プログラミング
Steel Bank Common Lisp version 2.6.7
Steel Bank Common Lisp version 2.6.7 (sbcl.org)
要約
Steel Bank Common Lisp (SBCL) のバージョン2.6.7がリリースされました。このバージョンでは、SBCLマニュアルをdocstringに統合した新しいcontribモジュールSB-MANUALの追加、ARM64でのSIMD命令サポートの拡充、コンパイラの最適化、および多数のバグ修正が含まれています。また、以前のバージョン2.6.6および2.6.5の変更点もリストアップされています。
全文翻訳
Steel Bank Common Lisp
- About
- News
- Download
- Getting Started
- Documentation
- History and Copyright
- Porting
- Maintainer Public Keys
- Project Page
- Bug Database
- Links
[画像: SourceForge Logo] [画像: (get 'sbcl)]
All News
SBCL の新しいバージョンは通常、毎月末にリリースされます。現在のバージョンを確認するには、Sourceforge File List を参照してください。すべての SBCL リリースの新機能は以下にリストされています。
- 2.6.9
- 2.6.8
- 2.6.7
- 2.6.6
- 2.6.5
- 2.6.4
- 2.6.3
- 2.6.2
- 2.6.1
- 2.6.0
- 2.5.11
- 2.5.10
- 2.5.9
- 2.5.8
- 2.5.7
- 2.5.6
- 2.5.5
- 2.5.4
- 2.5.3
- 2.5.2
- 2.5.1
- 2.5.0
- 2.4.11
- 2.4.10
- 2.4.9
- 2.4.8
- 2.4.7
- 2.4.6
- 2.4.5
- 2.4.4
- 2.4.3
- 2.4.2
- 2.4.1
- 2.4.0
- 2.3.11
- 2.3.10
- 2.3.9
- 2.3.8
- 2.3.7
- 2.3.6
- 2.3.5
- 2.3.4
- 2.3.3
- 2.3.2
- 2.3.1
- 2.3.0
- 2.2.11
- 2.2.10
- 2.2.9
- 2.2.8
- 2.2.7
- 2.2.6
- 2.2.5
- 2.2.4
- 2.2.3
- 2.2.2
- 2.2.1
- 2.2.0
- 2.1.11
- 2.1.10
- 2.1.9
- 2.1.8
- 2.1.7
- 2.1.6
- 2.1.5
- 2.1.4
- 2.1.3
- 2.1.2
- 2.1.1
- 2.1.0
- 2.0.11
- 2.0.10
- 2.0.9
- 2.0.8
- 2.0.7
- 2.0.6
- 2.0.5
- 2.0.4
- 2.0.3
- 2.0.2
- 2.0.1
- 2.0.0
- 1.5.9
- 1.5.8
- 1.5.7
- 1.5.6
- 1.5.5
- 1.5.4
- 1.5.3
- 1.5.2
- 1.5.1
- 1.5.0
- 1.4.16
- 1.4.15
- 1.4.14
- 1.4.13
- 1.4.12
- 1.4.11
- 1.4.10
- 1.4.9
- 1.4.8
- 1.4.7
- 1.4.6
- 1.4.5
- 1.4.4
- 1.4.3
- 1.4.2
- 1.4.1
- 1.4.0
- 1.3.21
- 1.3.20
- 1.3.19
- 1.3.18
- 1.3.17
- 1.3.16
- 1.3.15
- 1.3.14
- 1.3.13
- 1.3.12
- 1.3.11
- 1.3.10
- 1.3.9
- 1.3.8
- 1.3.7
- 1.3.6
- 1.3.5
- 1.3.4
- 1.3.3
- 1.3.2
- 1.3.1
- 1.3.0
- 1.2.16
- 1.2.15
- 1.2.14
- 1.2.13
- 1.2.12
- 1.2.11
- 1.2.10
- 1.2.9
- 1.2.8
- 1.2.7
- 1.2.6
- 1.2.5
- 1.2.4
- 1.2.3
- 1.2.2
- 1.2.1
- 1.2.0
- 1.1.18
- 1.1.17
- 1.1.16
- 1.1.15
- 1.1.14
- 1.1.13
- 1.1.12
- 1.1.11
- 1.1.10
- 1.1.9
- 1.1.8
- 1.1.7
- 1.1.6
- 1.1.5
- 1.1.4
- 1.1.3
- 1.1.2
- 1.1.1
- 1.1.0
- 1.0.58
- 1.0.57
- 1.0.56
- 1.0.55
- 1.0.54
- 1.0.53
- 1.0.52
- 1.0.51
- 1.0.50
- 1.0.49
- 1.0.48
- 1.0.47
- 1.0.46
- 1.0.45
- 1.0.44
- 1.0.43
- 1.0.42
- 1.0.41
- 1.0.40
- 1.0.39
- 1.0.38
- 1.0.37
- 1.0.36
- 1.0.35
- 1.0.34
- 1.0.33
- 1.0.32
- 1.0.31
- 1.0.30
- 1.0.29
- 1.0.28
- 1.0.27
- 1.0.26
- 1.0.25
- 1.0.24
- 1.0.23
- 1.0.22
- 1.0.21
- 1.0.20
- 1.0.19
- 1.0.18
- 1.0.17
- 1.0.16
- 1.0.15
- 1.0.14
- 1.0.13
- 1.0.12
- 1.0.11
- 1.0.10
- 1.0.9
- 1.0.8
- 1.0.7
- 1.0.6
- 1.0.5
- 1.0.4
- 1.0.3
- 1.0.2
- 1.0.1
- 1.0
- 0.9.18
- 0.9.17
- 0.9.16
- 0.9.15
- 0.9.14
- 0.9.13
- 0.9.12
- 0.9.11
- 0.9.10
- 0.9.9
- 0.9.8
- 0.9.7
- 0.9.6
- 0.9.5
- 0.9.4
- 0.9.3
- 0.9.2
- 0.9.1
- 0.9.0
- 0.8.21
- 0.8.20
- 0.8.19
- 0.8.18
- 0.8.17
- 0.8.16
- 0.8.15
- 0.8.14
- 0.8.13
- 0.8.12
- 0.8.11
- 0.8.10
- 0.8.9
- 0.8.8
- 0.8.7
- 0.8.6
- 0.8.5
- 0.8.4
- 0.8.3
- 0.8.2
- 0.8.1
- 0.8.0
- 0.8alpha.0
- 0.7.14
- 0.7.13
- 0.7.12
- 0.7.11
- 0.7.10
- 0.7.9
- 0.7.8
- 0.7.7
- 0.7.6
- 0.7.5
- 0.7.4
- 0.7.3
- 0.7.2
- 0.7.1
- 0.7.0
- 0.6.13
- 0.6.12
- 0.6.11
- 0.6.10
- 0.6.9
- 0.6.8
- 0.6.7
- 0.6.6
- 0.6.5
- 0.6.4
- 0.6.3
- 0.6.2
- 0.6.1
- 0.6.0
New in version 2.6.9, 2026-09-26
- minor incompatible change: structure slots defined as :READ-ONLY can be written using SLOT-VALUE with the constant slot name, matching the existing behaviour if the slot name is provided as a variable.
- enhancement: addition of u8.16-shuffle to SSE3 SB-SIMD support. (thanks to Qiantan Hong)
- enhancement: addition of some AVX512 support to SB-SIMD. (thanks to Aaron Estrada)
- bug fix: address a number of issues in the SB-SIMD contributed module. (thanks to Sylvia Harrington)
- bug fix: don't improperly inherit dynamic extent nature of transformations of lists constructed with dynamic extent.
- bug fix: don't call SB-GMP exponential function with a non-rational base. (thanks to Ilia Gradina)
- bug fix: compiler internal error from stack manipulation in inline- and dynamic-extent-heavy code. (#2156347, reported by Vasily Postnicov)
- bug fix: the SB-MANUAL contributed module should build more reliably on Windows. (#2165787)
- bug fix: be more conservative when inferring types of components of composite data structures lexically following a declaration in the presence of possible side effects. (#2165835)
- bug fix: compiler miscompilation of type checks of unused optional arguments. (#2165974, reported by Vasily Postnicov)
- bug fix: test consistency checks of system variables now correctly ignore SB-IMPL::READ-LINE-BUFFERS, affecting particularly large heaps. (#2167121, reported by 3b)
- bug fix: miscompilation through too-optimistic reuse of storage locations. (#2167208)
- bug fix: compiler error from off-by-one access to internal data structures. (#2167568)
- bug fix: compiler error from compilation of nested functions with non-local exits. (#2167725)
- bug fix: compiler error after incautious value coalescing. (#2167800)
- optimization: microoptimizations to the return sequence to generic arithmetic assembly routines on x86-64.
- optimization: checks for whether floats are integral are more compact.
- optimization: on arm64, integer-length is more efficient on machine-word-sized integers.
- optimization: certain basic operations on numeric RATIO types have a fast path.
- optimization: the GMP-INTEXP used in the SB-GMP implementation of EXPT is faster.
- optimization: self- and mutually-referential functions can be automatically stack-allocated when the system can prove that no function reference can escape.
- optimization: the compiler can propagate equality constraints over local calls.
- optimization: the compiler can make use of register pairs for unboxed signed 128-bit simple arithmetic on ARM64 and x86-64.
- optimization: constraint propagation is somewhat faster.
- optimization: provide the compiler with smarts for type inference for SB-SYS:WITH-INTERRUPTS. (#2167153, reported by Vasily Postnicov)
New in version 2.6.8, 2026-08-28
- minor incompatible change: FILE-POSITION on composite streams returns a more accurate result for CONCATENATED-STREAM, and NULL on TWO-WAY-STREAM and ECHO-STREAM.
- platform support:
- improvements to code generation for the ARM64 backend.
- improved support for AVX512 instruction encoding and decoding. (thanks to Arthur Miller)
- various enhancements to the SB-SIMD contributed module. (thanks to Sylvia Harrington and Arthur Miller)
- bug fix: FILE-POSITION on a stream after reading input into a list or generic sequence returns the correct result.
- bug fix: allow the manual to build on more platforms. (#2162021)
- bug fix: infinite loop in the compiler while attempting to remove data moves. (#2162597)
- bug fix: debugger failed to handle (COMPLEX DOUBLE-FLOAT) entities on the stack. (reported by Clint Seinen)
- bug fix: error reporting from STRING-TO-OCTETS reported using the wrong string indices. (#2164924)
- optimization: the compiler performs assignment conversion in nearly all possible cases now, along the lines of Fluet and Weeks' "Contification using Dominators" of 2001., which transforms state-machine-like code written in terms of local functions into the equivalent tagbody/go control structures.
- optimization: type inference of parameters is now possible for local functions which recursively call themselves. (#486416, reported by Tobias Rittweiler)
- optimization: special-case DPB with newbyte of 0 or -1.
- optimization: more improvements to UTF8 encoding and decoding.
- optimization: multiple structure TYPEP tests in sequence can elide redundant memory loads of layouts.
- optimization: transform ZEROP of (LDB ...) to an equivalent zero test with a mask. (#2162824, thanks to Vasily Postnicov)
New in version 2.6.7, 2026-07-28
- new contrib module: SB-MANUAL contains the SBCL manual in docstrings of section definitions, which tie together the docstrings of normal Lisp definitions. The manual can thus be explored interactively in the usual way (e.g. with Slime's M-.), and it is browsable with the MGL-PAX library (out of tree). Also, https://fixnum.com/ (similarly unrelated to the SBCL project) provides alternative renderings of the SBCL manual as heavily linked PDF and HTML documents as well as Markdown and plain text.
- new feature: DOCUMENTATION supports DOC-TYPE DECLARATION.
- platform support:
- the SB-SIMD contrib now supports ARM64. (Thanks to Sylvia Harrington)
- AVX512 instructions are now supported on X86-64. (Thanks to Robert Smith and Arthur Miller)
- additional support for SIMD instructions on ARM64 and X86-64. (Thanks to Arthur Miller)
- fix miscompilation of SAP-REF-N on ARM64. (Thanks to Hayley Patton)
- implement INTEGER-LENGTH on primitive types without a loop on MIPS and LoongArch.
- bug fix: READING with READ-SUPPRESS T no longer emits warnings like "<internal-feature> no longer present on FEATURES".
- bug fix: compiler type-error when compiling calls to CONCATENATE with conditional known non-sequence arguments. (#2160747)
- bug fix: (EQL <complex>) types were not being treated as numeric by the type system. (#2160429)
- bug fix: improve the handling of quiet (non-signalling) NaN inputs to LOG. (#2160268, reported by Woodrow Kiang)
- bug fix: miscompilation of MULTIPLE-VALUE-CALL. (#2160207, reported by Vasily Postnicov)
- optimization: passing constant complex numbers to local functions can be done without consing.
- optimization: where available, use enhanced SIMD routines for UTF-8 conversions.
- optimization: compiler transforms of COUNT are applicable with a wider variety of keyword arguments.
- optimization: remove at least one redundant instruction from SB-ALIEN:DEREF.
- optimization: the sparse set implementation in the compiler has been tuned to improve performance on real-world workloads.
- documentation: many typos and typesetting issues were fixed.
- documentation: SB-MANUAL:@FOREIGN-FUNCTION-INTERFACE now correctly states that arrays are row-major (not column-major). (#2158033, thanks to Scott L. Burson)
- documentation: internally, docstrings now conform to a subset of Markdown, but DOCUMENTATION (and thus DESCRIBE) strips some of this markup. The official manual is still generated from Texinfo, but the Texinfo files are generated from SB-MANUAL.
- documentation: the manual now has a separate index for declarations.
New in version 2.6.6, 2026-06-28
- minor incompatible change: FDEFINITION now returns the outermost wrapper (added e.g. by TRACE, PROFILE) like SYMBOL-FUNCTION. (#799533)
- minor incompatible change: in unsafe code, C strings with :EXTERNAL-FORMAT :ASCII are copied directly as byte-sized quantities without checking whether the top bit of the byte is set.
- platform support:
- fix the build on big-endian 64-bit PowerPC with ELFv2. (thanks to Piotr Kubaj)
- move the static space address for macOS 27 on ARM64. (#2156072, reported by Gary Palter)
- optimizations to SB-THREAD:BARRIER for ARM64. (thanks to Sahil Kang)
- fix a compiler crash in MULTIPLE-VALUE-LIST in argument forms on ARM64. (#2155788, reported by Gary Palter)
- bug fix: TRACE no longer fails when trying to print a return value that cannot be printed readably and PRINT-READABLY is true.
- optimization: the compiler is more precise in its type derivation of COERCE given constraints on its inputs.
- optimization: the compiler is better able to derive the return types of AREF and ELT.
- optimization: faster encoding and decoding of UTF-8 C strings.
- optimization: (length (intersection a b)) doesn't cons an intermediate list.
- documentation: the manual now includes a section for SB-INTROSPECT, which has also seen improvement in its documentation strings and comments.
- documentation: fixed many typesetting problems and typos in the user manual.
New in version 2.6.5, 2026-05-29
- minor incompatible change: the condition signalled when an accessed slot is missing from an object is no longer a TYPE-ERROR.
- minor incompatible change: the condition signalled when accessing an uninitialized structure slot is no longer a TYPE-ERROR.
- minor incompatible change: the implementations of standardized functions treating lists as sets, such as INTERSECTION and UNION, take more advantage of the freedom to return the elements of the result in any order.
- platform support:
- add low-level support for floating point state manipulation on PPC64/FreeBSD. (thanks to Piotr Kubaj)
- improve the software emulation of displaced instructions on ARM64.
- restore building the system using the musl C library. (#2153432, reported by Tom Gillespie)
- fix some SB-SIMD shifting instructions on AVX2. (#2152791, reported by Willem Broekema)
- enhancement: definition sources for alien callbacks are now findable by name in SB-INTROSPECT.
- bug fix: the SYMBOL restart for finding packages now actually performs a non-local exit. (#2153092, reported by Zach Shaftel)
- bug fix: TYPEP on large disjoint numeric union types compiles faster using fewer resources. (#2151818, reported by James McDonald)
- bug fix: strings of arbitrary size with fill-pointer set to 1 are character designators. (reported by _death)
- bug fix: the KEEP-OLD restart established by ADD-PACKAGE-LOCAL-NICKNAME keeps the old nickname instead of going ahead with the change (and the restart report function no longer returns from ADD-PACKAGE-LOCAL-NICKNAME).
- bug fix: when EXPORT results in a conflict between symbols exported by different used packages, the TAKE-NEW restart now shadowing-imports the new symbol instead of doing nothing and leaving the package in an inconsistent state.
- bug fix: the SB-EVAL interpreter checks program syntax more thoroughly.
- bug fix: compiler cross-reference data is decoded correctly for a functional with more than one entry for a given name.
- bug fix: TYPE-ERRORs signalled by SBCL are more likely to have a DATUM that is not of the condition's EXPECTED-TYPE.
- bug fix: the code walker no longer uses the stack to walk PROGN bodies.
- optimization: in various situations, INTERSECTION and UNION will use hash-tables to perform the operation more quickly.
New in version 2.6.4, 2026-04-29
- minor incompatible change: when DEFSETF is called on a name that was previously used as a (presumed) call to a function, it issues a single STYLE-WARNING (like DEFMACRO).
- minor incompatible change: SB-EXT:PROCESS-KILL no longer supports the :PTY-PROCESS-GROUP option (which was never correctly implemented).
- minor incompatible change: the :INITIAL-OFFSET argument for typed DEFSTRUCT, if given, no longer accepts NIL.
- platform support:
- more likely to support 32-bit linux's struct timeval with 64-bit time_t.
- the runtime's main function is now defined as a weak symbol for platform/compiler combinations that support it.
- on Windows, individual empty arguments for RUN-PROGRAM are escaped.
- add input/output speed fields for our definition of the termios structure, addressing a crash in sb-posix:tcsetattr on OpenBSD. (#2150499, thanks to Robert Palm)
- bug fix: address infinite loops in the compiler. (#2144911, #2148056)
- bug fix: if an FTYPE has been proclaimed for a function, don't mix NULL with explicitly-typed keyword arguments. (#2147050, reported by Vasily Postnicov)
- bug fix: compiler error from treatment of unused results. (#2147383)
- bug fix: compiler error from invalid dimension arguments to MAKE-ARRAY. (#2147384)
- bug fix: compiler error arising from continuing to optimize deleted nodes. (#2147385)
- bug fix: make sure linkage-table alien entries have base-string names. (#2147646, thanks to Seokjun Lee)
- bug fix: make sure CHECK-TYPE's expansion does not include internal non-externalizable objects. (#2148777, reported by Willem Broekema)
- bug fix: alien calls involving passing structs by value are less likely to read or write wrong areas of memory. (thanks to Jesse Bouwman)
- bug fix: lowering EQUALP to EQUAL handles constant dotted lists correctly.
- bug fix: a number of standard functions perform more explicit type checks on their arguments.
- bug fix: only return the primary value from (LIST*/APPEND/NCONC <values>).
- bug fix: correct treatment of escaped closing brackets in pathname patterns.
- bug fix: escape dots in pathnames more correctly.
- bug fix: the hash set for function names will incur collisions less frequently. (reported by Andrew Wolven)
- bug fix: the system is now capable of expressing select() on the whole range of possible file descriptors.
- optimization: compiler optimizations for (REPLACE vector list) now apply given :START1 and/or :END1 keyword arguments.
- optimization: CONCATENATE is faster for concatenating list arguments to a VECTOR.
- optimization: ROUND for integers is more compact.
- optimization: on 64-bit x86, implement TRUNCATE using the Lemire, Kaser, Kurz transform.
New in version 2.6.3, 2026-03-29
- minor incompatible change: (MAKE-ARRAY X :ELEMENT-TYPE 'UNDEFINED) now signals an error, consistent with (UPGRADED-ARRAY-ELEMENT-TYPE 'UNDEFINED).
- platform support:
- fix disassembler on ppc for the MFLR and ISEL instructions
- the Lisp Return Address object (as part of the Lisp calling convention) is no longer needed or supported on PPC, SPARC, MIPS or ARM. (This also frees up a widetag slot previously held by return-pc-widetag)
- remove sensitivity to SBCL init files when building embedcore-sbcl. (thanks to Robert Brown)
- add support for the ADCX and ADOX instructions on x86-64. (thanks to Robert Smith)
- on PPC64, indicate the number of return values through flags, making function calls four times faster.
- fix FFI involving int128 arguments on x86-64. (thanks to Andreas Franke)
- fix build on OpenIndiana/x86-64. (thanks to Andreas Wacknitz)
- fix build on Haiku/x86-64.
- bug fix: improved stability of (particularly) the mark-region garbage collector. (#2142711)
- bug fix: compiler type error in complicated expression involving BOOLE and conditionals. (#2142949)
- bug fix: dotted lists involving symbols whose names begins with "DEF" are not definitions. (#2143114, reported by Glenn Thompson)
- bug fix: STABLE-SORT miscompiled on declared union types involving both LIST and VECTOR. (#2143163, reported by akater, thanks to Vasily Postnicov)
- bug fix: more consistent results between compiler and interpreter in numerical computations involving negative zeros. (#2143383)
- bug fix: USE-PACKAGE once again signals the correct error if an attempt is made to use the KEYWORD package.
- bug fix: EQUALP on hash tables is no longer sensitive to irrelevant aspects of the hash table.
- bug fix: SB-INTROSPECT:DEFTYPE-LAMBDA-LIST is more robust against types defined in low debug situations.
- bug fix: ENSURE-GENERIC-FUNCTION ensures that the allocation of a generic function does not happen in an arena. (Thanks to Andreas Franke)
- optimization: actually return early when we hit the cache for a :MAKUNBOUND slot access. (thanks to John Mallery)
- optimization: streams opened with WITH-OPEN-FILE avoid having finalizers.
- optimization: improvement of COUNT on non-simple bit-vectors, or simple ones with :START/:END arguments. (#2142062, thanks to Andrew Berkley)
- optimization: SIMD routines for checking strings for ASCII-only content are more compact.
- optimization: the reader prefers base-string symbol-names of uninterned symbols if possible.
- optimization: (length (remove-duplicates a s)) doesn't cons an intermediate sequence.
- optimization: REMOVE-DUPLICATES uses hash-tables in more situations.
- optimization: UPGRADED-ARRAY-ELEMENT-TYPE is now faster.
New in version 2.6.2, 2026-02-27
- minor incompatible change: IMAGPART of a negative float returns 0.0, not -0.0, consistent with a treatment of reals as complexes with an imaginary part of strictly 0.0, but strictly incompatible with the requirement that the IMAGPART equal (* 0 <float>)
- platform support:
- support for Windows on arm64 has been added. (thanks to Masatoshi SANO)
- various mismatches and bugs related to mismatches between Win32 and Unix have been addressed.
- fixed an issue in unsigned 32-bit compare-and-swap on RISC-V (thanks to Andreas Schwab) and LoongArch.
- fixed the integration of the system with the (lack of) floating point traps on RISC-V. (thanks to Andreas Schwab)
- implemented the missing runtime breakpoint-related functions on RISC-V. (#2130944)
- fix for assembling large relative jumps on MIPS.
- fix for GC safety of function calling on RISC-V and LoongArch. (reported by Will Sinatra)
- support little-endian PPC64 to write cores in ELF format.
- fix for SB-POSIX:STAT on Windows with the UCRT C library.
- numerous other fixes related to architecture definitions, particularly on RISC-V, but also on LoongArch, MIPS, PPC and PPC64, and ARM64.
- enhancement: support stack allocation of results for struct return values from alien functions. (thanks to Jesse Bouwman)
- bug fix: rounding of floats converted from ratios. (#2139007)
- bug fix: SCALE-FLOAT and RATIONALIZE on denormals gave wrong answers, and converting ratios to denormals is both more correct and faster.
- bug fix: the ~E FORMAT directive scales its float more correctly. (#1854151, #2016431, #2125287, reported by Michał "phoe" Herda, Robert Dodier and Francis Wright)
- bug fix: the error when failing to bind in DESTRUCTURING-BIND is now a PROGRAM-ERROR.
- bug fix: the compiler respects INLINE and NOTINLINE declarations to control compiler macros that apply to macros. (#1490345)
- bug fix: converting bignums to float will trap or return floating point infinities as appropriate to the prevailing floating point modes.
- bug fix: division with a (COMPLEX FLOAT) result will be more consistent with results involving negative zeros.
- bug fix: allow the full range of hash values in weak hash-tables with user-defined hash functions. (#2141482, reported by Patrick Poitras)
- bug fix: malformed OPTIMIZE declarations no longer cause the compiler to stop.
- bug fix: symbols with terminating macro characters in their names (or non-terminating ones at the start) print with escapes when PRINT-ESCAPE is true.
- bug fix: documentation issues, in README and the manual. (thanks to Carl Gay)
- bug fix: compiler crash while transforming arithmetic operations on known non-numeric inputs. (#2142297)
- bug fix: unsafe concurrent access to synchronized weak hash tables. (#2142714)
- optimization: avoid consing when right-shifting a bignum gives a fixnum result.
- optimization: various type tests in the presence of intersecting constraints do less redundant work.
New in version 2.6.1, 2026-01-26
- minor incompatible change: the never-documented :NO-CONSTRUCTOR-DEFUN option to DEFSTRUCT is no longer supported.
- platform support:
- support for the LoongArch architecture has been added. (thanks to ZiLong Wang)
- support for FreeBSD on 32-bit and 64-bit PowerPC platforms has been added. (thanks to Piotr Kubaj)
- SB-ALIEN now supports passing and returning structures by value (rather than by reference) in accordance with the platform ABIs, on x86-64 and arm64 platforms. (#313202, thanks to Jesse Bouwman)
- wide compare-and-exchange is supported on the arm64 platform for Armv8.1-A or later.
- bug fix: signal a DIVISION-BY-ZERO error in calls of (/ 0 0). (#2137266, reported by khbit)
- bug fix: compiler infinite loop from previously-acceptable recursive inline expansions. (#2137380, reported by Frode Fjeld)
- bug fix: unnormalized inequality constraints causing a compiler infinite loop. (#2137422, reported by Jesse Bouwman)
- bug fix: compiler infinite loop from a deleted loop with an remnant TAGBODY. (#2137493)
- bug fix: miscompilation of REDUCE #'APPEND with non-null INITIAL-VALUE and :FROM-END NIL. (#2137736, reported by Vasily Postnicov)
- bug fix: encode XREF locations in a way that allows for higher form numbers. (#2137765, reported by Jim White)
- bug fix: lookups of optimized memory movers fails when PRINT-BASE is not 10. (#2138812, reported by Robert Dodier)
- bug fix: CLRHASH on a hash table with weak keys should not corrupt the table's index vector. (#2138965, reported by Patrick Poitras)
- bug fix: address a number of type derivation subtleties related to floating point zeros.
- optimization: various combinations of arithmetic operations with variables and constants are simplified to have fewer calls.
- optimization: elide full calls to CONJUGATE when its argument is known to be of type REAL. (#2137354, thanks to Vasily Postnicov)
- optimization: the compiler knows more about the type and results of ARRAY-DIMENSION.
- optimization: transform ARRAY-DIMENSIONS away on objects with array types of known dimensions. (#2138581, thanks to Vasily Postnicov)
- optimization: (equal (array-dimensions a) (array-dimensions b)) does not cons intermediate lists.
- optimization: (make-array (array-dimensions a)) does not cons a list.
New in version 2.6.0, 2025-12-28
- enhancement: the compiler will recognize certain combinations of high-level optimizations as expressible by shorter machine instruction sequences, documented in the manual under "Recognized idioms" in the "Efficiency" section.
- enhancement: the SB-COVER code coverage tool can emit a report in a format compatible with the LCOV open-source tool.
- bug fix: compiled code calling EXPT with constant integer exponent (or 1/2) and floating point base is more consistent with out-of-line EXPT. (#1899969, #2136082)
- bug fix: fix SCALE-FLOAT on denormal floats. (#2000178, re-reported by Barton Willis)
- bug fix: the system's test of constructing an ELF core is compatible with storing code coverage information. (#2131956)
- bug fix: inconsistent result from SUBTYPEP on array types. (#2132250)
- bug fix: the SB-COVER reporting utilities can now annotate source files containing array literals using #A(<dims> <eltype> . <contents>) syntax. (#2134290)
- bug fix: compiler error resulting from losing some already-computed derived type information. (#2136852)
- bug fix: miscompilation of DPB involving non-word-sized intermediate results but a word-sized final result. (#2137028)
- bug fix: compiler error when asserting the result of a known non-list to be of a type union involving a CONS with a given CAR. (#2137030)
- bug fix: miscompilation of DPB with constant byte positions above the number of bits in a word. (#2137046)
- bug fix: miscompilation of PHASE with a negative zero argument. (#2137068, #2137119)
- bug fix: failure to round-trip types involving positive and negative zeros of different floating point representations. (#2137140)
- optimization: machine arithmetic can be used when bit-shifting bignum inputs in a modular arithmetic context.
- optimization: extending an association list, including using backquote notation, is recognized as ACONS and is potentially stack-allocatable.
- optimization: some intermediate copies of lists are elided for calls to maybe-copying operations surrounded by a call to COPY-LIST.
- optimization: a number of comparison operations on rationals are simplified where possible.
- optimization: a number of arithmetic operations recognize and elide double negations or calls to ABS.
- optimization: tracking code with coverage information uses a weak vector per fasl file, rather than a list of per-function weak pointers.
- optimization: REDUCE has been tweaked for better performance both on lists and vectors.
- optimization: for simple-bit-vectors of the right alignment and length, REVERSE will operate a word-at-a-time.
New in version 2.5.11, 2025-11-30
- incompatible change: the compiler's internal representation of "source paths" for unquoted forms within backquotes has changed. Other developer tools using this representation, including callers of some exported SB-INTROSPECT functions, will misreport the location of signalled conditions and/or definitions in top-level forms including backquotes and commas.
- minor incompatible change: undefined syntaxes following READ-BASE-related reader macros (such as #B, #O, #X, #R) now signal reader errors.
- minor incompatible change: the convenience reader syntax pkg::(...) no longer triggers package locks for the PKG package.
- minor incompatible change: building with the SB-DEVEL feature inhibits identical code folding at the end of the build of the SBCL system itself.
- enhancement: improve the compiler's knowledge of the dimensions of the result of MAKE-ARRAY. (#2130477, thanks to Vasily Postnicov)
- enhancement: the SB-COVER contributed module has been made substantially more robust; collecting coverage no longer inhibits various CLOS optimizations. (For SBCL developers, it is now capable of reporting on the coverage of the SBCL system itself, provided it is built with the new :SB-COVER-FOR-INTERNALS build-time feature.)
- bug fix: REQUIREing the SB-MD5 contributed module no longer installs a compiler optimization policy restriction of SPACE being at least 1.
- bug fix: don't miscompute the sizes of garbage collector data structures for running with dynamic space heap sizes above 128GiB.
- bug fix: ENOUGH-NAMESTRING when the pathname and defaults arguments are both logical pathnames with the same host returns a shorter string than previously.
- bug fix: the compiler retains fewer temporary data structures when compiling code with coverage data.
- bug fix: requiring the SB-MD5 contrib module no longer installs a restriction on the SPACE optimization quality.
- bug fix: internal compiler floating point error while compiling certain calls to CEILING. (#2132231)
- bug fix: miscompilation of TYPEP on intersections of types including rational ranges. (#2132207)
- bug fix: miscompilation of MISMATCH from insufficiently cautious type derivation. (#2132187)
- bug fix: internal compiler error while compiling ASH from an incorrect consistency check. (#2132156)
- bug fix: internal compiler error from missing constant-folding stub for internal function. (#2132126)
- bug fix: miscompilation of GET-PROPERTIES at low SAFETY optimization qualities. (#2131985)
- bug fix: internal compiler error while generating code for multiplications of fixnums where the result is also asserted to be a fixnum. (#2131894)
- bug fix: the asserted real range of ATANH was incorrectly stated as [-1,1]. (#2131711)
- bug fix: incorrect type error resulting from wrong type derivation of REDUCE #'LOGIOR. (#2131699)
- bug fix: specialized XEPs should not be generated during block-compilation or interpretation. (#2131118)
- bug fix: fix a wrong compiler transform for MAKE-ARRAY leading to miscompilation. (#2131048)
- bug fix: miscompilation of type checks of (UNSIGNED-BYTE <X>) for large <X>. (#2130028)
- bug fix: it is now possible to inline functions containing SPECIAL declarations on symbols in locked packages. (#2129827, reported by Scott L. Burson)
- bug fix: various internal compiler errors. (#2132022, #2129649)
- bug fix: various wrong results and inconsistencies from SUBTYPEP on ARRAY types. (#2130374, #2130424, #2130441, #2130457, #2130508, #2130518, #2130526, #2130542, #2130588, #2130950, #2131079, #2131149, #2131179, #2131186, #2131271, #2131618, #2130262)
- optimization: LOGTEST participates in compiler transforms related to modular (machine-sized) arithmetic.
- optimization: more arithmetic combinations simplifications.
- optimization: (car (list a)) doesn't allocate a list.
New in version 2.5.10, 2025-10-27
- platform support:
- handling of "./" and "../" in pathname functions on Windows is improved. (#2125908, reported by khbit)
- use x29 for the control frame pointer on arm64, improving backtrace tooling.
- provide a plugin to lldb to display backtraces (contrib/lldb_bt.py).
- an experimental option for performing GC without interrupting and stopping foreign function calls. Enabled via --with-nonstop-foreign-call (for arm64, x86-64 (outside of Windows, where that's already the case.))
- bug fix: some interactions between TWO-WAY-STREAM (and ECHO-STREAM) and user-defined streams have been cleaned up.
- bug fix: the SB-COVER contributed module can now annotate source files containing COMPLEX literals. (A number of other more minor cosmetic issues have also been fixed).
- bug fix: compiler crash from reoptimizing with some stale type information. (#2125944)
- optimization: SLOT-VALUE calls with known slot-name on values which are of type (OR NULL <STRUCT>) are transformed to a null check and a structure access.
- optimization: the compiler will apply constraints to the result of calling RANDOM. (#2126978, thanks to Vasily Postnicov)
- optimization: the compiler will perform type derivation on CL:APPLY called with a known function.
- optimization: fusion of type checking and move of 64-bit integers is enabled on arm64 and x86-64.
- optimization: allocation fusion for (PUSH (CONS A B) LIST) on x86-64.
- optimization: improvements of type derivation for float rounding operations.
- optimization: constant folding when one of the arguments is (if v constant1 constant2)
New in version 2.5.9, 2025-09-29
- minor incompatible change: remove (SETF SB-EXT:POSIX-GETENV), which only ever existed as an operator in SBCL on Windows.
- minor incompatible change: (LOG -0.0) now returns SINGLE-FLOAT-NEGATIVE-INFINITY, more consistently with IEEE 754.
- minor incompatible change: (EXPT 0.0 0.0) now returns 1.0, rather than signalling an error.
- platform support
- restore functionality on NetBSD. (thanks to Masatoshi SANO)
- fix building SBCL as a shared library on ARM64. (#2122059, reported by Guillaume LE VAILLANT)
- optimization: TYPEP with array types does less work in many cases.
- optimization: COMPLEMENT on a known function can be transformed away in more cases.
- optimization: calls to local functions with &REST arguments can be optimized in more cases.
- optimization: bound checks can be eliminated in ROW-MAJOR-AREF based on constraints relating the index to the available array elements. (#2121253, thanks to Vasily Postnicov)
- optimization: function type declarations no longer inhibit inlining local functions. (#2121351, reported by kbhit)
- optimization: bulk movement of memory in the system is implemented with less overhead around memmove().
- optimization: MAKE-ARRAY with dimensions coming from ARRAY-DIMENSIONS on an array with known dimensions avoids consing an intermediate dimensions list.
- optimization: a number of arithmetic operators and relations in combination with some constant arguments do partial expression simplification at compile-time. (#2122063 for %NEGATE thanks to Vasily Postnicov)
New in version 2.5.8, 2025-08-29
- minor incompatible change: SB-THREAD:MAIN-THREAD-P can only be applied to threads, not arbitrary lisp objects.
- minor incompatible change: the instruction-combining (peephole) optimization pass does not run if COMPILATION-SPEED has a higher value than SPEED.
- platform support:
- on arm64, provide better backtraces in the statistical profiler's reporting, along with better detection of assembly routines, local functions and callers of foreign code.
- on ppc64le, make --dynamic-space-size behave as documented. (#2121255)
- on x86-64, handle more cases in the ALU+TEST peephole optimization.
- bug fix: for file-streams with :DIRECTION :IO, input and output file positions should no longer get out of sync. (#1600610, reported by Guillaume le Vaillant, test cases by Brent Benson)
- bug fix: an infinite loop in SUBTYPEP for types involving negations of CONS of specialized ARRAY types. (#2114755)
- bug fix: miscompilation of a CASE form with small numeric keys. (#2119035)
- bug fix: anonymous alien structs definitions are deduplicated, making it harder to overflow internal data structures. (#2114943, reported by Brooke Tilley)
- bug fix: allow ALU+TEST peephole optimizations to fire on x86-64. (#2120547, reported by Christoph Breitkopf)
- bug fix: miscompilation of a LOOP form with rational arithmetic on variables involved in termination tests. (#2121178, reported by 3b)
- bug fix: the compiler is better able to associate some forms in a macroexpanion with the original sources.
- optimization: improve array construction with LIST or SEQUENCE :INITIAL-CONTENTS.
New in version 2.5.7, 2025-07-26
- enhancement: the encapsulate mechanism can be used to wrap functions that are currently not FBOUNDP.
- bug fix: internal compiler error in a failure of stack analysis during propagation of dynamic-extent. (#2113935)
- bug fix: address regression in type inference for TRUNCATE and other division-related operators. (#2115305, reported by Vasily Postnicov)
- bug fix: cleanup of the main thread is performed more carefully when SBCL is used as a shared library. (#2115669, reported by Fedorov Alexander)
- bug fix: the compiler does not lose track of the types of specialized external entry points for user-defined functions. (#2115955, reported by Matt Kaufmann)
- bug fix: adjust compiler template argument acceptability for increased usage scope. (#2116150)
- bug fix: provide a stub for ROTATE-RIGHT-WORD for constant-folding during compilation. (#2117080)
- bug fix: provide a stub for %MAKE-DOUBLE-FLOAT for constant-folding during compilation. (reported by Eric Marsden)
- bug fix: don't loop infinitely in the presence of type-mismatching circular #S read syntax. (reported by Bohong Huang)
- optimization: calls to SLOT-VALUE (and related functions) within methods, on values that are not a specialized argument to those methods, are optimized similarly to calls to SLOT-VALUE in non-method code.
- optimization: calls to REPLACE with VECTOR first argument and LIST second argument are improved.
- optimization: TYPECASE over a set of structure types known not to be extensible is converted to an array lookup.
New in version 2.5.6, 2025-06-29
- enhancement: the compiler now recognizes when local functions (both named and anonymous) are used only as downward funargs in many situations and can stack allocate such closures even without explicit dynamic extent declarations. See the updated manual entry on stack allocation for more information and how user-code can declare funargs as downward.
- minor incompatible change: optimization notes for a variable declared to be of type LIST will not be emitted for various transforms which are defined to operate on (OR NULL VECTOR).
- minor incompatible change: some forms, including a THE form with an invalid type specifier, or a CASE form with bad entries, no longer produce a runtime error. (They continue to provide a full warning at compile-time).
- platform support
- on arm64, breakpoint-based stepping is now thread-safe.
- on arm64, backtraces after interrupts should be more correct.
- on x86-64 with the immobile-space feature, calling from assembly routines to lisp routines is more efficient.
- if arenas are enabled, users can define a lisp function to act as a handler to customize behaviour on arena exhaustion.
- bug fix: address several bugs related to dynamic-extent declarations, inference, and stack allocation.
- bug fix: the stack return page protection is temporarily disabled during GC, so that GC can complete even if it needs to write in the return page.
- bug fix: the compiler generates code to the right entry point for specialized functions. (#2111876, reported by Matt Kaufmann)
- bug fix: the compiler's constraint derivation would sometimes not terminate. (#2113747)
- bug fix: when the runtime structure representing a thread is re-used, the stack guard pages are restored.
- bug fix: type checks for &OPTIONAL arguments are done only once.
- bug fix: CEILING's docstring was wrong. (reported by Dave Tenny)
- bug fix: APPLY could be called with too many arguments when parsing MEMBER type specifications. (reported by Zach Beane)
- bug fix: the compiler could allow constant values of bad types to trigger optimizations. (#2113977)
- bug fix: internal compiler error when attempting to write out a type check for a value already proved to never exist (i.e. be of type NIL). (#2112475)
- optimization: better division for signed-word dividends and unsigned-word divisors on arm64 and x86-64.
- optimization: improvements to subtraction involving bignums and words on x86-64.
- optimization: perfect-hash-based transformations are applied to sequence functions with keys including fixnums and characters as well as symbols.
New in version 2.5.5, 2025-05-31
- minor incompatible change: the output from TRACE is now prefixed by a FRESH-LINE on TRACE-OUTPUT.
- platform support:
- On Linux, the system is better at negotiating with the kernel to find locations for Lisp memory spaces, succeeding more often than previously.
- bug fix: resolve signed/unsigned char mismatch in RUN-PROGRAM on Windows. (#2110525, reported by awlygj)
- bug fix: compiler confusion given sufficiently complex derived type constraints. (#2109902)
- bug fix: compiler inconsistency in low-level representation leading to inconsistent transformations. (#2109837)
- bug fix: return NIL from calls to DOCUMENTATION on illegal function names.
- bug fix: calls to APPLY or VALUES-LIST on some combinations of constant arguments could lose the constant nature after transformation. (thanks to Hayley Patton)
- optimization: some micro-improvements to bignum operations, particularly on x86-64 and arm64
- optimization: allow the result of MAKE-STRING to be allocated on the stack when :element-type is unknown.
- optimization: the compiler will recognize the use of ZEROP on the results of LENGTH and REM (on suitable operands) to avoid full computation of the intermediate result.
New in version 2.5.4, 2025-04-28
- enhancement: :FUN-END breakpoints now support the known values return convention when DEBUG > 0. This means that tracing local functions works in more situations.
- platform support:
- on x86-64, relocation of static space is always enabled.
- save-lisp-and-die with :callable-exports can be used for sbcl.dll on Windows.
- Building with UCRT64 on Windows is now fully supported.
- bug fix: :FUN-END breakpoints work on PowerPC, SPARC, and MIPS again.
- bug fix: incorrect rounding when converting some bignums to floats.
- bug fix: the second value of the truncation functions is more consistently computed for bignum floats.
- bug fix: fix code generation for constants being considered from conflicting type propagation information. (#2107652)
- bug fix: fix 32-bit range check code generation on x86-64. (#2106432)
- bug fix: types are correctly propagated from the keyword argument processor to their uses. (#2106358, reported by Vasily Postnicov)
- bug fix: fix compilation error from CHECK-TYPE when the value checked is a keyword argument and the type specifier argument is not a valid type specifier. (#2104089)
- bug fix: generate stack-manipulation code in the presence of non-local exits and dynamic-extent declarations even more carefully. (#2043242)
- optimization: (LOGIOR A (- (MASK-FIELD (BYTE 1 constantN) A))), or its equivalent (LOGIOR A (- (LOGAND (ASH 1 constantN) A))), is recognized as an idiom for sign-extending the N+1-bit field in A, and can be used for signed modular arithmetic.
- optimization: ROUND is faster for floats.
- optimization: TRUNCATE/FLOOR/etc. are faster on ratios.
- optimization: MAKE-SEQUENCE does not invoke the full type algebra when the provided type specifier is simple.
- optimization: don't attempt to align branch targets if the SPACE optimization quality is greater than 1.
- optimization: circularity detection for printing now places its temporary data structures on the stack.
- optimization: faster GCD on fixnums, especially when the difference in magnitude is large.
- optimization: the implementation of ISQRT has been replaced with the (faster) algorithm currently implemented in CPython.
New in version 2.5.3, 2025-03-30
- enhancement: breakpoint debugger commands have been added. Included is a stepper based on breakpoints requiring no extra instrumentation. However, it still has less functionality than the existing single stepper. See the new debugger manual section titled "Breakpoint Commands" for more information on the new commands.
- minor incompatible change: the behaviour of :save-runtime-options has been restored to match the documentation. (#2096995, reported by Zach Beane)
- minor incompatible change: invoking CHANGE-CLASS from user code no longer grabs the CLOS world lock. Callers must take responsibility for ordering execution of CHANGE-CLASS and any changes to the class hierarchy.
- platform support:
- (CAS SAP) is implemented on ARM v8.1 directly with CAS instructions.
- on x86-64, list constructors emit more compact code sequences, particularly in the presence of multiple references to the same object.
- on x86 and x86-64, fix the stack overflow check to use signed comparisons.
- on Darwin/arm64 and Linux/x86-64, provide a restart to disable floating-point exceptions of the type signalled, and another to disable all floating-point exceptions.
- bug fix: cycle detection in class precedence lists happens before adding classes to the direct subclasses of the parent.
- bug fix: stack-allocated unaligned cons cells no longer cause errors in the debugger.
- bug fix: local function type declarations no longer inhibit tail calls in (SAFETY 0) code. (#2039301)
- bug fix: bad or unknown type specifiers in CHECK-TYPE do not crash or slow down the compiler. (#2102644, #2102653, #2102714, #2104048)
- bug fix: numerous bug fixes relating to the type system's handling of arrays make SUBTYPEP more reliable and less likely to express a contradiction. (#1996980, #2100563, #2100728, #2100779, #2100784, #2100812, #2100825, #2101192, #2101215, #2101803, #2102684)
- bug fix: improve other aspects of the type system's self-consistency. (#2101073, #2101170, #2101183, #2101189, #2101399, #2101589)
- bug fix: fix compiler type error when deriving the type of FTRUNCATE. (#2101073)
- bug fix: fix compiler error when deriving constraints for single-floats. (#2102759)
- bug fix: startup tuning for particular microarchitectures no longer accidentally disables one of the optimizations.
- optimization: ROW-MAJOR-AREF is transformed to use the same array machinery as one-dimensional array references. (Thanks to Scott Burson for the suggestion)
- optimization: list constructors emit shorter code sequences on x86-64, particularly in the presence of multiple references to the same object.
- optimization: FLOOR and CEILING on ratios do not unnecessarily cons.
- optimization: provide specialized CALL-NEXT-METHOD functions for the no-argument and full-argument cases.
New in version 2.5.2, 2025-02-28
- minor incompatible change: in some instances when the compiler cannot prove that a NIL-valued branch is unreachable, where NIL is not compatible with the expected type, a type warning will no longer be issued.
- minor incompatible change: the compiler will more strictly treat type declarations for &OPTIONAL and &KEY arguments in FTYPE declarations, no longer effectively adding an implicit (OR ... <default>) type when the function itself has a default value not matching the declared type for that argument.
- enhancement: type errors in structure constructors are now restartable, with a USE-VALUE restart provided.
- enhancement: CHECK-TYPE warns about type conflicts at compile-time.
- enhancement: FTYPE declarations for functions which set their parameters are checked.
- enhancement: new print control variable SB-EXT:PRINT-CIRCLE-NOT-SHARED, when used in conjunction with PRINT-CIRCLE, prints #1# only for circularities and not simple sharing.
- platform support
- on Windows, make sure to commit memory after zeroing during save-lisp-and-die. (#2097197, reported by _3b)
- on Linux, add the TCP_USER_TIMEOUT constant to SB-BSD-SOCKETS. (thanks to Mihai Bazon)
- on *BSD, make TCP_KEEPCNT, TCP_KEEPIDLE and TCP_KEEPINTVL available where the OS supports it.
- on x86-64, optimize BOUNDP for known-global symbols.
- on x86-64, optimize KEYWORDP for some arguments.
- on arm64, don't trigger an assertion when using FMOV on complex single-float registers.
- on arm64, improve type checking for (AND SYMBOL (NOT NULL)).
- bug fix: using structure read macros with shared structure markers no longer signals type errors when the shared structure is in a slot with a type. (#308936)
- bug fix: non-conforming user macros which modify their source no longer trigger internal errors. (#1371719, reported by _3b)
- bug fix: the combination of CONSTANTLY and DYNAMIC-EXTENT declarations no longer causes an internal compiler error. (#2059950, reported by bohonghuang)
- bug fix: treat inlined functions analogously to constants in the compiler. (#2095560, reported by Vasily Postnicov)
- bug fix: FTYPE declarations for &optional and &key arguments do not include default values when checking types.
- bug fix: Storing coverage data no longer leads to miscompilations allowing reachability of unreachable code. (#2092451, reported by mrkissinger)
- optimization: elide bounds-checking for multidimensional arrays with known dimensions. (reported by aeth)
- optimization: alien callbacks are generally less heavyweight.
- optimization: REMOVE shares the tail of the input list when there's nothing to remove.
New in version 2.5.1, 2025-01-31
- minor incompatible change: SBCL now reveals details of its COMPLEX representations through UPGRADED-COMPLEX-PART-TYPE, rather than hiding them.
- minor incompatible change: the compiler will warn on the use of a SATISFIES type with an undefined function. (#576608, reported by Roman Marynchak)
- minor incompatible change: (room t) now counts the space taken by the internals of hash-tables and CLOS instances.
- platform support
- fixes to the included version of ASDF, and to sockets functions, for the Haiku operating system. (thanks to Alexandru Popa)
- add support for CAS (compare-and-swap) on SAPs for arm64, x86-64 and (partially) RISC-V. (#1894057, reported by Yukari Hafner)
- the system is now consistent with 64-bit time_t on 32-bit linux platforms. (#2063340, reported by Peter van Eynde)
- restore building on 32-bit ARM with newer gcc versions. (#1839783, reported by Sébastien Villemot)
- fix large stack allocation on 64-bit Windows.
- CL portability fixes to the definitions of certain compiler structures, detected by CLISP. (#2064301, #2064312, thanks to Robert Brown)
- bug fix: a misplaced assertion regarding weak hash tables would trigger if a garbage collection hit at just the wrong time. (#2096998)
- bug fix: structure BOA constructors with &REST arguments no longer cause structure slots named NIL or T to be unconditionally initialized with the values NIL and T respectively.
- bug fix: structure BOA constructors without values for some slots no longer cause compilation errors for initforms that are not a single variable.
- bug fix: sequence functions handle :TEST and :TEST-NOT both being given uniformly. (#309143)
- bug fix: the type system is better equipped to handle complicated unions of numeric types. (#308937, #1694839, #1734959, #2073544)
- bug fix: misoptimization of VALUES-LIST in the presence of intervening stack operations. (reported by haruhi.s)
- bug fix: apply the limit to inline expansions more selectively. (#2092518, reported by Andrew Kravchuk)
- bug fix: compiler-detected type mismatches are reported even given the presence of inlined functions. (#2092613, reported by Vasily Postnicov)
- bug fix: improved type error detection for inlined array construction forms. (#2092889, reported by Vasily Postnicov)
- bug fix: accesses to multidimensional arrays are now checked based on the (internal) INSERT-ARRAY-BOUNDS-CHECKS declaration, as with one-dimensional arrays. (#2095155, thanks to Vasily Postnicov)
- bug fix: sb-bsd-sockets:socket-connect handles EINTR caused by GC signals.
New in version 2.5.0, 2024-12-29
- platform support:
- improve support for the Haiku operating system. (thanks to Al Hoang, Estevan Castilho and Alexandru Popa)
- bug fix: generic functions with a large number of required arguments, with methods with specializations on exactly STANDARD-OBJECT or FUNCALLABLE-STANDARD-OBJECT, test the types of their arguments more correctly.
- bug fix: defining a method on SB-MOP:SLOT-VALUE-USING-CLASS where the object argument is specialized to a CONDITION-CLASS no longer leads to an internal error.
- bug fix: the dissassembler on x86-64 correctly disassembles the vcvttpd2dq AVX2 instruction.
- bug fix: ensure that the dispatch function for generic functions is compiled with a known compilation policy. (reported by Neil Goldman)
- bug fix: the compiler retains less intermediate data between COMPILE-FILE forms. (#1557590, reported by andy arvid)
- bug fix: the (invalid) :INITARGS slot option keyword is reported on. (#1887014, reported by Wayne Rittiman, Jr)
- bug fix: the SB-SIMD s16.8-maddubs accepts packs of 16 8-bit quantities, not 8 16-bit quantities. (#2069538, reported by Georgios Makris)
- bug fix: compiling a TYPECASE to dispatch between many user-defined classes no longer takes exponential time. (#2089311, reported by Tomas Hlavaty)
- bug fix: derive the new type for a variable when setting it to a function of its previous version. (#2090997, reported by Vasily Postnicov)
- bug fix: properly clear compiler annotations on variables set to new values involving functions of themselves. (#2090967, reported by Kirill A. Korinskiy)
- bug fix: handle BY in LOOP forms involving iteration on the reverse of a list. (#2091210, reported by James Kalenius)
- bug fix: fix miscompilation of IF where the consequent and alternative would have the same value but for an intervening side-effect. (#2092588, reported by JA)
- optimization: SLOT-VALUE and (SETF SLOT-VALUE) on method arguments specialized to structure classes are compiled to the corresponding structure accessor.
- optimization: calls to SLOT-VALUE (and related operators) on method arguments specialized to instances of SB-MOP:FUNCALLABLE-STANDARD-CLASS are optimized similarly to calls on method arguments specialized to instances of STANDARD-CLASS.
- optimization: (coerce (reverse list) 'vector) doesn't cons a list.
- optimization: (replace vector (reverse list)) doesn't cons a list.
New in version 2.4.11, 2024-11-30
- enhancement: define SB-EXT:DEFAULT-SOURCE-EXTERNAL-FORMAT as the external format for reading source files (for direct use in LOAD and COMPILE-FILE). On Windows, this defaults to an external format with CRLF line-endings. (#720517, reported by Mark David)
- minor incompatible change: the documentation of SB-SEQUENCE:MAKE-SEQUENCE-LIKE has been altered to match its implementation regarding the (un)initialization of the sequence if neither :INITIAL-CONTENTS nor :INITIAL-ELEMENT is provided.
- minor incompatible change: the outputs from SB-GROVEL no longer contain calls to SB-GROVEL::DEFINE-FOREIGN-ROUTINE, but call SB-ALIEN:DEFINE-ALIEN-ROUTINE directly; the definitions of some other SB-GROVEL utilities has also changed.
- platform support:
- The system is more likely to build with the musl C library. (thanks to Masatoshi SANO)
- It is possible to build 32-bit binaries on NetBSD/x86-64 systems. (thanks to Masatoshi SANO)
- Stale big-endian ARM code in callbacks is no longer present. (#2087866, reported by Rongcui Dong)
- Correct the encoding of the VPSHUFD AVX2 instruction. (reported by Dmitry Ignatiev)
- Implement the PINSRQ SSE instruction and provide access to it in SB-SIMD.
- Fix some signed/unsigned and 32-bit issues in the runtime leading to problems with large --dynamic-space-size. (#2087986)
- bug fix: cross-reference information about structure accessors is preserved when compilation policy requires it.
- bug fix: changing &ALLOW-OTHER-KEYS in a generic function's lambda list needs to invalidate the effective methods cache. (reported by Robert Strandh)
- bug fix: calling DISASSEMBLE on a method-function provides a more useful disassembly.
- bug fix: PROCESS-CLOSE no longer leaks a zombie process.
- bug fix: interaction between SYMBOL-MACROLET and SPECIAL declarations is handled more correctly in the code walker. (#1053198)
- bug fix: better scaling when compiling large numbers of calls to local functions. (#1379661, reported by 3b and Burton Samograd)
- bug fix: allow the compiler to approximate types involving large bignums or ratios with large numerator or denominator. (#2085637)
- bug fix: miscompilation of type tests involving STRUCTURE-OBJECT. (#2088417)
- optimization: CONCATENATE with consing arguments can elide some of the intermediate consing.
- optimization: the implementations of various external-formats have been sped up.
- optimization: elide %SAP-ALIEN calls if all uses dereference the resulting ALIEN object.
- optimization: faster (expt integer integer) when computing fixnum results.
- optimization: (ash unknown-integer right) can use modular arithmetic.
- optimization: (apply x ... list) avoids consing intermediate lists in more situations.
- optimizations for arm64, x86-64:
- AREF on non-simple arrays with known element type is faster, along with uses such as LOOP ACROSS, VECTOR-PUSH/POP/EXTEND.
- SIMD variants for POSITION for strings, 8 and 32 bit integer arrays.
- faster overflow checking for (the fixnum (+ fixnum fixnum))
New in version 2.4.10, 2024-10-30
- minor incompatible change: SB-POSIX::POSIX-FORK is no longer exported from SB-POSIX. (The interface function, SB-POSIX:FORK, remains exported).
- platform support:
- fix bugs in instruction encoding on RISC-V; (reported by Guillaume Le Vaillant)
- fix the location of the linkage-table comment in disassembly on 64-bit powerpc;
- elide allocation of empty number stack frames on arm64;
- fix crash on x86 platforms in compiling array dereferencing with computed offsets with negative intermediate results. (#2084943)
- enhancement: the error message from standard object slot typecheck functions in optimized constructors is clearer about the context of the failed type check.
- enhancement: BREAK is no longer tail-called, even when in tail position.
- enhancement: on arm64 and x86-64, specialized entry points for functions known to take or return fixed numbers of double floats are generated and can be automatically called without boxing intermediate floats.
- bug fix: RUN-PROGRAM no longer leaks memory by referencing otherwise unreachable stream instances.
- bug fix: exporting or unexporting symbols during package iteration no longer causes any symbol to be visited more times than expected.
- bug fix: DISASSEMBLE preserves the comment marker across line-breaks for long function or segment names. (#1889456, thanks to Fedorov Alexander)
- bug fix: the compiler no longer loops infinitely trying to compile NOTINLINE calls to known functions with source transform definitions. (#2085451, reported by Fedorov Alexander)
New in version 2.4.9, 2024-09-29
- minor incompatible change: FIND, POSITION (and variants) now check :START and :END arguments for validity as bounding index designators for list sequences.
- platform support:
- improve support for Solaris and variants on x86 and x86-64. (thanks to Masatoshi SANO)
- fix a bug in handling timeouts and interrupted system calls in SB-UNIX:UNIX-SIMPLE-POLL. (#2078824, thanks to Michał phoe Herda)
- fix a bug in the lisp understanding of ssize_t under Windows.
- fix large constant encoding in RISC-V. (#2077307, reported by Guillaume LE VAILLANT)
- more parsimonious low-level type tests on arm64.
- building from the result of git-archive should complete without error.
- bug fix: exporting a symbol during package iteration no longer skips other symbols. (#2080387, reported by kbhit)
- optimization: improvements to EQ hash tables and associated hash functions.
- optimization: type checking of string and string-designator is more efficient.
- optimization: the compiler better understands the nature of the results of CONCATENATE.
New in version 2.4.8, 2024-08-29
- bug fix: the elftool utility finds a writeable directory in more situations. (thanks to Shinmera)
- bug fix: SLOT-MAKUNBOUND does not attempt to dereference a PROGN variable in the interpreter.
- bug fix: READ-SEQUENCE into displaced arrays with a non-zero offset now writes to the right memory location.
- bug fix: fix some erroneous file position calculations in the editcore utility (exposed by a change in the libzstd compression implementation).
- bug fix: do not break the build on STYLE-WARNINGs for earlier SBCL build hosts. (#2064671, reported by Patrick Poitras)
- bug fix: various bug fixes for ppc64le (#2074275, reported by Claude R. C.)
- bug fix: address a rounding error in the TAN type deriver which led to a miscompile in cl-pdf. (#2077100, reported by Gian Pierro Carrubba)
- bug fix: overoptimization of FIND with a :TEST of CHAR-EQUAL. (#2077539)
- optimization: detection of duplicate names in loaded code now scales subquadratically.
- optimization: switch from Floyd's to Brent's cycle detection for lists.
- optimization: EQUAL on lists should be faster.
- optimization: fewer filesystem operations are performed when working out what to LOAD.
- optimization: various microoptimizations on hash tables and associated operations.
- optimization: strings are now hashed using FNV-1A, replacing Bob Jenkins' one-at-a-time hash.
- optimization: fewer redundant validations of the sequence bounding indices in POSITION on strings.
- optimization: many improvements to type derivation on the arguments to and results of standard functions
- optimization: adding many (hundreds) methods to a generic function can be done faster.
New in version 2.4.7, 2024-07-27
- minor incompatible change: the compiler will emit optimization notes related to complex type checks only at high SPEED optimization settings.
- minor incompatible change: the GET-FOREGROUND symbol is now exported from the SB-THREAD package. (thanks to Philipp Marek)
- minor incompatible change: code objects are now printed with their ending address as well as their start address.
- platform support:
- fix occasional saved core corruption on Win32. (reported by Luís Borges de Oliveira)
- address some crashing cases in arena support. (reported by Andreas Franke)
- fix a hard-coded path to
cp. (thanks to Hraban Luyat) - address relocation issues on PIC-flavoured arm64. (#2067153, thanks to leafze)
- fix a crash in the arm64-dissassembler if a code component is missing.
- fix a crash on 32-bit musl libc systems. (reported by Will Sinatra)
- fix building with link-time optimization. (#2072800, reported by Eli Schwartz)
- fix building with newer llvm. (#2071545, reported by Yan)
- mitigate the lack of gc-safety in SB-VM::REMOVE-STATIC-LINKS. (#2045433)
- bug fix: COMPILE installs its second argument as the value of its non-null NAME argument, even if the second argument was already compiled. (#2071324, reported by Tim Bradshaw)
- bug fix: allow the hashing routines in sb-md5 to work on arrays with more than 2^29 elements.
- bug fix: allow READ-SEQUENCE and WRITE-SEQUENCE to read and write files bigger than 4GB.
- bug fix: READ-SEQUENCE returns the index of the first unmodified element of its sequence. (reported by Janis Dzerins)
- optimization: various improvements to floating point rounding routines.
- optimization: PHASE called on the result of COMPLEX now elides the intermediate COMPLEX object.
- optimization: the compiler is more aware of the types of COMPLEX, ATAN, APPEND and NCONC.
New in version 2.4.6, 2024-06-29
- enhancement: name conflicts resulting from colliding symbols in IMPORT and USE-PACKAGE are resolved once for each name, rather than between pairwise colliding symbols.
- enhancement: calls to structure constructors with type mismatches in default initforms cause compile-time warnings.
- platform support:
- fix constant-folding of %log1p and %log2 on 32-bit x86.
- fix the encoding of popcntd on ppc64
- bug fix: EXPORT could be tricked into exporting two distinct symbols of the same name from the same package.
- bug fix: two-argument calls to LOG with arguments of different precision do not lose accuracy through insufficiently-precise intermediate values.
- bug fix: :NEWLINE options in DEFAULT-EXTERNAL-FORMAT are respected when opening files. (reported by Marco Antoniotti)
- bug fix: extend type declarations for the iteration variable of DOLIST with NULL during the evaluation of the result clause. (#942237)
- bug fix: #\uE0 (LATIN CAPITAL LETTER A WITH GRAVE) was incorrectly not downcased with STRING-DOWNCASE. (#2067841, reported by Matt Kaufmann)
- bug fix: backquoted lists as arguments to MAKE-ARRAY were miscompiled. (#2069345, reported by Dan Bothell)
- bug fix: resolve the circularity between the type system and the CLOS metaobject protocol more robustly. (#2069502, reported by Jan Moringen)
- bug fix: misplaced tilde in a logical pathname error message. (#2069995, thanks to Marco Heisig)
- optimization: various speedups to UTF-8 based external format streams and other stream routines.
New in version 2.4.5, 2024-05-30
- incompatible change: attempting to pass or return alien structs by value now signals errors. Previously, such attempts would silently tend to corrupt registers or memory; there is work in progress to allow such calls conforming to the platform ABI. (thanks to Rongcui Dong)
- minor incompatible change: function debug info is compressed only if the system is compiled with libzstd, rather than falling back to a pure-lisp compression implementation.
- minor incompatible change: the compiler will warn in more cases when it can detect incorrectly-typed arguments to FORMAT directives ~C and ~R
- minor incompatible change: strings are converted to a more compact representation if possible before being used as docstrings.
- platform support:
- explicitly include stdlib.h where needed on OpenBSD. (thanks to Sebastien Marie)
- make the editcore utility work on Windows. (thanks to Luís Borges de Oliveira)
- fix memory initialization in RUN-PROGRAM on Windows.
- resolve a deadlock in %INTERRUPT-THREAD on Darwin. (#2062940, reported by Yan)
- hang on startup under some terminal emulators on OS X 14.5. (#2067313, reported by Richard M Kreuter)
- enhancement: when (> debug 1), try to find source locations for code expanded from macros that copy their input forms.
- bug fix: the method-combination generic function caches were missing some of the generic functions constructed during the PCL build itself. (reported by Didier Verna)
- bug fix: SXHASH on equal PATHNAME objects could return different values in different cores. (reported by Luís Borges de Oliveira)
- bug fix: spurious compiler warnings on SORT in some circumstances. (#2065609, reported by William G Lederer)
- optimization: faster bignum addition and negation on arm64, x86-64.
- optimization: faster type tests for (unsigned-byte X) for X being multiples of the platform word size, on arm64, x86-64.
New in version 2.4.4, 2024-04-29
- minor incompatible change: after-GC hooks are now called in the finalizer thread on threaded builds.
- platform support:
- on win32 on 64-bit systems, clear x87 state as well as SSE state after an exception. (#2000435, reported by David Scherfgen)
- enhancement: type assertions resulting from declarations involving variable numbers of multiple values are now checked.
- enhancement: support for memory allocation arenas is now available on the arm64 platform.
- enhancement: the cross-referencing facility is now able to handle anonymous functions.
- enhacnement: the SB-INTROSPECT contrib finds callees in methods of generic functions.
- bug fix: inconsistency in VOP definitions implementing DPB. (#2059842)
- bug fix: MASK-FIELD miscompilation. (#2059849)
- bug fix: caching a state-dependent type function leads to wrong type inference further down the line. (#2059888)
- bug fix: internal consistency violation after failing to apply a MAKE-ARRAY transform. (#2060083)
- bug fix: check the array index before constant-folding an array reference. (#2060347)
- bug fix: don't try to apply type constraints from SATISFIES types. (#2060756)
- bug fix: the JOIN-THREAD-PROBLEM function is now exported from the SB-THREAD package. (#2063385, thanks to Benjamin Lee)
- bug fix: compare-and-swap works as expected on SLOT-VALUE of FUNCALLABLE-STANDARD-OBJECT instances. (reported by qhong)
- bug fix: on arm64, and x86-64 with some non-default build options, integer rounding functions treat minus zero more consistently.
- bug fix: disassembly of stack allocation of vectors caused a crash. (reported by bohonghuang)
- bug fix: don't leak memory when decompressing a compressed core.
- optimization: the amount of space needed for debug-info is about 50% less, leading to a total default image size reduction of more than 10%.
- optimization: on arm64 and x86-64, encoding constants in machine code is sometimes smaller.
- optimization: on arm64 and x86-64, list accumulation is done with less space overhead.
- optimization: APPLY can apply a function to the REST of a &REST list without additional consing.
New in version 2.4.3, 2024-03-29
- enhancement: when dumping debug information to fasl files, the system respects the SOURCE_DATE_EPOCH environment variable (if set) as the latest timestamp to be used.
- contributed module: the sb-perf contributed module, an interface originally by Luke Gorrie to Linux's perf, is now included by default on Linux. (Thanks to Philipp Marek)
- platform support:
- on Linux and BSD variants, timezone querying is now faster.
- Arm64/Darwin: allow configuring the system with a relocatable static space.
- PPC64: allow configuring the system with fasteval.
- bug fix: erroneous transform of EQUALP on characters (#2055425)
- bug fix: float rounders (FTRUNCATE and related functions) handle minus zero more consistently, returning minus zero as the primary value when rounding to zero from a negative value.
- bug fix: type checks for (VECTOR T) were giving the wrong answer for vectors displayed to simple multi-dimensional arrays. (reported by _death)
- bug fix: do not transform away division by BIT when the BIT might still include 0. (#2056184, reported by xizang)
- bug fix: various errors related to imperfections in the perfect hash generator. (#2055794, #2056341)
- bug fix: respect declarations for symbol macros inside defmethod. (#2056514, reported by Jonathan Braud)
- bug fix: failure on x86-64 to set structure slots to floats. (#2058148, reported by Bibek Panthi)
- optimization: improvements to the implementation of CASE.
- optimization: faster PARSE-INTEGER :radix 10/16 on word-sized integers.
- optimization: improvements to LOGBITP.
New in version 2.4.2, 2024-02-29
- bug fix: restore the ability to inherit from both SEQUENCE and SB-MOP:FUNCALLABLE-STANDARD-OBJECT. (#2050088, reported by Christophe Junke)
- bug fix: COERCE will not convert lambda forms to functions if given a type naming a (strict) subclass of FUNCTION.
- bug fix: LOG with a double-float and a ratio argument (in either order) do not lose precision through a single-float intermediate argument.
- bug fix: LOG to the base 2 of integer powers of 2 are more likely to get the mathematically precise answer.
- bug fix: LOG on ratios very near 1 with numerator or denominator being near a power of 2 will use log1p and so will lose less precision.
- bug fix: the utf-8 external format with Unix line-endings updates its character size information when taking the fast path for a buffer of ascii characters. (#2054169, reported by John Carroll)
- bug fix: don't print the contents of a possibly no-longer-valid dynamic-extent cons in PRINT-OBJECT method for THREAD objects. (#2026195, reported by Jake Connor)
- bug fix: place external entry points for functions consistently before any local functions. (#2051169, reported by Fedorov Alexander)
- bug fix: remove unactionable optimization notes for backquoted forms and ordinary calls to APPEND at high speed. (#2051401, reported by Robert Brown)
- bug fix: infinite loop in COPY-SEQ on zero-length arrays of element-type NIL. (#2051759, reported by Devon Sean McCullough)
- bug fix: fix compilation of non-top-level struct constructors. (#2052329, reported by Robert Poitras)
- bug fixes in SB-SIMD:
- optimization: a number of internal tables, particularly those related to Unicode support have been converted to use perfect hash mechanisms, improving both speed and space.
- optimization: FIND, POSITION, ASSOC and RASSOC with constant sequence arguments containing symbols as keys are compiled to perfect hash lookups.
- optimization: the compiler runs a jump-to-jump elimination pass on x86-64.
- system integrity: compiling the system itself on x86-64/linux now produces bitwise-identical cross-compiled fasls whether the build host is cmucl, ccl, clisp or sbcl itself.
New in version 2.4.1, 2024-01-28
- enhancement: compact instance headers are partially supported with the mark-region parallel garbage collector.
- enhancement: functions with declared return types have their return values type-checked in optimization regimes with high SAFETY and (DEBUG 3).
- platform support:
- disable ASLR on FreeBSD. (#2047655, thanks to Konstantin Belousov)
- link to libpthread on FreeBSD. (thanks to Konstantin Belousov)
- restore build on 64-bit riscv. (#2034713, #2048869, reported by Guillaume LE VAILLANT)
- restore build on 64-bit ppc.
- fix case in referring to a header file. (#2047726, thanks to Andrew Kravchuk)
- the fastrem-32 feature (for optimized computations of FLOOR) is now available on all platforms.
- bug fix: resweep moved lines after compaction in the mark-region parallel garbage collector.
- bug fix: infinite loops in the compiler on some constructs with SATISFIES types. (#2047289, #2047706, #2049631)
- optimization: various hash tables implementing part of the system (packages, Unicode data tables) have been converted to use perfect hash functions.
- optimization: TYPECASE on structure class hierarchies is implemented using a perfect hash.
- optimization: eliminate bound checks with relative offsets. (#1830314)
- optimization: the compiler has more knowledge of how to optimize DIGIT-CHAR.
- optimization: the compiler can elide intermediates for some calls to APPLY, CONCATENATE and MAKE-ARRAY with arguments that are freshly-consed modifications of existing sequences.
- optimization: (LOOP FOR X IN (REVERSE LIST) ...) is now faster and conses less.
- optimization: (LOOP ... APPEND ...) is more compact, and does less work if appending NIL.
- optimization: type tests of various array types are faster and shorter.
New in version 2.4.0, 2023-12-28
- minor incompatible change: COMPILE-VERBOSE and LOAD-VERBOSE are bound to NIL when the system is started with the --script command-line argument. (reported by Hraban Luyat, thanks to Nicolas Martyanoff)
- minor incompatible change: when looking for its core file, the system checks the validity of whatever is pointed to by /proc/self/exe, and assesses argv[0] if /proc/self/exe is invalid. (thanks to Philipp Marek)
- minor incompatible change: the system no longer provides type names on the standard (CL) symbols ARRAY-RANK, ARRAY-TOTAL-SIZE, PATHNAME-HOST, PATHNAME-TYPE, PATHNAME-DIRECTORY, FLOAT-RADIX or FLOAT-DIGITS. (#2045559)
- platform support:
- the mark-region parallel garbage collector can be enabled on arm64. (Thanks to Hayley Patton)
- fix build on modern FreeBSDs. (#2046966, thanks to David J. Flander)
- bug fix: restore compiler type inference correctness on calls to REDUCE with :INITIAL-VALUE but no :FROM-END. (#2044856, reported by Patrick Poitras)
- bug fix: compiler error when declaring SB-EXT:MUFFLE-CONDITIONS on an unknown type. (#2045442)
- bug fix: the disassembler provided non-pretty output for registers in some cases. (#2046004, reported by Fedorov Alexander)
- bug fix: the system is slightly less likely to exhaust the stack again when reporting a control stack exhaustion error.
- optimization: GC write barriers are eliminated in more cases.
- optimization: improved type derivation of iteration variables with mixed types.
- optimization: remove unused initial values from LET bindings, improving register type selection.
- optimization: lower EQUALP/EQUAL/EQL to EQL/EQ in FIND/MEMBER based on the input types.
- optimization: better type derivation for DPB, LOGIOR.
New in version 2.3.11, 2023-11-27
- minor incompatible change: streams with an external-format specified with :REPLACEMENT will use their replacement data once per stream unit that causes a decoding error (rather than, in some cases, once for a sequence of bytes none of which is a valid character start position for that external-format).
- minor incompatible change: external-format designators with unsupported or unrecognized options now signal an error when used.
- enhancement: During generic function dispatch, for a generic function using standard- or short-method-combination, if there are no applicable primary methods the system will call the generic function SB-PCL:NO-PRIMARY-METHOD, whose default behaviour is to signal an error. Users may define methods on this generic function.
- enhancement: external formats for unibyte encodings and utf-8 now support newline variants.
- enhancement: character decoding and encoding errors signalled by stream or octet functions now provide a USE-VALUE restart for handlers to provide replacement input or output.
- enhancement: READ-SEQUENCE and WRITE-SEQUENCE support user-defined sequences; the default implementation proceeds element-by-element, reading or writing single bytes or characters to or from the stream as appropriate.
- bug fix: OCTETS-TO-STRING using unibyte external formats with unallocated codepoints (e.g. iso-8859-3) correctly signal or use replacements rather than taking bits from the address of NIL and converting those bits to a character.
- bug fix: FILE-STRING-LENGTH now returns NIL if the input datum is not encodable in the stream's external format.
- bug fix: table-based multibyte external formats (EUC-JP, Shift-JIS, GBK) now honour a replacement character (in the external format or through restarts) when encoding to octets.
- bug fix: converting from octets using the UCS-2, UCS-4 and UTF-32 external formats no longer reads past the end of an octet array with a non-integral number of two- or four-byte units.
- bug fix: converting from octets using the UCS-2, UCS-4, UTF-16 or UTF-32 external formats now returns a simple string, as required by the type declaration of OCTETS-TO-STRING.
- bug fix: providing an invalid external format argument to OPEN or WITH-OPEN-FILE (or the internal MAKE-FD-STREAM) no longer leaks a file descriptor.
- bug fix: SB-ROTATE-BYTE recognizes out-of-relevant-range BYTE specifications for integers before attempting to cons up enormous bignums for masking and shifting. (#2042937)
- bug fix: fix type derivation on compiling SB-ROTATE-BYTE forms with non-zero POSITION in the byte specifier. (#2042775)
- bug fix: fix multiple assembler errors when compiling MAKE-ARRAY, MAKE-STRING and similar with a large constant size. (#2037347, #2038744)
- bug fix: fix internal error when compiling (SETF SBIT) with a large constant index. (#2037415)
- bug fix: fix internal compiler error on invalid lambda list parameters in LABELS. (#2040334)
- bug fix: fix internal compiler error when compiling some infinitely-recursive LABELS forms. (#2042704)
- bug fix: fix internal compiler error when attempting to inline a jump to a label that has been deleted. (#2043262)
- bug fix: FILL-POINTER should never be made to go negative. (#2042452)
- optimization: external formats with :REPLACEMENT no longer bind handlers for coding errors around conversion functions, and so should cons less and be faster.
- optimization: when the :EXTERNAL-FORMAT argument to STRING-TO-OCTETS or OCTETS-TO-STRING is a compile-time constant, the external format is resolved at load time rather than on each call.
- optimization: the compiler is able to constrain the types of inputs to some functions given a derived or asserted type of the function's return value.
- optimization: the compiler performs fewer redundant type checks in ASSOC, GETF and similar functions.
New in version 2.3.10, 2023-10-30
- enhancement: The compiler now allows stack allocating vectors of any size on all safety levels, not just those which it can prove are of sub-page sizes. It can do this because it now inserts code to check for stack overflow explicitly on higher safety levels.
- enhancements to the disassembler:
- on arm64, x86-64, DISASSEMBLE annotates references to static symbols.
- bug fix: calls to generic functions now detect erroneous keywords (in the sense of CLHS 7.6.5) passed as arguments even when auxiliary methods are applicable.
- bug fix: the standard method on SB-MOP:COMPUTE-EFFECTIVE-METHOD no longer inserts calls to implementation-defined local macros. (reported by Daniel Kochmański)
- bug fix: compiler error from state-machine-like LABELS forms in some circumstances. (#2037318)
- bug fix: fix compile-time error in constant-folding RATIONAL on literal float infinities. (#2037455)
- bug fix: failure on x86-64 to assemble code for EQL tests of comparisons with immediates. (#2037456)
- bug fix: infinite loop in the compiler for simplification of type tests of complicated union types. (#2038112, reported by Paul M. Rodriguez)
- bug fix: inability to dump a literal displaced array containing copies of its displacement target. (#2038233, reported by James Kalenius)
- bug fix: compiler error in LOGBITP type derivation. (#2038241)
- bug fix: compiler error in AREF type derivation. (#2038659)
- bug fix: compiler internal consistency failure in overflow type checks. (#2038736)
- bug fix: work around an infinite loop in type simplification by not providing such types from the compiler. (#2038980, reported by Richard Holcombe)
- bug fix: spurious run-time argument count errors from generic function calls on arm64. (#2039006, reported by fiddlerwoaroof)
- bug fix: errors in SCALE-FLOAT on floating-point infinities. (#2039613)
- bug fix: ROOM is slightly more robust to incompletely-initialized instances at the point of running ROOM. (Reported by Andreas Franke)
- bug fix: finalizers saved through SAVE-LISP-AND-DIE and subsequently executed do not trigger memory faults. (Reported by Bohong Huang)
- optimization: improvements to type derivation for ISQRT, INTEGER-LENGTH, LOGCOUNT, LOG, DENOMINATOR.
New in version 2.3.9, 2023-09-28
- enhancement: stack allocation via DYNAMIC-EXTENT now applies to all values that a variable can take on (for example via SETQ), not just the initial binding. This permits for example building complex or recursive structures on the stack more easily via iteration. See the updated manual entry for more details.
- minor incompatible change: some interfaces in the SB-POSIX contrib module adhere to the spec that a NULL result from the C library is an error if and only if errno was altered by the call. SYSCALL-ERROR will be signaled if so.
- enhancement: the SB-POSIX contrib module provides DO-PASSWDS and DO-GROUPS to allow users to iterate over password and group databases safely.
- platform support:
- support for Darwin on x86 and PowerPC has been restored. (#2033287, thanks to Kirill A. Korinsky, Sergey Fedorov and barracuda156)
- bug fix: miscompilation due to erroneous type derivation in the presence of multiplication of fixnums by ratios. (#2033695, reported by Patrick Dussud)
- bug fix: compiler error when compiling signed- and unsigned 64-bit type checks in some cases. (#2033997, reported by Eric Smith)
- bug fix: compiler error when the :INITIAL-CONTENTS argument to MAKE-ARRAY is a constant non-sequence. (#2037328)
- bug fix: compiler error when constant-folding sequence functions with :TEST or :KEY functions erroring on the given sequence. (#2037341)
- bug fix: compiler error when arguments to array or sequence functions imply a very large sequence size. (#2037443, #2037348)
- bug fix: compiler error when the return value of ADJUST-ARRAY is not used. (#2037450)
- optimization: function types derived by the compiler can in some cases be propagated backwards through the intermediate representation.
- optimization: better type derivations for LDB, LOGBITP, RATIO.
- optimization: eliminate bound checks in more cases involving transitive comparisons.
New in version 2.3.8, 2023-08-28
- enhancement: a mark-region parallel garbage collector is available as a build-time option; it can be enabled by adding
--without-gencgc --with-mark-region-gcto the build command line. (Thanks to Hayley Patton) - enhancement: Stack allocation via DYNAMIC-EXTENT is now possible for conditionals even when not all branches are stack-allocatable. Previously all branches needed to be stack-allocatable for the otherwise-inaccessible subparts to get stack allocated.
- platform support:
- on Mac OS X Sonoma, loading the memory image no longer fails. (#2029430, reported by cladur)
- on Darwin, we no longer reimplement nanosleep().
- on PPC64, undefined function errors now work.
- on ARM64/OpenBSD, enable the GCC TLS feature. (thanks to Sébastien Marie)
- when building the system, only display a reasonable amount of timing precision. (thanks to Philipp Marek)
- bug fix: handling of inlining functions compiled to return unboxed values no longer crashes the compiler. (#2029020, reported by Pascal J. Bourguignon)
- bug fix: the source location for code executed within a top-level (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) form is now more useful.
- bug fix: address a race between user threads cancelling finalizers and the finalizer thread executing them. (#2029306)
- bug fix: complex division returns the same value when evaluated inline and out-of-line. (#2030097)
- bug fix: the pretty-printer no longer deletes syntactically-significant whitespace immediately preceding a newline. (#1985814, reported by Mark David)
- optimization: the compiler is more aware of the result type of the NUMERATOR function.
- optimization: EQUAL and EQUALP compile to more efficient code when the two arguments are known to be of the same nullable type.
- optimization: converting bignums to floats uses no intermediate memory.
New in version 2.3.7, 2023-07-29
- minor incompatible change: MACROLET macro functions are now compiled with (SPEED 1), leading to fewer efficiency notes being emitted when compiled in otherwise high-SPEED environments.
- minor incompatible change: when coalescing list data, the file compiler respects substructure equality more accurately, with the side-effect of coalescing along CDR chains as well as CARs. (#2025086)
- minor incompatible change: FUNCTION type declarations for local variables generate assertions around their use when called.
- platform support
- on OpenBSD, the regression test suite expectations have been updated. (#2026809, thanks to Sebastien Marie)
- on OpenBSD, the data limit is now 1GB. (#2027536, thanks to Sebastien Marie)
- on Darwin with the SB-FUTEX feature, do not use unpaired mach_thread_self() syscalls, avoiding resource leaks when creating threads.
- on 64-bit RISCV, add support for some REM-by-multiplication optimizations.
- on Windows, work around a C compiler bug relating to SYSV_ABI.
- bug fix: FILE-POSITION on string output streams no longer crashes or causes arbitrary memory overwrites. (#1839040)
- bug fix: the compiler no longer constant-folds POSITION to NIL if the START or END arguments are not valid.
- optimization: the compiler derives types of &KEY arguments in local calls. (#655562)
- optimization: type tests of values of known union type can be faster if the type being tested for has a non-trivial intersection with the known type.
- optimization: the low-level implementation of NUMBERP, REALP and RATIONALP has been improved on x86-64 and arm64.
- optimization: the compiler removes known-NIL arguments from calls to APPEND and NCONC, and empty sequences from calls to CONCATENATE.
- optimization: checks for symbols being bindable are now memoized, speeding up compiled uses of PROGV.
- optimization: SLOT-VALUE on STRUCTURE-OBJECTs with non-constant slot-name argument is faster.