- Jan 23, 2003
-
-
pmai authored
Corrects the problems uncovered by Paul F. Dietz's ANSI test cases vector.type.13 and vector.type.15, where typep would return two values for certain array types, e.g. (typep "ABC" '(vector t)) => NIL, T instead of only NIL
-
- Jan 21, 2003
-
-
pw authored
argument to dlsym (FreeBSD port). This may be source of some segfault reports while trying to use foreign functions.
-
toy authored
From a recent thread on comp.lang.lisp, it seems that CMUCL is non conforming in the way that it handles the coalescing of constants in the functions EVAL and COMPILE: Literal objects appearing in code processed by the compile function are neither copied nor coalesced. The code resulting from the execution of compile references objects that are eql to the corresponding objects in the source code.
-
- Jan 16, 2003
-
-
cracauer authored
Backtrace print tuning: Replace the hardcoded default for the keyword parameter "verbosity" when printing backtrace frames with a parameter debug:*default-print-frame-call-verbosity* Default is 1, which is the old value. Setting it >= 2 gets you the source location printed for every frame print in a backtrace (when the source is available).
-
- Jan 14, 2003
-
-
cracauer authored
Typo fixes by Ed Wang (thanks, Ed!).
-
- Jan 10, 2003
-
-
toy authored
denominator are different lengths. Should be log, not log2.
-
- Jan 08, 2003
- Jan 07, 2003
- Jan 06, 2003
- Jan 03, 2003
-
-
pmai authored
Commited patch by Gerd Moellmann that fixes a bug in the accessing of class-slots for redefined classes in certain cases, due to caches not being flushed because of reused wrappers.
-
toy authored
-
pmai authored
As noted by Gerd Moellmann, PCL needs a method on finalize-inheritance specialized for forward-referenced-class, which signals a relevant error, since forward-referenced-classes can't be finalized.
-
pmai authored
Fix to ensure that finalize-inheritance (the bug's name is a misnomer) is called on class-finalization, from Gerd Moellmann via SBCL.
-
pmai authored
of this file, which was otherwise unchanged.
-
pmai authored
Updated the fix for the broken handling of the :arguments option to define-method-combination to Gerd Moellmann's latest patch. Also includes various indentation fixes and a smallish documentation/comment fix.
-
toy authored
inheritance bug (SBCL bug 127). Slightly modified to make building a little easier. Todo: After the next release (18e) is done, go through and remove the old bootstrap-conc-name reader conditionals. We want the #-bootstrap-conc-name versions.
-
toy authored
-
- Jan 02, 2003
-
-
pmai authored
As per ANSI, define-method-combination should return the name of the method combination.
-
- Dec 31, 2002
-
-
toy authored
single-float or double-float. Thus (log (expt 2 32768) 2) will return 32768.0 instead of causing an error.
-
- Dec 29, 2002
-
-
pmai authored
The :initial-contents argument to adjust-array isn't restricted to be a list, as the defknown entry erroneously claims. This fix requires an L2 rebuild (or loading of the changed defknown prior to rebuilding), in order for the information to propagate properly.
-
- Dec 22, 2002
-
-
pmai authored
Added needed second half of patch, from Gerd Moellmann via SBCL, to cure newly introduced problem with inheritance of class slots. Also committed modified version (again via SBCL) of Gerd's COMPUTE-SLOTS revamp for more modular computation of slot locations.
-
- Dec 20, 2002
- Dec 18, 2002
-
-
pmai authored
Merged patches and ideas by Gerd Moellmann and Christophe Rhodes to improve error reporting on missing applicable primary methods for standard method- combination to report the arguments that were passed to the GF in question.
-
pmai authored
Committed a slightly adjusted patch from Gerd Moellmann, which makes the GF slot-definition-allocation return :class for class slots, as specified by AMOP, instead of the class object, as it did previously.
-
pmai authored
This commits a patch by Gerd Moellmann which elides the elaborate cacheing of GF caches, since modern GCs are usually better at handling this. Performance on x86 with generational GC seems unharmed by this, other ports using the non-generational GC will have to investigate. Should this turn up performance problems, the patch might get reverted.
-
pmai authored
This commit fixes the problem by ripping out all call-next-method and next-method-p optimizations done by PCL through code walking, since the CMUCL compiler is smart enough to do all those optimizations (like removing unused local functions, and/or inlining the code if it isn't closed over) better by itself. This leads to the elimination of quite a bit of hairy code from PCL. Since there currently exists no way to locally override a user-supplied ignore declaration on a variable for macro-generated code, we need to remove ignore declarations for method arguments, in order not to generate superfluous warnings. This is suboptimal, and hence should be corrected through compiler enhancements at some time.
-
- Dec 13, 2002
- Dec 12, 2002
- Dec 09, 2002
-
-
toy authored
(Noted by Paul Dietz.)
-
- Dec 07, 2002
-
-
toy authored
* fix DESCRIBE for arrays of zero rank. Bug and fix reported for SBCL on the sbcl-help list by Lutz Euler. (Slightly modified)
-
toy authored
- fixes to the internal consistency of IR2, ported from the corresponding changes to SBCL by Alexey Dejneka - modification of the internal IR1 consistency checking code to accept objects of type HEAP-ALIEN-INFO in the *FREE-VARIABLES* list. With this change, it is possible to rebuild with c::*check-consistency* enabled, with only the "Unseen function ~S in ~S." check disabled.
-
toy authored
CLtS.
-
toy authored
by explicitly checking for this case. o In member :complex-subtypep-arg2, handle the case of the second being of type NULL. This makes (subtypep 'atom 'list) work. All the type stuff needs more work. May use some ideas from SBCL?
-
- Dec 04, 2002
-
-
toy authored
is the type T. (This still needs more work.) o Put back the original hairy :complex-subtypep-arg1 case to handle AND specifiers. This still needs more work too.
-