sbcl-2.0.10 released
Christophe Rhodes <[email protected]> Tue, 27 Oct 2020 16:46:21 +0000
| Newsgroups | gmane.lisp.steel-bank.announce |
|---|---|
| Message-ID | <87a6w7vcpu.fsf@shin> |
Dear all,
I have released SBCL 2.0.10, and pushed sources and a binary to
SourceForge; a summary of the changes is appended below.
There is a known bug in spilling of AVX registers to stack in this
release, which is fixed in the development branch for the next release.
This is only likely to affect you if you (or systems that you depend on)
explicitly perform vectorized operations on the x86-64 platform using
the SB-EXT:SIMD-PACK-256 type; normal user code is unaffected by this
bug.
Best wishes,
Christophe
changes in sbcl-2.0.10 relative to sbcl-2.0.9:
* minor incompatible change: the funarg given to SB-SPROF:MAP-TRACES
does not receive a wallclock time with each trace.
* minor incompatible change: INTERNAL-TIME-UNITS-PER-SECOND has been
increased to 10^6 on 64-bit architectures.
* minor incompatible change: SIGPIPE is ignored by default again. (lp#1897624)
* minor incompatible change: the system code compiled under the
:LINKAGE-TABLE feature is now unconditionally compiled in, and the
corresponding entry in *FEATURES* has been removed.
* enhancement: style-warnings are issued for variables which have an
assignment but no "for-value reference" (per CLHS glossary entry)
* bug fix: SB-CLTL2:MACROEXPAND-ALL did not expand MULTIPLE-VALUE-BIND
and MULTIPLE-VALUE-SETQ
* bug fix: CPUID-based feature detection had an index/mask confusion
(lp#1899239)
* bug fix: fix a deadlock on Windows (lp#1896802)
* bug fix: eliminate type errors when wall clocks go back (lp#1028026,
lp#1032111)
* bug fix: fix EOF handling in read-char-no-hang on concatenenated streams
(lp#690408, reported by Willem Broekema)
* bug fix: fix MAP-INTO on extended sequences (lp#1855375, thanks to James
Kalenius)
* bug fix: SB-GMP can now raise -1, 0 and 1 to the power of a bignum.
(thanks to Aaron Chen)
* bug fixes in tests:
** add a C function declaration (lp#1897627, thanks to Bob Felts)
** parse vmmap output more liberally (lp#1897722, reported by Bob Felts)