[Sbcl-announce] sbcl-2.3.6 released

1 view
Skip to first unread message

Christophe Rhodes

unread,
Jun 28, 2023, 3:28:08 AM6/28/23
to sbcl-a...@lists.sourceforge.net
Dear all,

I have performed the release incantations for SBCL 2.3.6; the fruits of
this month's efforts are available in source form (and a x86-64/linux
binary) at SourceForge, and binaries for some other platforms may follow
shortly. Changes in this release are summarized below.

There is also some work going on in various longer-lived branches. I
myself have been working on improving external-format support, initially
motivated by adding support for newline ending conventions, but in the
process addressing a number of other issues. Development is not yet
complete but the branch at
https://github.com/csrhodes/sbcl/tree/external-format-line-endings
should be in a usable state, and any testing, experience reports or
comments about it would be welcome.

Christophe

changes in sbcl-2.3.6 relative to sbcl-2.3.5:
* platform support:
** restore building contribs on riscv; (lp#2002930)
** shorter constant-loading sequences on riscv;
** on OpenBSD, map the stack without executable permission (thanks to
Sébastien Marie)
** Restore OpenBSD/arm64 for OpenBSD 7.3 (lp#2024003, reported by Robert
Palm)
* bug fix: AREF on multidimensional arrays with the wrong number of indices
now signals an error. (lp#2022327, reported by EU)
* bug fix: the nature of NIL as both STRING and SEQUENCE is correctly handled
in the compiler's handling of string functions. (lp#2023118, reported by
Patrick Poitras)
* bug fix: the sb-bsd-sockets tests no longer fail on systems configured
without IPv6. (reported by Will Senn)
* bug fix: the compiler no longer transforms into incompletely-known functions
(lp#1824351)
* optimization: better (the word-sized-type (ASH word-sized word-sized))
when the result can overflow. For arm64 and x86-64.
* optimization: better unoptimized calls to CEILING, FLOOR.
* optimization: functions that involve coercing floats now cons less.
For arm64 and x86-64.
* optimization: array displacement to simple arrays is slightly faster.
(reported by Shubhamkar Ayare)

_______________________________________________
Sbcl-announce mailing list
Sbcl-a...@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-announce

Reply all
Reply to author
Forward
0 new messages