Do you mean WITH-STANDARD-IO-SYNTAX?
WITH-STANDARD-IO is not in the standard.
Pail
Jeff wrote:
Hi,
I recently started using gcl to learn common lisp, but ran into a
problem or two.
If I'm not mistaken, with-standard-io does not appear to be in the
2.6.7 release, I get the message "Error:
No I didn't. How great. Is that how he came up with some of his ANSI
tests?
There's a file in ansi-tests (misc.lsp) containing tests generated by
the random compiler tester. Each test represents a failure in some
lisp implementation, either free or commercial. The random tester
found bugs in
Frode Vatvedt Fjeld wrote:
[Is this a good place to report problems with ansi-tests?]
It seems to me there's a problem with ansi-tests/rt.lsp. It does
(defvar *entries* '(nil)) and (defvar *entries-tail* *entries*), and
then proceeds to do (setf (cdr *entries-tail*) ..) in the add-entry,
thereb
I have not been able to install gcl built on cvs head,
apparently due to problems building gcl-tk. This is on
an up-to-date system running Fedora Core 4. Is this
a known problem?
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists.gnu
>(compile nil `(lambda () ,(code-char #x1C)))
Error in CONDITIONS::CLCS-COMPILE-FILE [or a callee]: Unexpected end of
file on #.
This is showing up in the random type prop tester.
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists
Camm,
I've run the random tester on gcl again, and found a high
frequency bug involving MULTIPLE-VALUE-SETQ. See test misc.632
in ansi-tests/misc.lsp:
(funcall (compile nil '(lambda () (let (b) (multiple-value-setq (b)
10)
==> NIL (incorrect, should be 10)
This is with --enable-ans
This is from the current cvs head:
bash-2.05b$ ./configure --enable-ansi --prefix=/home/dietz
configure: error: cannot find sources (src/mpi_defglue.lsp) in . or ..
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists.gnu.org/mailma
Camm Maguire wrote:
If not, we need to expand compile and function
to accept arguments in the other lambda forms, that is if the spec
allows. Paul?
COMPILE does not specify what happens if the definition argument
is not a lambda expression or function, which means the behavior
is undefined.
I don't know if this was the problem you were talking about,
but this took over an hour to build on my machine (Athlon XP+ 1800),
and then ended with a READ error in gcl_pcl_impl_low.lisp.
That time is a large slowdown from even the previous build
time (25 minutes on the same machine).
P
Camm Maguire wrote:
Just my thoughts, Feeback most welcome.
subtypep doesn't appear to be particularly slow, as judged by
the speed of the random subtypep tester (it appears to be
faster than sbcl, actually). Incidently, that tester has just
found a bug, which I turned into a test case in sub
Camm Maguire wrote:
Also inlined are: length and reverse, endp and identity are expanded,
the former to throw an error only when safety >=1.
There is an unused heap sort in the code too, but I can't see any gain
here, as the memory requirements are virtually the same. The
quick-sort is non-rec
Camm Maguire wrote:
Greetings!
Paul, could you try just removing the declaration for logl from pltd.h
and restarting make from the top directory?
Ok, that seems to work. Thanks!
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lis
Camm Maguire wrote:
Can you tell me how logl is defined at
/usr/include/bits/mathcalls.h:110: previous declaration of `logl'
Is there a manpage giving its prototype?
I couldn't find the man page. I've attached math.h, and three files
it includes. logl appears to be used in mathinline.h.
Camm Maguire wrote:
Greetings! What is logl? What kind of machine is this? C references
to log apparrently reference both log and logl as external symbols.
No logl on my system.
This is Redhat 9, with gcc 3.2.2.
Paul
___
Gcl-devel maili
Camm Maguire wrote:
Greetings, and thanks! Please see if my latest commit to array.c
fixes this and let me know if problems persist.
Now there's another problem (make output attached).
Paul
(cd o && make ../h/new_decl.h)
make[1]: Entering directory `/home/dietz/gcl/o'
gcc -o grab_d
I've attached the output of make; it dies
about 1 minute into the build.
Paul
(cd o && make ../h/new_decl.h)
make[1]: Entering directory `/home/dietz/gcl/o'
gcc -o grab_defs grab_defs.c
gcc -c -Wall -DVOL=volatile -fsigned-char -pipe -O3 -fomit-frame-pointer
-I/home/dietz/gcl/o -I../h
Camm Maguire wrote:
Greetings! As many may know, GCL autodeclares constant let bindings,
and this allows for many optimizations to proceed. We could also
autodeclare bindings which are changed, but changed to a subtype of
the original type, let alone autodeclaring loop counters after
checking l
Camm Maguire wrote:
Greetings! Have retored connectivity now. Am wondering if no news is
good news :-)
Oh, sorry! It built, with 1302 failures on ansi-tests.
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists.gnu.org/mai
Camm Maguire wrote:
Greetings!
1) I've checked in my recent stuff. Bug list diff below. If you have
a moment, please check that I haven't messed it up through omission
or commission.
Wow! Lots of changes. There's a build problem, though. I've attached
the output of 'make'. It's try
Camm Maguire wrote:
What is the typespec for an improper list?
There isn't a standardized one. I recommend something
like (satisfies si::proper-list-p).
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists.gnu.org/mailman/lis
I changed the setf expander for GETF to resemble
that in SBCL. This is to fix setf-getf.order.2
in ansi-tests/getf.lsp. However, the result
is odd -- it passes when rt::*compile-tests*
is true, but fails when that variable is bound
to nil. This looks like it may be a bug in
gcl's EVAL function.
Camm Maguire wrote:
Greetings!
What is '(array nil) suposed to mean?
It's the type of arrays with element type nil.
These arrays are specialized to hold nothing at all.
Weird, but since (upgraded-array-element-type nil) is required
to be nil, the type has to exist.
It's probably a spec bug t
Camm Maguire wrote:
Yes, it does seem as a construct to aid in array compilation
optimization. The point to me is that I don't see the overwhelming
benefit given the definition, though there may be some. The big issue
in our array optimization is to be able to know the element type, rank
and s
Whoops, I made a mistake reading the page for simple-array. Yes,
you're right, you can do that.
I want to add that you may not *want* to do it, if it forces
array operations to be slow even if the program has a simple-array
declaration.
Paul
Paul F. Dietz wrote:
Camm Maguire wrote:
Greetings! Am I correct in assuming that I can make all arrays simple
arrays?
No. simple-arrays aren't displaced and can't have fill pointers.
Whoops, I made a mistake reading the page for simple-array. Yes,
you're right,
Camm Maguire wrote:
Greetings! Am I correct in assuming that I can make all arrays simple
arrays?
No. simple-arrays aren't displaced and can't have fill pointers.
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.org
http://lists.gnu.org/m
Juho Snellman benchmarked some code he wrote for a Google
programming project, and found gcl wasn't very good at
all:
http://www.cs.helsinki.fi/u/jesnellm/blog/archive/2005-08-27b.html
He told me the biggest problem was gcl wasn't able to compile
defuns in non-null lexical environments:
;; comp
I've checked in some changes, fixing several more bugs.
since fixed two of them (a typep problem and a REMF bug) and
reduced the number of ansi-test failures to 1725 (mostly
from the typep change) and checked in these changes. Should
I be updating the debian changelog for this?
I've been us
I think gcl is getting this right.
,@,@ would mean that the value returned by the innermost
,@ is being used in a position where its value is being
used directly rather than being spliced into a list.
That's undefined.
Paul
Camm Maguire wrote:
Greetings, and thanks for the report!
Th
I've added the file BUGS in the gcl/ directory. I intend
to diagnose the bugs found by ansi-tests and list them here,
so they'll be easier to track and resolve. I've started
with the subtypep bug on structure-object and all the ansi-test
bugs from the 'cons' section of the standard.
Pau
There are 95 test failures in the structure tests that
have a common cause. If the value of S is a structure
type name (or a structure class object), then
(subtypep S 'structure-object)
should return t, t (or, more properly, two non-nil
values). However, it is returning nil, t.
I don't know
Camm Maguire wrote:
Greetings!
test-random-types3 spit this out after a 1 10 run:
(((AND (OR (OR (NOT (AND NIL RATIO NIL))) (COMPLEX (INTEGER * 12800
(NOT (CONS (CONS (CONS T T)
(CONS (COMPLEX (REAL 78460 78460)) (EQL -3905)))
(EQL -285003020)
do
This was found with the test-random-types3:
(deftest subtypep.cons.39
(values (subtypep t '(and (not (cons cons (cons cons t)))
(not (cons t cons)
nil)
(expected value: nil, actual value: t)
Paul
___
Gcl-d
Camm Maguire wrote:
Thanks! I didn't look carefully at typep, (or anything else other
than subtypep really), so if you see anything else please let me know.
I've checked in fixes to those problems, and also
made TYPEP work on type designators that are class objects.
This last change fixed near
Looking over gcl_predlib.lsp, I see some problems with TYPEP:
>(typep nil nil)
T
(Remember, NIL is the empty type; NULL is the type containing NIL.)
Also, typep-int contains this code fragment:
(case tp
(cons (and (consp object) (typep-int (car object) (car i)) (typep-int (cdr
object) (c
I wrote:
I'm getting some
failures from the test-random-types3; I'll reduce them to test cases
and send them in this evening when I have time.
Here's a problem:
CL-TEST>(subtypep t '(or (not (real 0 10)) (not (real -100 -50
NIL;; incorrect
T
CL-TEST>(subtypep t '(not (and (real 0 1
Camm Maguire wrote:
Just a note -- if either of you very experienced lisp users might have
a few cycles to look over the new gcl_predlib.lsp and offer any
comments/suggestions/advice, this would of course be most welcome.
The code still needs commenting and a little polish -- in particular
comple
I've added two more random tester-found bugs to misc.lsp.
misc.613 is a failure involving throws in lambda forms
in REDUCE. The enclosing catch form isn't being seen.
misc.614 is a failure related to dynamic-extent.
;;; Error in APPLY [or a callee]: The tag CT1 is undefined.
(deftest misc.613
Camm Maguire wrote:
Greetings, and thanks! This should be fixed in t4 and later -- I
believe this was one of Mike's mingw options which crept in to the unix
build.
Yes, all fixed. Thanks!
Paul
___
Gcl-devel mailing list
Gcl-devel@gnu.
The following rather odd bug has showed up (from running
random-type-prop.typep.1):
(deftest misc.612
(funcall
(compile nil '(lambda (p1 p2)
((lambda (x y) (typep x (type-of y))) p1
(the (member "foo" #\- :b "bar") p2
#*1 :b)
nil)
CL-TEST>(do-t
I've been writing tests that check if the implementation
signals errors in such cases, even if not required by ANSI.
See ansi-tests/beyond-ansi/. This is still a work in progress.
Paul
Camm Maguire wrote:
Greetings! This is my understanding too, and general goal that the
compiler shou
Output of 'make' attached.
The version of ld is: GNU ld version 2.13.90.0.18 20030206
And gcc: gcc (GCC) 3.2.2 20030222 (Red Hat Linux 3.2.2-5)
Paul
(cd o && make ../h/new_decl.h)
make[1]: Entering directory `/home/dietz/gcl/o'
gcc -o grab_defs grab_defs.c
gcc -c -Wall -DVOL=volatile
Camm Maguire wrote:
Greetings! cmucl and clisp both seem to fail this too. Is string
supposed to be `(or (vector standard-char) (vector base-char) (vector
extended-char) (vector character))?
The three tests string-is-not-vector-of-character.[135]
have the note :nil-vectors-are-strings. This
Camm Maguire wrote:
Greetings! I need a more powerful subtypep for the compiler, and it
appears I have one basically implemented. I'd like to make use of the
second value to indicate whether the first type is in the complement
of the second type -- i.e. nil t means the types are disjoint (so th
The random tester is finding a problem in cvs head:
>(compile nil '(lambda () (let ((x (values 0))) 0)))
Error in WHEN [or a callee]: The GO tag #:G3614 is missing.
This is also test misc.603 in ansi-tests.
Paul
___
Gcl-devel mailing list
Camm Maguire wrote:
Greetings! Paul, are there any overlapping types in Common Lisp
neither of which is a subtype of the other?
FIXNUM and UNSIGNED-BYTE
(I assume you mean builtin types with type specifiers that
are symbols.)
Paul
___
G
Robert Dodier wrote:
Well, I'm not sure that anyone actually expects the
input string "NaN" to read the same as the floating point
object of which it is a representation.
One property that is important: if *PRINT-READABLY* is true,
then either NaN should print in a readable form (that is,
in a
Camm Maguire wrote:
Its mostly intended to scratch the itch I always feel having to
remember to add :test 'eq to the end of member for a simple symbol
lookup, etc. Hopefully others have the same itch. GCL even had its
own si::memq hack to the same effect. Without this of course one has
the eq
Camm Maguire wrote:
Thanks! These are very useful. Do you have any plans for same on
sequence functions, e.g. map, map-into, every, some, etc?
I do plan to. I also plan to fix the problem with destructive
operations.
Did this testing find any problems for you?
Paul
_
Camm Maguire wrote:
Thank you, and my apologies for the question. I'm trying to hurry
this summer, and am using your wonderful tet suite largely as a
replacement for me having to read and parse the spec in detail
myself. I should have known that you'd get it right, but I just had
the nagging qu
Camm Maguire wrote:
2) Paul, if *compile-verbose* defaults to t, must the output of compile-file
have
semi-colons?
From COMPILE-FILE:
"If verbose is true, compile-file prints a message in the form of a comment
(i.e., with a leading semicolon) to standard output indicating what
fi
Pascal Bourguignon wrote:
Camm Maguire writes:
Greetings! Is there a type specifier which restricts elements of a
list, e.g. '(list symbol)?
Not in Common Lisp.
But in Common Lisp, you can write such a type specifier:
[...]
The only difficulty is that SATISFIES expects a symbol and nothin
Camm Maguire wrote:
The compile-file tests try to fmakunbound the test function before it
is defined, and don't catch the error. Is this intended?
That shouldn't be an error. FMAKUNBOUND is described
as doing this:
"Removes the function or macro definition, if any,
of name in the globa
Camm Maguire wrote:
Greetings! Is there a type specifier which restricts elements of a
list, e.g. '(list symbol)?
No. DEFTYPE isn't required to be recursive, which would be the
obvious way to do it.
Paul
___
Gcl-devel mailing list
Gcl-d
The build on Version_2_7_0t1 segfaults for me:
[...]
Loading binary of GCL_PCL_CACHE...
Loading gcl_pcl_cache.o
Segmentation violation: c stack ok:signalling error
Error: Caught fatal error [memory may be damaged]: Segmentation violation.
Fast links are on: do (si::use-fast-links nil) for debuggi
Camm Maguire wrote:
Thank you so much for your report! As you know, ansi is still a work
in progress for gcl. It would be most helpful if you could double
check that the specific errors you see are covered in the existing
failures reported in Paul's ansi test suite. They are sure to get
fixed
Robert Boyer <[EMAIL PROTECTED]> writes:
According to one place on the web
(http://primes.utm.edu/primes/search.php?Number=100) the largest known
prime is:
(- (expt 2 25964951) 1)
GCL can compute this (in a rather long 7 seconds), but then causes a
segmentation fault trying to print it.
I wrote:
I think it would be. SBCL does it for both sizes. They cache
the SXHASH for symbols into the symbol rather than precomputing
it (and the hash doesn't depend on the package), but the idea
is similar.
Actually, SXHASH cannot depend on the package of a symbol,
since it must not change if th
Camm Maguire wrote:
Greetings, and thanks for the feedback!
"Paul F. Dietz" <[EMAIL PROTECTED]> writes:
Camm,
Judging by the space allocation numbers from TIME, objects
are also constrained to be an even number of words in Allegro CL
and in SBCL. I think this is a common idea.
Camm,
Judging by the space allocation numbers from TIME, objects
are also constrained to be an even number of words in Allegro CL
and in SBCL. I think this is a common idea.
If you have an extra field in some objects, consider uses for
it. For example, we might want to precompute a hash key
f
Camm Maguire wrote:
Paul, how does one instruct your random tester to zero in on a
particular function like expt?
This is properly the function of the random type prop tester.
In the ansi-tests directory:
(load "gclload1.lsp")
(compile-and-load "random-int-form.lsp")
(compile-and-load "random-type-
Camm Maguire wrote:
Thanks for the report. Should be fixed in CVS now, as was Paul's
remaining fmakunbound bug regarding setf function names. Paul,
perhaps you want a test in your suite for fmakunbound on undefined
function names?
I'll add that.
Paul
_
Peter Wood wrote:
(let ((foo 'bar) (bar 'foo)) (defun fubar () (list bar foo)))
FUBAR
(symbol-function 'fubar)
(LAMBDA-BLOCK-CLOSURE ((BAR FOO) (FOO BAR)) () () FUBAR ()
(LIST BAR FOO))
(compile 'fubar)
LAMBDA-BLOCK-CLOSURE fell through ECASE expression.
Wanted one of (LAMBDA LAMBDA-BLOCK).
63 matches
Mail list logo