Fix %in-bounds-constraint.
Delete xfloat-math.lisp-expr
Delete unused code and fixup comments
Adjust compile-time cross-float tests
Implement oracles for ATAN and COSH
Implement FROUND and friends portably
Implement SQRT portably on flonums
Implement +, -, * and / portably on flonums
Implement MIN/MAX on flonums portably
Implement comparisons on flonums portably
Support denormals in COMPUTE-FLONUM-VALUE
Implmement FLOAT-INFINITY-P and friends on flonum bits
Break the build on float cache mismatch
Implement oracle for needed calls to LOG on flonums
Implement SCALE-FLOAT portably on flonums
Implement EXPT portably on flonums
Implement CEILING, FLOOR, ROUND, TRUNCATE portably on flonums
Implement FLOAT-SIGN portably on flonums
More rigorous check of cross-float results in warm.lisp
Implement FLOAT portably on flonums
Implement ABS directly on FLONUMs
Implement COERCE portably
Define and use FLONUM-MINUS-ZERO-P
Implement RATIONALIZE in terms of (portable) RATIONAL
Implement RATIONAL portably on cross-floats
Compute values of parsed cross-floats
Allow implicit progn in WITH-MEMOIZED-MATH-OP
Make separate host SINGLE-FLOAT and DOUBLE-FLOAT structs
Return slightly more helpful values from CONTAINS-IRRATIONAL
Track a bug with a failing test rather than an intrigued comment
or-eq-to-jump-table: better exhaustive checks.
Split or-eq-to-jump-table.
or-eq-to-jump-table: don't use a jump-table for a single target.
x86-64: better (eql ratio x).
Avoid a cross-compile-time use of LOG
Move permgen layout allocator into alloc.lisp
Fix S-L-A-D crash if #+permgen
Fix integer rounders' interaction with minus zero
Convert COND + OR + EQ into jump-tables.
Handle generic functions in sb-introspect:find-function-callees
Don't transform mask-field with out of range size.
Find XREF for anonymous functions.
Skip a test.
x86-64: correctly allow %dpb VOPs.
x86-64: Decode imm32 operand as immobile layout if possible
Type-check (nth-value n (values ... &rest type)).
Type-check (values ... &rest type).
Shave a few instructions from RETURN-MULTIPLE
Delete stray character
arm64: add support for arenas.
Remove dynbind.h
Optionally preserve EFLAGS in WITH-REGISTERS-PRESERVED
Fix load-other-pointer-widetag + non-null-symbol-p.
More compact list accumulation.
arm64: add :targets
x86-64: save one byte in >-integer-fixnum.
Fix *handler-clusters* initialization in warm.lisp
Inline FIND if OR-EQ-TRANSFORM can handle it.
arm64: more compact widetag-of and layout-of
Warn about type proclaimations conflicting with values.
Check complex type assertions.
get-defined-fun: handle :declared-verify.
with-array-data: inform about array bounds.
arm64: encode as immediates (--mod64 x negative-constant)
x86-64: encode as immediates (--mod64 x negative-constant)
or-eq-transform: restore tagged fixnum handling.
win 2.4.3
Update for 2.4.3
2.4.3: will be tagged as "sbcl-2.4.3"
Adjust NEWS.
Draft NEWS for sbcl-2.4.3
Check for (boundp 'sb-c:*compilation*)
Remove an unused variable.
Hide sb-c:jump-table from third party code walkers.
Add a NEWS entry.
Non-T arrays are simple constants too.
case-to-jump-table: accept only simple constants.
case-to-jump-table: add load-time-value around constants.
Disable perfectly hashed packages except for "CL"
sb-introspect: Make function form-number work even on debug 0.
or-eq-transform: disable untagged tests when subtraction is needed.
sb-introspect: Guard against (DEBUG 0) functions.
Add WITHOUT-PACKAGE-LOCKS when interning debug-var symbols.
parse-integer transform: correctly accept all keywords.
Properly exit from make-perfect-hash-lambda.
Optimize parse-integer for known :radix 10 and 16.
Perfectly hash initial pprint-dispatch cons entries
or-eq-transform: operate on signed tagged fixnums.
Update xfloat-math
arm64: fix the DPB vops when size+posn is > 64.
or-eq-transform: correctly tag negative fixnum constants.
coalesce-debug-info: Don't iterate over the debug-info-fun-map.
Implement consistent minus zero handling for FROUND