sbcl-1.5.6 released
Christophe Rhodes <[email protected]> Mon, 26 Aug 2019 10:59:07 +0100
| Newsgroups | gmane.lisp.steel-bank.announce |
|---|---|
| Message-ID | <87v9ukz1qs.fsf@shin> |
Dear all,
I have released and uploaded sbcl-1.5.6 to all the usual places. The
major change in this release is that the half-finished 64-bit PowerPC
port done by Brian Bokser last year has been three-quarter or
nine-tenths finished: I have seen reports that the system is able to run
SLIME on version 1 of the powerpc64 ABI, and that there is also support
for version 2 of the ABI. Bug reports (and success reports) for this
and other features are welcome as always. A summary of the changes in
this release is appended below.
Best wishes,
Christophe
changes in sbcl-1.5.6 relative to sbcl-1.5.5:
* platform support:
- experimental support for 64-bit PowerPC running Linux (extending Brian
Bokser's work from 2018), on both the v1 and v2 ABIs
* new feature: SB-EXT:SEARCH-ROOTS discovers paths from live objects to the
roots keeping them alive.
* enhancement: string output streams created with :ELEMENT-TYPE 'BASE-CHAR
use internal buffers of BASE-STRING instead of UCS-4 strings restricted
to the ASCII range, yielding a theoretical 4:1 space reduction.
* optimization: improved make-array type derivation for multi-dimensional
arrays. (lp#1838442)
* bug fix: compliant redefinition of classes whose previous definition
caused argument mismatch errors does not generate errors any more.
(lp#1840595, reported by 3b on #sbcl)
* bug fixes for issues caught by the random tester:
- never derive the type of TRUNCATE on arbitrary numbers as the empty
type. (lp#1838267)
- provide out-of-line definitions for internal machinery related to
FLOAT-SIGN. (lp#1838337)
- include COMPLEX in the derived type of SIGNUM when appropriate.
(lp#1838333)
- more correct internal type testing for function types. (lp#1838808,
lp#1838888, lp#1838986)
- don't assume that all objects of type (NOT SIMPLE-ARRAY) have an array
header. (lp#1838827)
- recognize that PHASE on non-positive numbers can return 0 as well as PI.
(lp#1838892)