 14:09:03 +00:00
jjgarcia
0772b295cc
Signal also an arithmetic error when dividing by a floating point 0
2003-12-19 13:00:29 +00:00
jjgarcia
d0eeede5e0
(LCM) => 1
2003-12-19 10:21:45 +00:00
jjgarcia
99e1acec1c
CLASS-OF is now in the C core, and the parameter specializers of a method are always either class objects, EQL specializers or NIL.
2003-12-19 10:10:13 +00:00
