- Nov 29, 2001
-
-
pmai authored
Linux in-line with all other non-MACH platforms, and with the Linux Filesystem Standard.
-
- Nov 26, 2001
-
-
toy authored
includes and libraries on Solaris. (Noted by Eric Marsden.)
-
- Nov 22, 2001
-
-
pw authored
-
- Nov 21, 2001
-
-
pmai authored
:displaced-to option, as required by the tightened error-checking done in MAKE-ARRAY.
-
pmai authored
didn't check whether the specified element-type was a subtype of the array-element-type of the displaced-to array. o One case in ADJUST-ARRAY didn't check for an array-header before calling %array-displaced-p on an array, so that sometimes an array element was accessed and checked instead. Fixed this, and prevented ADJUST-ARRAY from shrinking non-adjustable arrays in-place, since this can violate naive user expectations for little gain.
-
- Nov 20, 2001
-
-
pw authored
the rest of the world. Noticed by Ray.
-
- Nov 05, 2001
-
-
toy authored
31 bits. Modified by RLT to get rid of an extra branch instruction. o Update right shifts for (signed-byte 32) to get rid of an extra branch instruction too.
-
- Nov 03, 2001
-
-
pw authored
-
- Oct 31, 2001
- Oct 30, 2001
-
-
pmai authored
-
pmai authored
to be loaded in-full, which is the only useful behaviour when trying to link with static libraries. Currently this change only corrects the situation for Linux and Solaris, since I don't know the relevant flags for other platforms, and/or don't have access to machines for testing.
-
- Oct 28, 2001
- Oct 27, 2001
-
-
pmai authored
to be of type program-error, as required by section 3.5.1.6 of the standard. From a patch by Eric Marsden.
-
- Oct 16, 2001
-
-
toy authored
o Add :external-format option to LOAD, as specified by ANSI. Doesn't do anything currently.
-
- Oct 15, 2001
-
-
pw authored
symbol-macros by changing CADR to CDR. Probably a typo? Now (declare (ignore xxx)) in WITH-SLOTS (xxx ...) works.
-
- Oct 12, 2001
-
-
toy authored
cross-compilation not to loop forever. From Douglas.
-
- Oct 05, 2001
-
-
toy authored
o Delete an unneeded instruction.
-
- Oct 04, 2001
-
-
toy authored
Loop doesn't handle loops like (loop with (a b) = '(1 2) and (c d) = '(3 4) return (list a b c d)) correctly because it fails to destructure the variables correctly.
-
- Oct 03, 2001
-
-
toy authored
o Remove the unused check-list and check-instance that generated invalid trap instructions on Sparc V9. o Use fixnum-tag-mask instead of #x3.
-
toy authored
-
toy authored
o Clean up the code a little, and delete the unused stuff about Sparc V9 not liking the trap instructions (that we don't generate anymore).
-
toy authored
-
toy authored
usage.
-
- Sep 28, 2001
-
-
toy authored
common situation where we don't have to call rational to get the correct result. (About 15 times worse on the frpoly/float benchmark if we don't do this.)
-
- Sep 27, 2001
- Sep 25, 2001
- Sep 24, 2001
-
-
toy authored
need. gcc 3.0 does this even with -O0. (I think this is a bug in gcc.) Try to arrange things so that gcc can't possibly delete them by moving functions to reference_random_symbols_table.
-
toy authored
register that is used if the offset is too big to fit into the 13-bit offset field of the LD/ST instruction.
-
toy authored
causing unexpected NaN, infinities, and loss of precision. o Get rid of the temp TN by recycling a register.
-
toy authored
too large to fit into the 13-bit offset field of a Sparc instruction. Handle that by loading the value into a register first. (I think, Eric Marsden has some test code that tickles this bug.)
-
toy authored
-
toy authored
operation for the denominator). o Add corresponding deftransform for real/complex.
-
- Sep 23, 2001
-
-
pmai authored
cmucl-help: The optional order argument to method combinations defined by the short form of define-method-combination was being quietly ignored, resulting in the default :most-specific-first behaviour, even if :most-specific-last was specified. This fix makes it respect the specified order.
-
- Sep 22, 2001
-
-
pw authored
represses compiler warnings about undefined class when compiling defmethods in the same file as the defclass forms. It does this by calling (lisp:make-class :name class-name) which creates a minimal entry such that lisp:find-class will return non-nil. The class entry is filled out at load time. A consequence of this is that until load time, any attempt to treat that minimal class as valid will fail, probably because lisp:class-layout => NIL. This doesn't seem a problem in current practice.
-
- Sep 21, 2001
-
-
pw authored
problems are fixed now. This revision fixes an omission from a previous rev that added support for :print-object and :print-function defstruct options. The default method created was calling DEFAULT-STRUCTURE-PRINT with only two arguments. It seems to work correctly now.
-