Custom Query (1030 matches)
Results (583 - 585 of 1030)
| Ticket | Resolution | Summary | Owner | Reporter |
|---|---|---|---|---|
| #209 | fixed | No support for vectors with element-type FIXNUM in COMPILE-FILE/FASLOAD | ||
| Description |
Even though arrays/vectors with element-type FIXNUM have existed in OpenMCL for over a year, neither COMPILE-FILE nor the fasl loader have any support for dumping/loading them when they appear as constants. The big ETYPECASE in FASL-DUMP-DISPATCH misses them and falls into a clause which tries to dump them as GVECTORs; whether that "works" or not depends on how the bits in the vector are (mis)interpreted as tagged objects, as in the example below. (defconstant *matrix*
(let ((matrix (make-array '(80) :element-type 'fixnum)))
(dotimes (i 80 matrix)
(setf (aref matrix i) (random 1000)))))
(defun problem81 ()
(let ((j 50))
;;(aref *matrix* 50) ; no problem
(aref *matrix* j) ; bad
))
|
|||
| #267 | fixed | No runtime warnings on call to (eql) | ||
| Description |
? (defun foo () (eql)) ;Compiler warnings : ; In FOO: In the call to EQL with arguments (), ; 0 arguments were provided, but at least 2 are required ; by the current global definition of EQL ? (foo) T ? Minor, but it would be nice to get a runtime error, in case you miss the compile-time warnings for whatever reason. |
|||
| #271 | fixed | No error for unknown defclass class options | ||
| Description |
? (defclass erroneous-class.13 () (a b c) (#.(gensym))) #<STANDARD-CLASS ERRONEOUS-CLASS.13> ? (make-instance 'erroneous-class.13) #<ERRONEOUS-CLASS.13 #x300044399DDD> ? The spec says a program-error is required for unsupported options. This may have been augmented by the MOP, I don't know. But in any case, some indication of a problem should be given at some point. After a little poking about, it seems like the problem is this method: (defmethod initialize-instance :before ((class class) &key &allow-other-keys) (setf (%class.ctype class) (make-class-ctype class))) The &allow-other-keys in this method causes keyword arg checking for class objects to be disabled. I wonder it it would be worthwhile (or even allowed) to treat methods that have &allow-other-keys without &rest as not disabling initialization arg checking. |
|||
