OCaml releases 4.12.0 and 4.11.2
Florian Angeletti <[email protected]> Wed, 24 Feb 2021 19:03:03 +0100
| Newsgroups | gmane.comp.lang.caml.announce |
|---|---|
| Message-ID | <[email protected]> |
This is a multi-part message in MIME format.
--------------43F7E16313203F4ED826936E
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Dear OCaml users,
We have the pleasure of celebrating the birthday of Jacques de Vaucanson
by announcing the joint releases of OCaml version 4.12.0 and 4.11.2 .
Some of the highlights in the 4.12.0 release are:
- Major progress in reducing the difference between the mainline and
multicore runtime
- A new configuration option `ocaml-option-nnpchecker` which emits an alarm
when the garbage collector finds out-of-heap pointers that could
cause a crash
in the multicore runtime
- Support for macOS/arm64
- Mnemonic names for warnings
- Many quality of life improvements
- Many bug fixes
The 4.11.2 release is a collection of safe bug fixes, cherry-picked from
the 4.12.0 development
cycle. If you were using OCaml 4.11.1 and cannot yet upgrade to 4.12.0,
this release is for you.
The full list of changes can be found in the changelogs below.
Those releases are available as OPAM switches, and as a source download
here:
https://github.com/ocaml/ocaml/archive/4.12.0.tar.gz
https://caml.inria.fr/pub/distrib/ocaml-4.12/
and there:
https://github.com/ocaml/ocaml/archive/4.11.2.tar.gz
https://caml.inria.fr/pub/distrib/ocaml-4.11/
Happy hacking,
-- Florian Angeletti for the OCaml team.
OCaml 4.11.2 (24 February 2021)
-----------------------------------------------
### Build system:
- #9938, #9939: Define __USE_MINGW_ANSI_STDIO=0 for the mingw-w64 ports to
prevent their C99-compliant snprintf conflicting with ours.
(David Allsopp, report by Michael Soegtrop, review by Xavier Leroy)
### Runtime system:
- #10056: Memprof: ensure young_trigger is within the bounds of the minor
heap in caml_memprof_renew_minor_sample (regression from #8684)
(David Allsopp, review by Guillaume Munch-Maccagnoni and
Jacques-Henri Jourdan)
- #9654: More efficient management of code fragments.
(Xavier Leroy, review by Jacques-Henri Jourdan, Damien Doligez, and
Stephen Dolan)
### Tools:
- #9606, #9635, #9637: fix performance regression in the debugger
(behaviors quadratic in the size of the debugged program)
(Xavier Leroy, report by Jacques Garrigue and Virgile Prevosto,
review by David Allsopp and Jacques-Henri Jourdan)
### Code generation and optimizations:
- #9969, #9981: Added mergeable flag to ELF sections containing mergeable
constants. Fixes compatibility with the integrated assembler in
clang 11.0.0.
(Jacob Young, review by Nicolás Ojeda Bär)
### Bug fixes:
- #9970, #10010: fix the declaration scope of extensible-datatype
constructors.
A regression that dates back to 4.08 makes extensible-datatype
constructors
with inline records very fragile, for example:
type 'a t += X of {x : 'a}
(Gabriel Scherer, review by Thomas Refis and Leo White,
report by Nicolás Ojeda Bär)
- #9096, #10096: fix a 4.11.0 performance regression in classes/objects
declared within a function
(Gabriel Scherer, review by Leo White, report by Sacha Ayoun)
- #9326, #10125: Gc.set incorrectly handles the three `custom_*` fields,
causing a performance regression
(report by Emilio Jesús Gallego Arias, analysis and fix by Stephen
Dolan,
code by Xavier Leroy, review by Hugo Heuzard and Gabriel Scherer)
OCaml 4.12.0 (24 February 2021)
-----------------------------------------------
### Supported platforms (highlights):
- #9699: add support for iOS and macOS on ARM 64 bits
(Eduardo Rafael, review by Xavier Leroy, Nicolás Ojeda Bär
and Anil Madhavapeddy, additional testing by Michael Schmidt)
### Standard library (highlights):
- #9797: Add Sys.mkdir and Sys.rmdir.
(David Allsopp, review by Nicolás Ojeda Bär, Sébastien Hinderer and
Xavier Leroy)
* #9765: add init functions to Bigarray.
(Jeremy Yallop, review by Gabriel Scherer, Nicolás Ojeda Bär, and
Xavier Leroy)
* #9668: List.equal, List.compare
(This could break code using "open List" by shadowing
Stdlib.{equal,compare}.)
(Gabriel Scherer, review by Nicolás Ojeda Bär, Daniel Bünzli and
Alain Frisch)
- #9066: a new Either module with
type 'a Either.t = Left of 'a | Right of 'b
(Gabriel Scherer, review by Daniel Bünzli, Thomas Refis, Jeremy Yallop)
- #9066: List.partition_map :
('a -> ('b, 'c) Either.t) -> 'a list -> 'b list * 'c list
(Gabriel Scherer, review by Jeremy Yallop)
- #9865: add Format.pp_print_seq
(Raphaël Proust, review by Nicolás Ojeda Bär)
### Compiler user-interface and warnings (highlights):
- #9657: Warnings can now be referred to by their mnemonic name. The
names are
displayed using `-warn-help` and can be utilized anywhere where a
warning list
specification is expected.
ocamlc -w +fragile-match
...[@@ocaml.warning "-fragile-match"]
Note that only a single warning name at a time is supported for now:
"-w +foo-bar" does not work, you must use "-w +foo -w -bar".
(Nicolás Ojeda Bär, review by Gabriel Scherer, Florian Angeletti and
Leo White)
- #8939: Command-line option to save Linear IR before emit.
(Greta Yorsh, review by Mark Shinwell, Sébastien Hinderer and
Frédéric Bour)
- #9003: Start compilation from Emit when the input file is in Linear IR
format.
(Greta Yorsh, review by Jérémie Dimino, Gabriel Scherer and Frédéric
Bour)
### Language features (highlights):
* #9500, #9727, #9866, #9870, #9873: Injectivity annotations
One can now mark type parameters as injective, which is useful for
abstract types:
module Vec : sig type !'a t end = struct type 'a t = 'a array end
On non-abstract types, this can be used to check the injectivity of
parameters. Since all parameters of record and sum types are by
definition
injective, this only makes sense for type abbreviations:
type !'a t = 'a list
Note that this change required making the regularity check stricter.
(Jacques Garrigue, review by Jeremy Yallop and Leo White)
### Runtime system (highlights):
- #9534, #9947: Introduce a naked pointers checker mode to the runtime
(configure option --enable-naked-pointers-checker). Alarms are printed
when the garbage collector finds out-of-heap pointers that could
cause a crash in no-naked-pointers mode.
(Enguerrand Decorne, KC Sivaramakrishnan, Xavier Leroy, Stephen Dolan,
David Allsopp, Nicolás Ojeda Bär review by Xavier Leroy, Nicolás
Ojeda Bär)
* #1128, #7503, #9036, #9722, #10069: EINTR-based signal handling.
When a signal arrives, avoid running its OCaml handler in the middle
of a blocking section. Instead, allow control to return quickly to
a polling point where the signal handler can safely run, ensuring that
I/O locks are not held while it runs. A polling point was removed from
caml_leave_blocking_section, and one added to caml_raise.
(Stephen Dolan, review by Goswin von Brederlow, Xavier Leroy, Damien
Doligez, Anil Madhavapeddy, Guillaume Munch-Maccagnoni and Jacques-
Henri Jourdan)
* #5154, #9569, #9734: Add `Val_none`, `Some_val`, `Is_none`, `Is_some`,
`caml_alloc_some`, and `Tag_some`. As these macros are sometimes
defined by
authors of C bindings, this change may cause warnings/errors in case of
redefinition.
(Nicolás Ojeda Bär, review by Stephen Dolan, Gabriel Scherer, Mark
Shinwell,
and Xavier Leroy)
* #9674: Memprof: guarantee that an allocation callback is always run
in the same thread the allocation takes place
(Jacques-Henri Jourdan, review by Stephen Dolan)
- #10025: Track custom blocks (e.g. Bigarray) with Memprof
(Stephen Dolan, review by Leo White, Gabriel Scherer and Jacques-Henri
Jourdan)
- #9619: Change representation of function closures so that code pointers
can be easily distinguished from environment variables
(Xavier Leroy, review by Mark Shinwell and Damien Doligez)
- #9654: More efficient management of code fragments.
(Xavier Leroy, review by Jacques-Henri Jourdan, Damien Doligez, and
Stephen Dolan)
### Other libraries (highlights):
- #9573: reimplement Unix.create_process and related functions without
Unix.fork, for better efficiency and compatibility with threads.
(Xavier Leroy, review by Gabriel Scherer and Anil Madhavapeddy)
- #9575: Add Unix.is_inet6_addr
(Nicolás Ojeda Bär, review by Xavier Leroy)
- #9930: new module Semaphore in the thread library, implementing
counting semaphores and binary semaphores
(Xavier Leroy, review by Daniel Bünzli and Damien Doligez,
additional suggestions by Stephen Dolan and Craig Ferguson)
* #9206, #9419: update documentation of the threads library;
deprecate Thread.kill, Thread.wait_read, Thread.wait_write,
and the whole ThreadUnix module.
(Xavier Leroy, review by Florian Angeletti, Guillaume Munch-Maccagnoni,
and Gabriel Scherer)
### Manual and documentation (highlights):
- #9755: Manual: post-processing the html generated by ocamldoc and
hevea. Improvements on design and navigation, including a mobile
version, and a quick-search functionality for the API.
(San Vũ Ngọc, review by David Allsopp and Florian Angeletti)
- #9468: HACKING.adoc: using dune to get merlin support
(Thomas Refis, review by Gabriel Scherer)
- #9684: document in address_class.h the runtime value model in
naked-pointers and no-naked-pointers mode
(Xavier Leroy and Gabriel Scherer)
### Internal/compiler-libs changes (highlights):
- #9464, #9493, #9520, #9563, #9599, #9608, #9647: refactor
the pattern-matching compiler
(Thomas Refis and Gabriel Scherer, review by Florian Angeletti)
- #9696: ocamltest now shows its log when a test fails. In addition, the
log
contains the output of executed programs.
(Nicolás Ojeda Bär, review by David Allsopp, Sébastien Hinderer and
Gabriel
Scherer)
### Build system (highlights):
- #9824, #9837: Honour the CFLAGS and CPPFLAGS variables.
(Sébastien Hinderer, review by David Allsopp)
- #10063: (Re-)enable building on illumos (SmartOS, OmniOS, ...) and
Oracle Solaris; x86_64/GCC and 64-bit SPARC/Sun PRO C compilers.
(partially revert #2024).
(Tõivo Leedjärv and Konstantin Romanov,
review by Gabriel Scherer, Sébastien Hinderer and Xavier Leroy)
### Language features:
- #1655: pattern aliases do not ignore type constraints
(Thomas Refis, review by Jacques Garrigue and Gabriel Scherer)
- #9429: Add unary operators containing `#` to the parser for use in ppx
rewriters
(Leo White, review by Damien Doligez)
### Runtime system:
* #9697: Remove the Is_in_code_area macro and the registration of DLL code
areas in the page table, subsumed by the new code fragment management
API
(Xavier Leroy, review by Jacques-Henri Jourdan)
- #9756: garbage collector colors change
removes the gray color from the major gc
(Sadiq Jaffer and Stephen Dolan reviewed by Xavier Leroy,
KC Sivaramakrishnan, Damien Doligez and Jacques-Henri Jourdan)
* #9513: Selectively initialise blocks in `Obj.new_block`. Reject
`Custom_tag`
objects and zero-length `String_tag` objects.
(KC Sivaramakrishnan, review by David Allsopp, Xavier Leroy, Mark
Shinwell
and Leo White)
- #9564: Add a macro to construct out-of-heap block header.
(KC Sivaramakrishnan, review by Stephen Dolan, Gabriel Scherer,
and Xavier Leroy)
- #9951: Ensure that the mark stack push optimisation handles naked
pointers
(KC Sivaramakrishnan, reported by Enguerrand Decorne, review by Gabriel
Scherer, and Xavier Leroy)
- #9678: Reimplement `Obj.reachable_words` using a hash table to
detect sharing, instead of temporary in-place modifications. This
is a prerequisite for Multicore OCaml.
(Xavier Leroy, review by Jacques-Henri Jourdan and Sébastien Hinderer)
- #1795, #9543: modernize signal handling on Linux i386, PowerPC, and
s390x,
adding support for Musl ppc64le along the way.
(Xavier Leroy and Anil Madhavapeddy, review by Stephen Dolan)
- #9648, #9689: Update the generic hash function to take advantage
of the new representation for function closures
(Xavier Leroy, review by Stephen Dolan)
- #9649: Update the marshaler (output_value) to take advantage
of the new representation for function closures
(Xavier Leroy, review by Damien Doligez)
- #10050: update {PUSH,}OFFSETCLOSURE* bytecode instructions to match new
representation for closures
(Nathanaël Courant, review by Xavier Leroy)
- #9728: Take advantage of the new closure representation to simplify the
compaction algorithm and remove its dependence on the page table
(Damien Doligez, review by Jacques-Henri Jourdan and Xavier Leroy)
- #2195: Improve error message in bytecode stack trace printing and load
debug information during bytecode startup if OCAMLRUNPARAM=b=2.
(David Allsopp, review by Gabriel Scherer and Xavier Leroy)
- #9466: Memprof: optimize random samples generation.
(Jacques-Henri Jourdan, review by Xavier Leroy and Stephen Dolan)
- #9628: Memprof: disable sampling when memprof is suspended.
(Jacques-Henri Jourdan, review by Gabriel Scherer and Stephen Dolan)
- #10056: Memprof: ensure young_trigger is within the bounds of the minor
heap in caml_memprof_renew_minor_sample (regression from #8684)
(David Allsopp, review by Guillaume Munch-Maccagnoni and
Jacques-Henri Jourdan)
- #9508: Remove support for FreeBSD prior to 4.0R, that required explicit
floating-point initialization to behave like IEEE standard
(Hannes Mehnert, review by David Allsopp)
- #8807, #9503: Use different symbols for do_local_roots on bytecode and
native
(Stephen Dolan, review by David Allsopp and Xavier Leroy)
- #9670: Report full major collections in Gc stats.
(Leo White, review by Gabriel Scherer)
- #9675: Remove the caml_static_{alloc,free,resize} primitives, now unused.
(Xavier Leroy, review by Gabriel Scherer)
- #9710: Drop "support" for an hypothetical JIT for OCaml bytecode
which has never existed.
(Jacques-Henri Jourdan, review by Xavier Leroy)
- #9742, #9989: Ephemerons are now compatible with infix pointers occurring
when using mutually recursive functions.
(Jacques-Henri Jourdan, review by François Bobot)
- #9888, #9890: Fixes a bug in the `riscv` backend where register t0 was
not
saved/restored when performing a GC. This could potentially lead to a
segfault.
(Nicolás Ojeda Bär, report by Xavier Leroy, review by Xavier Leroy)
- #9907: Fix native toplevel on native Windows.
(David Allsopp, review by Florian Angeletti)
- #9909: Remove caml_code_area_start and caml_code_area_end globals (no
longer
needed as the pagetable heads towards retirement).
(David Allsopp, review by Xavier Leroy)
- #9949: Clarify documentation of GC message 0x1 and make sure it is
displayed every time a major cycle is forcibly finished.
(Damien Doligez, review by Xavier Leroy)
- #10062: set ARCH_INT64_PRINTF_FORMAT correctly for both modes of
mingw-w64
(David Allsopp, review by Xavier Leroy)
### Code generation and optimizations:
- #9551: ocamlc no longer loads DLLs at link time to check that
external functions referenced from OCaml code are defined.
Instead, .so/.dll files are parsed directly by pure OCaml code.
(Nicolás Ojeda Bär, review by Daniel Bünzli, Gabriel Scherer,
Anil Madhavapeddy, and Xavier Leroy)
- #9620: Limit the number of parameters for an uncurried or untupled
function. Functions with more parameters than that are left
partially curried or tupled.
(Xavier Leroy, review by Mark Shinwell)
- #9752: Revised handling of calling conventions for external C functions.
Provide a more precise description of the types of unboxed arguments,
so that the ARM64 iOS/macOS calling conventions can be honored.
(Xavier Leroy, review by Mark Shinwell and Eduardo Rafael)
- #9838: Ensure that Cmm immediates are generated as Cconst_int where
possible, improving instruction selection.
(Stephen Dolan, review by Leo White and Xavier Leroy)
- #9864: Revised recognition of immediate arguments to integer operations.
Fixes several issues that could have led to producing assembly code
that is rejected by the assembler.
(Xavier Leroy, review by Stephen Dolan)
- #9969, #9981: Added mergeable flag to ELF sections containing mergeable
constants. Fixes compatibility with the integrated assembler in
clang 11.0.0.
(Jacob Young, review by Nicolás Ojeda Bär)
### Standard library:
- #9781: add injectivity annotations to parameterized abstract types
(Jeremy Yallop, review by Nicolás Ojeda Bär)
* #9554: add primitive __FUNCTION__ that returns the name of the current
method
or function, including any enclosing module or class.
(Nicolás Ojeda Bär, Stephen Dolan, review by Stephen Dolan)
- #9075: define to_rev_seq in Set and Map modules.
(Sébastien Briais, review by Gabriel Scherer and Nicolás Ojeda Bär)
- #9561: Unbox Unix.gettimeofday and Unix.time
(Stephen Dolan, review by David Allsopp)
- #9570: Provide an Atomic module with a trivial purely-sequential
implementation, to help write code that is compatible with Multicore
OCaml.
(Gabriel Scherer, review by Xavier Leroy)
- #10035: Make sure that flambda respects atomicity in the Atomic module.
(Guillaume Munch-Maccagnoni, review by Gabriel Scherer)
- #9571: Make at_exit and Printexc.register_printer thread-safe.
(Guillaume Munch-Maccagnoni, review by Gabriel Scherer and Xavier Leroy)
- #9587: Arg: new Rest_all spec to get all rest arguments in a list
(this is similar to Rest, but makes it possible to detect when there
are no arguments (an empty list) after the rest marker)
(Gabriel Scherer, review by Nicolás Ojeda Bär and David Allsopp)
- #9655: Obj: introduce type raw_data and functions raw_field,
set_raw_field
to manipulate out-of-heap pointers in no-naked-pointer mode,
and more generally all other data that is not a well-formed OCaml value
(Xavier Leroy, review by Damien Doligez and Gabriel Scherer)
- #9663: Extend Printexc API for raw backtrace entries.
(Stephen Dolan, review by Nicolás Ojeda Bär and Gabriel Scherer)
- #9763: Add function Hashtbl.rebuild to convert from old hash table
formats (that may have been saved to persistent storage) to the
current hash table format. Remove leftover support for the hash
table format and generic hash function that were in use before OCaml
4.00.
(Xavier Leroy, review by Nicolás Ojeda Bär)
- #10070: Fix Float.Array.blit when source and destination arrays coincide.
(Nicolás Ojeda Bär, review by Alain Frisch and Xavier Leroy)
### Other libraries:
- #8796: On Windows, make Unix.utimes use FILE_FLAG_BACKUP_SEMANTICS flag
to allow it to work with directories.
(Daniil Baturin, review by Damien Doligez)
- #9593: Use new flag for non-elevated symbolic links and test for
Developer
Mode on Windows
(Manuel Hornung, review by David Allsopp and Nicolás Ojeda Bär)
* #9601: Return EPERM for EUNKNOWN -1314 in win32unix (principally affects
error handling when Unix.symlink is unavailable)
(David Allsopp, review by Xavier Leroy)
- #9338, #9790: Dynlink: make sure *_units () functions report accurate
information before the first load.
(Daniel Bünzli, review by Xavier Leroy and Nicolás Ojeda Bär)
* #9757, #9846, #10161: check proper ownership when operating over mutexes.
Now, unlocking a mutex held by another thread or not locked at all
reliably raises a Sys_error exception. Before, it was undefined
behavior, but the documentation did not say so.
Likewise, locking a mutex already locked by the current thread
reliably raises a Sys_error exception. Before, it could
deadlock or succeed (and do recursive locking), depending on the OS.
(Xavier Leroy, report by Guillaume Munch-Maccagnoni, review by
Guillaume Munch-Maccagnoni, David Allsopp, and Stephen Dolan)
- #9802: Ensure signals are handled before Unix.kill returns
(Stephen Dolan, review by Jacques-Henri Jourdan)
- #9869, #10073: Add Unix.SO_REUSEPORT
(Yishuai Li, review by Xavier Leroy, amended by David Allsopp)
- #9906, #9914: Add Unix._exit as a way to exit the process immediately,
skipping any finalization action
(Ivan Gotovchits and Xavier Leroy, review by Sébastien Hinderer and
David Allsopp)
- #9958: Raise exception in case of error in Unix.setsid.
(Nicolás Ojeda Bär, review by Stephen Dolan)
- #9971, #9973: Make sure the process can terminate when the last thread
calls Thread.exit.
(Xavier Leroy, report by Jacques-Henri Jourdan, review by David Allsopp
and Jacques-Henri Jourdan).
### Tools:
- #9551: ocamlobjinfo is now able to display information on .cmxs shared
libraries natively; it no longer requires libbfd to do so
(Nicolás Ojeda Bär, review by Daniel Bünzli, Gabriel Scherer,
Anil Madhavapeddy, and Xavier Leroy)
* #9299, #9795: ocamldep: do not process files during cli parsing. Fixes
various broken cli behaviours.
(Daniel Bünzli, review by Nicolás Ojeda Bär)
### Debugging and profiling:
- #9606, #9635, #9637: fix 4.10 performance regression in the debugger
(behaviors quadratic in the size of the debugged program)
(Xavier Leroy, report by Jacques Garrigue and Virgile Prevosto,
review by David Allsopp and Jacques-Henri Jourdan)
- #9948: Remove Spacetime.
(Nicolás Ojeda Bär, review by Stephen Dolan and Xavier Leroy)
### Manual and documentation:
- #10142, #10154: improved rendering and latex code for toplevel code
examples.
(Florian Angeletti, report by John Whitington, review by Gabriel
Scherer)
- #9745: Manual: Standard Library labeled and unlabeled documentation
unified
(John Whitington, review by Nicolás Ojeda Bär, David Allsopp,
Thomas Refis, and Florian Angeletti)
- #9877: manual, warn that multi-index indexing operators should be
defined in
conjunction of single-index ones.
(Florian Angeletti, review by Hezekiah M. Carty, Gabriel Scherer,
and Marcello Seri)
- #10233: Document `-save-ir-after scheduling` and update `-stop-after`
options.
(Greta Yorsh, review by Gabriel Scherer and Florian Angeletti)
### Compiler user-interface and warnings:
- #1931: rely on levels to enforce principality in patterns
(Thomas Refis and Leo White, review by Jacques Garrigue)
* #9011: Do not create .a/.lib files when creating a .cmxa with no modules.
macOS ar doesn't support creating empty .a files (#1094) and MSVC
doesn't
permit .lib files to contain no objects. When linking with a .cmxa
containing
no modules, it is now not an error for there to be no .a/.lib file.
(David Allsopp, review by Xavier Leroy)
- #9560: Report partial application warnings on type errors in
applications.
(Stephen Dolan, report and testcase by whitequark, review by Gabriel
Scherer
and Thomas Refis)
- #9583: when bytecode linking fails due to an unavailable module, the
module
that requires it is now included in the error message.
(Nicolás Ojeda Bär, review by Vincent Laviron)
- #9615: Attach package type attributes to core_type.
When parsing constraints on a first class module, attributes found
after the
module type were parsed but ignored. Now they are attached to the
corresponding core_type.
(Etienne Millon, review by Thomas Refis)
- #6633, #9673: Add hint when a module is used instead of a module type or
when a module type is used instead of a module or when a class type
is used
instead of a class.
(Xavier Van de Woestyne, report by whitequark, review by Florian
Angeletti
and Gabriel Scherer)
- #9754: allow [@tailcall true] (equivalent to [@tailcall]) and
[@tailcall false] (warns if on a tailcall)
(Gabriel Scherer, review by Nicolás Ojeda Bär)
- #9751: Add warning 68. Pattern-matching depending on mutable state
prevents the remaining arguments from being uncurried.
(Hugo Heuzard, review by Leo White)
- #9783: Widen warning 16 (Unerasable optional argument) to more cases.
(Leo White, review by Florian Angeletti)
- #10008: Improve error message for aliases to the current compilation
unit.
(Leo White, review by Gabriel Scherer)
- #10046: Link all DLLs with -static-libgcc on mingw32 to prevent
dependency
on libgcc_s_sjlj-1.dll with mingw-w64 runtime 8.0.0 (previously this was
only needed for dllunix.dll).
(David Allsopp, report by Andreas Hauptmann, review by Xavier Leroy)
- #9634: Allow initial and repeated commas in `OCAMLRUNPARAM`.
(Nicolás Ojeda Bär, review by Gabriel Scherer)
### Internal/compiler-libs changes:
- #8987: Make some locations more accurate
(Thomas Refis, review by Gabriel Scherer)
- #9216: add Lambda.duplicate which refreshes bound identifiers
(Gabriel Scherer, review by Pierre Chambart and Vincent Laviron)
- #9376: Remove spurious Ptop_defs from #use
(Leo White, review by Damien Doligez)
- #9604: refactoring of the ocamltest codebase.
(Nicolás Ojeda Bär, review by Gabriel Scherer and Sébastien Hinderer)
- #9498, #9511: make the pattern-matching analyzer more robust to
or-pattern explosion, by stopping after the first counter-example to
exhaustivity
(Gabriel Scherer, review by Luc Maranget, Thomas Refis and Florian
Angeletti,
report by Alex Fedoseev through Hongbo Zhang)
- #9514: optimize pattern-matching exhaustivity analysis in the
single-row case
(Gabriel Scherer, review by Stephen DOlan)
- #9442: refactor the implementation of the [@tailcall] attribute
to allow for a structured attribute payload
(Gabriel Scherer, review by Vladimir Keleshev and Nicolás Ojeda Bär)
- #9688: Expose the main entrypoint in compilerlibs
(Stephen Dolan, review by Nicolás Ojeda Bär, Greta Yorsh and David
Allsopp)
- #9715: recheck scope escapes after normalising paths
(Matthew Ryan, review by Gabriel Scherer and Thomas Refis)
- #9778: Fix printing for bindings where polymorphic type annotations and
attributes are present.
(Matthew Ryan, review by Nicolás Ojeda Bär)
- #9797, #9849: Eliminate the routine use of external commands in
ocamltest.
ocamltest no longer calls the mkdir, rm and ln external commands (at
present,
the only external command ocamltest uses is diff).
(David Allsopp, review by Nicolás Ojeda Bär, Sébastien Hinderer and
Xavier Leroy)
- #9801: Don't ignore EOL-at-EOF differences in ocamltest.
(David Allsopp, review by Damien Doligez, much input and thought from
Daniel Bünzli, Damien Doligez, Sébastien Hinderer, and Xavier Leroy)
- #9889: more caching when printing types with -short-path.
(Florian Angeletti, review by Gabriel Scherer)
- #9591: fix pprint of polyvariants that start with a core_type, closed,
not low (Chet Murthy, review by Florian Angeletti)
- #9590: fix pprint of extension constructors (and exceptions) that rebind
(Chet Murthy, review by octachron@)
- #9963: Centralized tracking of frontend's global state
(Frédéric Bour and Thomas Refis, review by Gabriel Scherer)
- #9631: Named text sections for caml_system__code_begin/end symbols
(Greta Yorsh, review by Frédéric Bour)
- #9896: Share the strings representing scopes, fixing some regression
on .cmo/.cma sizes
(Alain Frisch and Xavier Clerc, review by Gabriel Scherer)
### Build system:
- #9332, #9518, #9529: Cease storing C dependencies in the codebase. C
dependencies are generated on-the-fly in development mode. For
incremental
compilation, the MSVC ports require GCC to be present.
(David Allsopp, review by Sébastien Hinderer, YAML-fu by Stephen Dolan)
- #7121, #9558: Always have the autoconf-discovered ld in PACKLD, with
extra flags in new variable PACKLD_FLAGS. For
cross-compilation, this means the triplet-prefixed version will
always be
used.
(David Allsopp, report by Adrian Nader, review by Sébastien Hinderer)
- #9527: stop including configuration when running 'clean' rules
to avoid C dependency recomputation.
(Gabriel Scherer, review by David Allsopp)
- #9804: Build C stubs of libraries in otherlibs/ with debug info.
(Stephen Dolan, review by Sébastien Hinderer and David Allsopp)
- #9938, #9939: Define __USE_MINGW_ANSI_STDIO=0 for the mingw-w64 ports to
prevent their C99-compliant snprintf conflicting with ours.
(David Allsopp, report by Michael Soegtrop, review by Xavier Leroy)
- #9895, #9523: Avoid conflict with C++20 by not installing VERSION to
the OCaml
Standard Library directory.
(Bernhard Schommer, review by David Allsopp)
- #10044: Always report the detected ARCH, MODEL and SYSTEM, even for
bytecode-
only builds (fixes a "configuration regression" from 4.08 for the
Windows
builds)
(David Allsopp, review by Xavier Leroy)
- #10071: Fix bug in tests/misc/weaklifetime.ml that was reported in #10055
(Damien Doligez and Gabriel Scherer, report by David Allsopp)
### Bug fixes:
- #7538, #9669: Check for misplaced attributes on module aliases
(Leo White, report by Thomas Leonard, review by Florian Angeletti)
- #7813, #9955: make sure the major GC cycle doesn't get stuck in Idle
state
(Damien Doligez, report by Anders Fugmann, review by Jacques-Henri
Jourdan)
- #7902, #9556: Type-checker infers recursive type, even though
-rectypes is
off.
(Jacques Garrigue, report by Francois Pottier, review by Leo White)
- #8746: Hashtbl: Restore ongoing traversal status after filter_map_inplace
(Mehdi Bouaziz, review by Alain Frisch)
- #8747, #9709: incorrect principality warning on functional updates of
records
(Jacques Garrigue, report and review by Thomas Refis)
* #8907, #9878: `Typemod.normalize_signature` uses wrong environment
(Jacques Garrigue, report and review by Leo White)
- #9421, #9427: fix printing of (::) in ocamldoc
(Florian Angeletti, report by Yawar Amin, review by Damien Doligez)
- #9440: for a type extension constructor with parameterised arguments,
REPL displayed <poly> for each as opposed to the concrete values used.
(Christian Quinn, review by Gabriel Scherer)
- #9433: Fix package constraints for module aliases
(Leo White, review by Jacques Garrigue)
- #9469: Better backtraces for lazy values
(Leo White, review by Nicolás Ojeda Bär)
- #9521, #9522: correctly fail when comparing functions
with Closure and Infix tags.
(Gabriel Scherer and Jeremy Yallop and Xavier Leroy,
report by Twitter user @st_toHKR through Jun Furuse)
- #9611: maintain order of load path entries in various situations: when
passing
them to system linker, ppx contexts, etc.
(Nicolás Ojeda Bär, review by Jérémie Dimino and Gabriel Scherer)
- #9633: ocamltest: fix a bug when certain variables set in test scripts
would
be ignored (eg `ocamlrunparam`).
(Nicolás Ojeda Bär, review by Sébastien Hinderer)
- #9681, #9690, #9693: small runtime changes
for the new closure representation (#9619)
(Xavier Leroy, Sadiq Jaffer, Gabriel Scherer,
review by Xavier Leroy and Jacques-Henri Jourdan)
- #9739, #9747: Avoid calling type variables, types that are not
variables in
recursive occurrence error messages
(for instance, "Type variable int occurs inside int list")
(Florian Angeletti, report by Stephen Dolan, review by Armaël Guéneau)
- #9759, #9767: Spurious GADT ambiguity without -principal
(Jacques Garrigue, report by Thomas Refis,
review by Thomas Refis and Gabriel Scherer)
- #9799, #9803: make pat_env point to the correct environment
(Thomas Refis, report by Alex Fedoseev, review by Gabriel Scherer)
- #9825, #9830: the C global variable caml_fl_merge and the C function
caml_spacetime_my_profinfo (bytecode version) were declared and
defined with different types. This is undefined behavior and
cancause link-time errors with link-time optimization (LTO).
(Xavier Leroy, report by Richard Jones, review by Nicolás Ojeda Bär)
- #9753: fix build for Android
(Eduardo Rafael, review by Xavier Leroy)
- #9848, #9855: Fix double free of bytecode in toplevel
(Stephen Dolan, report by Sampsa Kiiskinen, review by Gabriel Scherer)
- #9858, #9861: Compiler fails with Ctype.Nondep_cannot_erase exception
(Thomas Refis, report by Philippe Veber, review by Florian Angeletti)
- #9860: wrong range constraint for subtract immediate on zSystems / s390x
(Xavier Leroy, review by Stephen Dolan)
- #9868, #9872, #9892: bugs in {in,out}_channel_length and seek_in
for files opened in text mode under Windows
(Xavier Leroy, report by Alain Frisch, review by Nicolás Ojeda Bär
and Alain Frisch)
- #9925: Correct passing -fdebug-prefix-map to flexlink on Cygwin by
prefixing
it with -link.
(David Allsopp, review by Xavier Leroy)
- #9927: Restore Cygwin64 support.
(David Allsopp, review by Xavier Leroy)
- #9940: Fix unboxing of allocated constants from other compilation units
(Vincent Laviron, report by Stephen Dolan, review by Xavier Leroy and
Stephen Dolan)
- #9991: Fix reproducibility for `-no-alias-deps`
(Leo White, review by Gabriel Scherer and Florian Angeletti)
- #9998: Use Sys.opaque_identity in CamlinternalLazy.force
This removes extra warning 59 messages when compiling afl-instrumented
code with flambda -O3.
(Vincent Laviron, report by Louis Gesbert, review by Gabriel Scherer and
Pierre Chambart)
- #9999: fix -dsource printing of the pattern (`A as x | (`B as x)).
(Gabriel Scherer, report by Anton Bachin, review by Florian Angeletti)
- #9970, #10010: fix the declaration scope of extensible-datatype
constructors.
A regression that dates back to 4.08 makes extensible-datatype
constructors
with inline records very fragile, for example:
type 'a t += X of {x : 'a}
(Gabriel Scherer, review by Thomas Refis and Leo White,
report by Nicolás Ojeda Bär)
- #10048: Fix bug with generalized local opens.
(Leo White, review by Thomas Refis)
- #10106, #10112: some expected-type explanations were forgotten
after some let-bindings
(Gabriel Scherer, review by Thomas Refis and Florian Angeletti,
report by Daniil Baturin)
--------------43F7E16313203F4ED826936E
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: 8bit
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
<div class="moz-text-flowed" style="font-family: -moz-fixed;
font-size: 12px;" lang="x-unicode">Dear OCaml users,
<br>
<br>
We have the pleasure of celebrating the birthday of Jacques de
Vaucanson
<br>
by announcing the joint releases of OCaml version 4.12.0 and
4.11.2 .
<br>
<br>
Some of the highlights in the 4.12.0 release are:
<br>
<br>
- Major progress in reducing the difference between the mainline
and
<br>
multicore runtime
<br>
- A new configuration option `ocaml-option-nnpchecker` which emits
an alarm
<br>
when the garbage collector finds out-of-heap pointers that could
cause a crash
<br>
in the multicore runtime
<br>
- Support for macOS/arm64
<br>
- Mnemonic names for warnings
<br>
- Many quality of life improvements
<br>
- Many bug fixes
<br>
<br>
The 4.11.2 release is a collection of safe bug fixes,
cherry-picked from the 4.12.0 development
<br>
cycle. If you were using OCaml 4.11.1 and cannot yet upgrade to
4.12.0, this release is for you.
<br>
<br>
The full list of changes can be found in the changelogs below.
<br>
<br>
Those releases are available as OPAM switches, and as a source
download here:
<br>
<br>
<a class="moz-txt-link-freetext"
href="https://github.com/ocaml/ocaml/archive/4.12.0.tar.gz">https://github.com/ocaml/ocaml/archive/4.12.0.tar.gz</a>
<br>
<a class="moz-txt-link-freetext"
href="https://caml.inria.fr/pub/distrib/ocaml-4.12/">https://caml.inria.fr/pub/distrib/ocaml-4.12/</a>
<br>
<br>
and there:
<br>
<br>
<a class="moz-txt-link-freetext"
href="https://github.com/ocaml/ocaml/archive/4.11.2.tar.gz">https://github.com/ocaml/ocaml/archive/4.11.2.tar.gz</a>
<br>
<a class="moz-txt-link-freetext"
href="https://caml.inria.fr/pub/distrib/ocaml-4.11/">https://caml.inria.fr/pub/distrib/ocaml-4.11/</a>
<br>
<br>
<br>
Happy hacking,
<br>
<br>
-- Florian Angeletti for the OCaml team.
<br>
<br>
OCaml 4.11.2 (24 February 2021)
<br>
-----------------------------------------------
<br>
<br>
### Build system:
<br>
<br>
- #9938, #9939: Define __USE_MINGW_ANSI_STDIO=0 for the mingw-w64
ports to
<br>
prevent their C99-compliant snprintf conflicting with ours.
<br>
(David Allsopp, report by Michael Soegtrop, review by Xavier
Leroy)
<br>
<br>
### Runtime system:
<br>
<br>
- #10056: Memprof: ensure young_trigger is within the bounds of
the minor
<br>
heap in caml_memprof_renew_minor_sample (regression from #8684)
<br>
(David Allsopp, review by Guillaume Munch-Maccagnoni and
<br>
Jacques-Henri Jourdan)
<br>
<br>
- #9654: More efficient management of code fragments.
<br>
(Xavier Leroy, review by Jacques-Henri Jourdan, Damien Doligez,
and
<br>
Stephen Dolan)
<br>
<br>
### Tools:
<br>
<br>
- #9606, #9635, #9637: fix performance regression in the debugger
<br>
(behaviors quadratic in the size of the debugged program)
<br>
(Xavier Leroy, report by Jacques Garrigue and Virgile Prevosto,
<br>
review by David Allsopp and Jacques-Henri Jourdan)
<br>
<br>
### Code generation and optimizations:
<br>
<br>
- #9969, #9981: Added mergeable flag to ELF sections containing
mergeable
<br>
constants. Fixes compatibility with the integrated assembler in
clang 11.0.0.
<br>
(Jacob Young, review by Nicolás Ojeda Bär)
<br>
<br>
### Bug fixes:
<br>
<br>
- #9970, #10010: fix the declaration scope of extensible-datatype
constructors.
<br>
A regression that dates back to 4.08 makes extensible-datatype
constructors
<br>
with inline records very fragile, for example:
<br>
type 'a t += X of {x : 'a}
<br>
(Gabriel Scherer, review by Thomas Refis and Leo White,
<br>
report by Nicolás Ojeda Bär)
<br>
<br>
- #9096, #10096: fix a 4.11.0 performance regression in
classes/objects
<br>
declared within a function
<br>
(Gabriel Scherer, review by Leo White, report by Sacha Ayoun)
<br>
<br>
- #9326, #10125: Gc.set incorrectly handles the three `custom_*`
fields,
<br>
causing a performance regression
<br>
(report by Emilio Jesús Gallego Arias, analysis and fix by
Stephen Dolan,
<br>
code by Xavier Leroy, review by Hugo Heuzard and Gabriel
Scherer)
<br>
<br>
<br>
<br>
OCaml 4.12.0 (24 February 2021)
<br>
-----------------------------------------------
<br>
<br>
### Supported platforms (highlights):
<br>
<br>
- #9699: add support for iOS and macOS on ARM 64 bits
<br>
(Eduardo Rafael, review by Xavier Leroy, Nicolás Ojeda Bär
<br>
and Anil Madhavapeddy, additional testing by Michael Schmidt)
<br>
<br>
### Standard library (highlights):
<br>
<br>
- #9797: Add Sys.mkdir and Sys.rmdir.
<br>
(David Allsopp, review by Nicolás Ojeda Bär, Sébastien Hinderer
and
<br>
Xavier Leroy)
<br>
<br>
* #9765: add init functions to Bigarray.
<br>
(Jeremy Yallop, review by Gabriel Scherer, Nicolás Ojeda Bär,
and
<br>
Xavier Leroy)
<br>
<br>
* #9668: List.equal, List.compare
<br>
(This could break code using "open List" by shadowing
<br>
Stdlib.{equal,compare}.)
<br>
(Gabriel Scherer, review by Nicolás Ojeda Bär, Daniel Bünzli and
Alain Frisch)
<br>
<br>
- #9066: a new Either module with
<br>
type 'a Either.t = Left of 'a | Right of 'b
<br>
(Gabriel Scherer, review by Daniel Bünzli, Thomas Refis, Jeremy
Yallop)
<br>
<br>
- #9066: List.partition_map :
<br>
('a -> ('b, 'c) Either.t) -> 'a list -> 'b list * 'c
list
<br>
(Gabriel Scherer, review by Jeremy Yallop)
<br>
<br>
- #9865: add Format.pp_print_seq
<br>
(Raphaël Proust, review by Nicolás Ojeda Bär)
<br>
<br>
### Compiler user-interface and warnings (highlights):
<br>
<br>
- #9657: Warnings can now be referred to by their mnemonic name.
The names are
<br>
displayed using `-warn-help` and can be utilized anywhere where
a warning list
<br>
specification is expected.
<br>
ocamlc -w +fragile-match
<br>
...[@@ocaml.warning "-fragile-match"]
<br>
Note that only a single warning name at a time is supported for
now:
<br>
"-w +foo-bar" does not work, you must use "-w +foo -w -bar".
<br>
(Nicolás Ojeda Bär, review by Gabriel Scherer, Florian Angeletti
and
<br>
Leo White)
<br>
<br>
- #8939: Command-line option to save Linear IR before emit.
<br>
(Greta Yorsh, review by Mark Shinwell, Sébastien Hinderer and
Frédéric Bour)
<br>
<br>
- #9003: Start compilation from Emit when the input file is in
Linear IR format.
<br>
(Greta Yorsh, review by Jérémie Dimino, Gabriel Scherer and
Frédéric Bour)
<br>
<br>
### Language features (highlights):
<br>
<br>
* #9500, #9727, #9866, #9870, #9873: Injectivity annotations
<br>
One can now mark type parameters as injective, which is useful
for
<br>
abstract types:
<br>
module Vec : sig type !'a t end = struct type 'a t = 'a array
end
<br>
On non-abstract types, this can be used to check the injectivity
of
<br>
parameters. Since all parameters of record and sum types are by
definition
<br>
injective, this only makes sense for type abbreviations:
<br>
type !'a t = 'a list
<br>
Note that this change required making the regularity check
stricter.
<br>
(Jacques Garrigue, review by Jeremy Yallop and Leo White)
<br>
<br>
### Runtime system (highlights):
<br>
<br>
- #9534, #9947: Introduce a naked pointers checker mode to the
runtime
<br>
(configure option --enable-naked-pointers-checker). Alarms are
printed
<br>
when the garbage collector finds out-of-heap pointers that could
<br>
cause a crash in no-naked-pointers mode.
<br>
(Enguerrand Decorne, KC Sivaramakrishnan, Xavier Leroy, Stephen
Dolan,
<br>
David Allsopp, Nicolás Ojeda Bär review by Xavier Leroy, Nicolás
Ojeda Bär)
<br>
<br>
* #1128, #7503, #9036, #9722, #10069: EINTR-based signal handling.
<br>
When a signal arrives, avoid running its OCaml handler in the
middle
<br>
of a blocking section. Instead, allow control to return quickly
to
<br>
a polling point where the signal handler can safely run,
ensuring that
<br>
I/O locks are not held while it runs. A polling point was
removed from
<br>
caml_leave_blocking_section, and one added to caml_raise.
<br>
(Stephen Dolan, review by Goswin von Brederlow, Xavier Leroy,
Damien
<br>
Doligez, Anil Madhavapeddy, Guillaume Munch-Maccagnoni and
Jacques-
<br>
Henri Jourdan)
<br>
<br>
* #5154, #9569, #9734: Add `Val_none`, `Some_val`, `Is_none`,
`Is_some`,
<br>
`caml_alloc_some`, and `Tag_some`. As these macros are sometimes
defined by
<br>
authors of C bindings, this change may cause warnings/errors in
case of
<br>
redefinition.
<br>
(Nicolás Ojeda Bär, review by Stephen Dolan, Gabriel Scherer,
Mark Shinwell,
<br>
and Xavier Leroy)
<br>
<br>
* #9674: Memprof: guarantee that an allocation callback is always
run
<br>
in the same thread the allocation takes place
<br>
(Jacques-Henri Jourdan, review by Stephen Dolan)
<br>
<br>
- #10025: Track custom blocks (e.g. Bigarray) with Memprof
<br>
(Stephen Dolan, review by Leo White, Gabriel Scherer and
Jacques-Henri
<br>
Jourdan)
<br>
<br>
- #9619: Change representation of function closures so that code
pointers
<br>
can be easily distinguished from environment variables
<br>
(Xavier Leroy, review by Mark Shinwell and Damien Doligez)
<br>
<br>
- #9654: More efficient management of code fragments.
<br>
(Xavier Leroy, review by Jacques-Henri Jourdan, Damien Doligez,
and
<br>
Stephen Dolan)
<br>
<br>
### Other libraries (highlights):
<br>
<br>
- #9573: reimplement Unix.create_process and related functions
without
<br>
Unix.fork, for better efficiency and compatibility with threads.
<br>
(Xavier Leroy, review by Gabriel Scherer and Anil Madhavapeddy)
<br>
<br>
- #9575: Add Unix.is_inet6_addr
<br>
(Nicolás Ojeda Bär, review by Xavier Leroy)
<br>
<br>
- #9930: new module Semaphore in the thread library, implementing
<br>
counting semaphores and binary semaphores
<br>
(Xavier Leroy, review by Daniel Bünzli and Damien Doligez,
<br>
additional suggestions by Stephen Dolan and Craig Ferguson)
<br>
<br>
* #9206, #9419: update documentation of the threads library;
<br>
deprecate Thread.kill, Thread.wait_read, Thread.wait_write,
<br>
and the whole ThreadUnix module.
<br>
(Xavier Leroy, review by Florian Angeletti, Guillaume
Munch-Maccagnoni,
<br>
and Gabriel Scherer)
<br>
<br>
### Manual and documentation (highlights):
<br>
<br>
- #9755: Manual: post-processing the html generated by ocamldoc
and
<br>
hevea. Improvements on design and navigation, including a
mobile
<br>
version, and a quick-search functionality for the API.
<br>
(San Vũ Ngọc, review by David Allsopp and Florian Angeletti)
<br>
<br>
- #9468: HACKING.adoc: using dune to get merlin support
<br>
(Thomas Refis, review by Gabriel Scherer)
<br>
<br>
- #9684: document in address_class.h the runtime value model in
<br>
naked-pointers and no-naked-pointers mode
<br>
(Xavier Leroy and Gabriel Scherer)
<br>
<br>
### Internal/compiler-libs changes (highlights):
<br>
<br>
- #9464, #9493, #9520, #9563, #9599, #9608, #9647: refactor
<br>
the pattern-matching compiler
<br>
(Thomas Refis and Gabriel Scherer, review by Florian Angeletti)
<br>
<br>
- #9696: ocamltest now shows its log when a test fails. In
addition, the log
<br>
contains the output of executed programs.
<br>
(Nicolás Ojeda Bär, review by David Allsopp, Sébastien Hinderer
and Gabriel
<br>
Scherer)
<br>
<br>
### Build system (highlights):
<br>
<br>
- #9824, #9837: Honour the CFLAGS and CPPFLAGS variables.
<br>
(Sébastien Hinderer, review by David Allsopp)
<br>
<br>
- #10063: (Re-)enable building on illumos (SmartOS, OmniOS, ...)
and
<br>
Oracle Solaris; x86_64/GCC and 64-bit SPARC/Sun PRO C compilers.
<br>
(partially revert #2024).
<br>
(Tõivo Leedjärv and Konstantin Romanov,
<br>
review by Gabriel Scherer, Sébastien Hinderer and Xavier Leroy)
<br>
<br>
<br>
### Language features:
<br>
<br>
- #1655: pattern aliases do not ignore type constraints
<br>
(Thomas Refis, review by Jacques Garrigue and Gabriel Scherer)
<br>
<br>
- #9429: Add unary operators containing `#` to the parser for use
in ppx
<br>
rewriters
<br>
(Leo White, review by Damien Doligez)
<br>
<br>
### Runtime system:
<br>
<br>
* #9697: Remove the Is_in_code_area macro and the registration of
DLL code
<br>
areas in the page table, subsumed by the new code fragment
management API
<br>
(Xavier Leroy, review by Jacques-Henri Jourdan)
<br>
<br>
- #9756: garbage collector colors change
<br>
removes the gray color from the major gc
<br>
(Sadiq Jaffer and Stephen Dolan reviewed by Xavier Leroy,
<br>
KC Sivaramakrishnan, Damien Doligez and Jacques-Henri Jourdan)
<br>
<br>
* #9513: Selectively initialise blocks in `Obj.new_block`. Reject
`Custom_tag`
<br>
objects and zero-length `String_tag` objects.
<br>
(KC Sivaramakrishnan, review by David Allsopp, Xavier Leroy,
Mark Shinwell
<br>
and Leo White)
<br>
<br>
- #9564: Add a macro to construct out-of-heap block header.
<br>
(KC Sivaramakrishnan, review by Stephen Dolan, Gabriel Scherer,
<br>
and Xavier Leroy)
<br>
<br>
- #9951: Ensure that the mark stack push optimisation handles
naked pointers
<br>
(KC Sivaramakrishnan, reported by Enguerrand Decorne, review by
Gabriel
<br>
Scherer, and Xavier Leroy)
<br>
<br>
- #9678: Reimplement `Obj.reachable_words` using a hash table to
<br>
detect sharing, instead of temporary in-place modifications.
This
<br>
is a prerequisite for Multicore OCaml.
<br>
(Xavier Leroy, review by Jacques-Henri Jourdan and Sébastien
Hinderer)
<br>
<br>
- #1795, #9543: modernize signal handling on Linux i386, PowerPC,
and s390x,
<br>
adding support for Musl ppc64le along the way.
<br>
(Xavier Leroy and Anil Madhavapeddy, review by Stephen Dolan)
<br>
<br>
- #9648, #9689: Update the generic hash function to take advantage
<br>
of the new representation for function closures
<br>
(Xavier Leroy, review by Stephen Dolan)
<br>
<br>
- #9649: Update the marshaler (output_value) to take advantage
<br>
of the new representation for function closures
<br>
(Xavier Leroy, review by Damien Doligez)
<br>
<br>
- #10050: update {PUSH,}OFFSETCLOSURE* bytecode instructions to
match new
<br>
representation for closures
<br>
(Nathanaël Courant, review by Xavier Leroy)
<br>
<br>
- #9728: Take advantage of the new closure representation to
simplify the
<br>
compaction algorithm and remove its dependence on the page table
<br>
(Damien Doligez, review by Jacques-Henri Jourdan and Xavier
Leroy)
<br>
<br>
- #2195: Improve error message in bytecode stack trace printing
and load
<br>
debug information during bytecode startup if OCAMLRUNPARAM=b=2.
<br>
(David Allsopp, review by Gabriel Scherer and Xavier Leroy)
<br>
<br>
- #9466: Memprof: optimize random samples generation.
<br>
(Jacques-Henri Jourdan, review by Xavier Leroy and Stephen
Dolan)
<br>
<br>
- #9628: Memprof: disable sampling when memprof is suspended.
<br>
(Jacques-Henri Jourdan, review by Gabriel Scherer and Stephen
Dolan)
<br>
<br>
- #10056: Memprof: ensure young_trigger is within the bounds of
the minor
<br>
heap in caml_memprof_renew_minor_sample (regression from #8684)
<br>
(David Allsopp, review by Guillaume Munch-Maccagnoni and
<br>
Jacques-Henri Jourdan)
<br>
<br>
- #9508: Remove support for FreeBSD prior to 4.0R, that required
explicit
<br>
floating-point initialization to behave like IEEE standard
<br>
(Hannes Mehnert, review by David Allsopp)
<br>
<br>
- #8807, #9503: Use different symbols for do_local_roots on
bytecode and native
<br>
(Stephen Dolan, review by David Allsopp and Xavier Leroy)
<br>
<br>
- #9670: Report full major collections in Gc stats.
<br>
(Leo White, review by Gabriel Scherer)
<br>
<br>
- #9675: Remove the caml_static_{alloc,free,resize} primitives,
now unused.
<br>
(Xavier Leroy, review by Gabriel Scherer)
<br>
<br>
- #9710: Drop "support" for an hypothetical JIT for OCaml bytecode
<br>
which has never existed.
<br>
(Jacques-Henri Jourdan, review by Xavier Leroy)
<br>
<br>
- #9742, #9989: Ephemerons are now compatible with infix pointers
occurring
<br>
when using mutually recursive functions.
<br>
(Jacques-Henri Jourdan, review by François Bobot)
<br>
<br>
- #9888, #9890: Fixes a bug in the `riscv` backend where register
t0 was not
<br>
saved/restored when performing a GC. This could potentially lead
to a
<br>
segfault.
<br>
(Nicolás Ojeda Bär, report by Xavier Leroy, review by Xavier
Leroy)
<br>
<br>
- #9907: Fix native toplevel on native Windows.
<br>
(David Allsopp, review by Florian Angeletti)
<br>
<br>
- #9909: Remove caml_code_area_start and caml_code_area_end
globals (no longer
<br>
needed as the pagetable heads towards retirement).
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #9949: Clarify documentation of GC message 0x1 and make sure it
is
<br>
displayed every time a major cycle is forcibly finished.
<br>
(Damien Doligez, review by Xavier Leroy)
<br>
<br>
- #10062: set ARCH_INT64_PRINTF_FORMAT correctly for both modes of
mingw-w64
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
### Code generation and optimizations:
<br>
<br>
- #9551: ocamlc no longer loads DLLs at link time to check that
<br>
external functions referenced from OCaml code are defined.
<br>
Instead, .so/.dll files are parsed directly by pure OCaml code.
<br>
(Nicolás Ojeda Bär, review by Daniel Bünzli, Gabriel Scherer,
<br>
Anil Madhavapeddy, and Xavier Leroy)
<br>
<br>
- #9620: Limit the number of parameters for an uncurried or
untupled
<br>
function. Functions with more parameters than that are left
<br>
partially curried or tupled.
<br>
(Xavier Leroy, review by Mark Shinwell)
<br>
<br>
- #9752: Revised handling of calling conventions for external C
functions.
<br>
Provide a more precise description of the types of unboxed
arguments,
<br>
so that the ARM64 iOS/macOS calling conventions can be honored.
<br>
(Xavier Leroy, review by Mark Shinwell and Eduardo Rafael)
<br>
<br>
- #9838: Ensure that Cmm immediates are generated as Cconst_int
where
<br>
possible, improving instruction selection.
<br>
(Stephen Dolan, review by Leo White and Xavier Leroy)
<br>
<br>
- #9864: Revised recognition of immediate arguments to integer
operations.
<br>
Fixes several issues that could have led to producing assembly
code
<br>
that is rejected by the assembler.
<br>
(Xavier Leroy, review by Stephen Dolan)
<br>
<br>
- #9969, #9981: Added mergeable flag to ELF sections containing
mergeable
<br>
constants. Fixes compatibility with the integrated assembler in
clang 11.0.0.
<br>
(Jacob Young, review by Nicolás Ojeda Bär)
<br>
<br>
### Standard library:
<br>
<br>
- #9781: add injectivity annotations to parameterized abstract
types
<br>
(Jeremy Yallop, review by Nicolás Ojeda Bär)
<br>
<br>
* #9554: add primitive __FUNCTION__ that returns the name of the
current method
<br>
or function, including any enclosing module or class.
<br>
(Nicolás Ojeda Bär, Stephen Dolan, review by Stephen Dolan)
<br>
<br>
- #9075: define to_rev_seq in Set and Map modules.
<br>
(Sébastien Briais, review by Gabriel Scherer and Nicolás Ojeda
Bär)
<br>
<br>
- #9561: Unbox Unix.gettimeofday and Unix.time
<br>
(Stephen Dolan, review by David Allsopp)
<br>
<br>
- #9570: Provide an Atomic module with a trivial purely-sequential
<br>
implementation, to help write code that is compatible with
Multicore
<br>
OCaml.
<br>
(Gabriel Scherer, review by Xavier Leroy)
<br>
<br>
- #10035: Make sure that flambda respects atomicity in the Atomic
module.
<br>
(Guillaume Munch-Maccagnoni, review by Gabriel Scherer)
<br>
<br>
- #9571: Make at_exit and Printexc.register_printer thread-safe.
<br>
(Guillaume Munch-Maccagnoni, review by Gabriel Scherer and
Xavier Leroy)
<br>
<br>
- #9587: Arg: new Rest_all spec to get all rest arguments in a
list
<br>
(this is similar to Rest, but makes it possible to detect when
there
<br>
are no arguments (an empty list) after the rest marker)
<br>
(Gabriel Scherer, review by Nicolás Ojeda Bär and David Allsopp)
<br>
<br>
- #9655: Obj: introduce type raw_data and functions raw_field,
set_raw_field
<br>
to manipulate out-of-heap pointers in no-naked-pointer mode,
<br>
and more generally all other data that is not a well-formed
OCaml value
<br>
(Xavier Leroy, review by Damien Doligez and Gabriel Scherer)
<br>
<br>
- #9663: Extend Printexc API for raw backtrace entries.
<br>
(Stephen Dolan, review by Nicolás Ojeda Bär and Gabriel Scherer)
<br>
<br>
- #9763: Add function Hashtbl.rebuild to convert from old hash
table
<br>
formats (that may have been saved to persistent storage) to the
<br>
current hash table format. Remove leftover support for the hash
<br>
table format and generic hash function that were in use before
OCaml 4.00.
<br>
(Xavier Leroy, review by Nicolás Ojeda Bär)
<br>
<br>
- #10070: Fix Float.Array.blit when source and destination arrays
coincide.
<br>
(Nicolás Ojeda Bär, review by Alain Frisch and Xavier Leroy)
<br>
<br>
### Other libraries:
<br>
<br>
- #8796: On Windows, make Unix.utimes use
FILE_FLAG_BACKUP_SEMANTICS flag
<br>
to allow it to work with directories.
<br>
(Daniil Baturin, review by Damien Doligez)
<br>
<br>
- #9593: Use new flag for non-elevated symbolic links and test for
Developer
<br>
Mode on Windows
<br>
(Manuel Hornung, review by David Allsopp and Nicolás Ojeda Bär)
<br>
<br>
* #9601: Return EPERM for EUNKNOWN -1314 in win32unix (principally
affects
<br>
error handling when Unix.symlink is unavailable)
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #9338, #9790: Dynlink: make sure *_units () functions report
accurate
<br>
information before the first load.
<br>
(Daniel Bünzli, review by Xavier Leroy and Nicolás Ojeda Bär)
<br>
<br>
* #9757, #9846, #10161: check proper ownership when operating over
mutexes.
<br>
Now, unlocking a mutex held by another thread or not locked at
all
<br>
reliably raises a Sys_error exception. Before, it was undefined
<br>
behavior, but the documentation did not say so.
<br>
Likewise, locking a mutex already locked by the current thread
<br>
reliably raises a Sys_error exception. Before, it could
<br>
deadlock or succeed (and do recursive locking), depending on the
OS.
<br>
(Xavier Leroy, report by Guillaume Munch-Maccagnoni, review by
<br>
Guillaume Munch-Maccagnoni, David Allsopp, and Stephen Dolan)
<br>
<br>
- #9802: Ensure signals are handled before Unix.kill returns
<br>
(Stephen Dolan, review by Jacques-Henri Jourdan)
<br>
<br>
- #9869, #10073: Add Unix.SO_REUSEPORT
<br>
(Yishuai Li, review by Xavier Leroy, amended by David Allsopp)
<br>
<br>
- #9906, #9914: Add Unix._exit as a way to exit the process
immediately,
<br>
skipping any finalization action
<br>
(Ivan Gotovchits and Xavier Leroy, review by Sébastien Hinderer
and
<br>
David Allsopp)
<br>
<br>
- #9958: Raise exception in case of error in Unix.setsid.
<br>
(Nicolás Ojeda Bär, review by Stephen Dolan)
<br>
<br>
- #9971, #9973: Make sure the process can terminate when the last
thread
<br>
calls Thread.exit.
<br>
(Xavier Leroy, report by Jacques-Henri Jourdan, review by David
Allsopp
<br>
and Jacques-Henri Jourdan).
<br>
<br>
### Tools:
<br>
<br>
- #9551: ocamlobjinfo is now able to display information on .cmxs
shared
<br>
libraries natively; it no longer requires libbfd to do so
<br>
(Nicolás Ojeda Bär, review by Daniel Bünzli, Gabriel Scherer,
<br>
Anil Madhavapeddy, and Xavier Leroy)
<br>
<br>
* #9299, #9795: ocamldep: do not process files during cli parsing.
Fixes
<br>
various broken cli behaviours.
<br>
(Daniel Bünzli, review by Nicolás Ojeda Bär)
<br>
<br>
### Debugging and profiling:
<br>
<br>
- #9606, #9635, #9637: fix 4.10 performance regression in the
debugger
<br>
(behaviors quadratic in the size of the debugged program)
<br>
(Xavier Leroy, report by Jacques Garrigue and Virgile Prevosto,
<br>
review by David Allsopp and Jacques-Henri Jourdan)
<br>
<br>
- #9948: Remove Spacetime.
<br>
(Nicolás Ojeda Bär, review by Stephen Dolan and Xavier Leroy)
<br>
<br>
### Manual and documentation:
<br>
<br>
- #10142, #10154: improved rendering and latex code for toplevel
code examples.
<br>
(Florian Angeletti, report by John Whitington, review by Gabriel
Scherer)
<br>
<br>
- #9745: Manual: Standard Library labeled and unlabeled
documentation unified
<br>
(John Whitington, review by Nicolás Ojeda Bär, David Allsopp,
<br>
Thomas Refis, and Florian Angeletti)
<br>
<br>
- #9877: manual, warn that multi-index indexing operators should
be defined in
<br>
conjunction of single-index ones.
<br>
(Florian Angeletti, review by Hezekiah M. Carty, Gabriel
Scherer,
<br>
and Marcello Seri)
<br>
<br>
- #10233: Document `-save-ir-after scheduling` and update
`-stop-after` options.
<br>
(Greta Yorsh, review by Gabriel Scherer and Florian Angeletti)
<br>
<br>
### Compiler user-interface and warnings:
<br>
<br>
- #1931: rely on levels to enforce principality in patterns
<br>
(Thomas Refis and Leo White, review by Jacques Garrigue)
<br>
<br>
* #9011: Do not create .a/.lib files when creating a .cmxa with no
modules.
<br>
macOS ar doesn't support creating empty .a files (#1094) and
MSVC doesn't
<br>
permit .lib files to contain no objects. When linking with a
.cmxa containing
<br>
no modules, it is now not an error for there to be no .a/.lib
file.
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #9560: Report partial application warnings on type errors in
applications.
<br>
(Stephen Dolan, report and testcase by whitequark, review by
Gabriel Scherer
<br>
and Thomas Refis)
<br>
<br>
- #9583: when bytecode linking fails due to an unavailable module,
the module
<br>
that requires it is now included in the error message.
<br>
(Nicolás Ojeda Bär, review by Vincent Laviron)
<br>
<br>
- #9615: Attach package type attributes to core_type.
<br>
When parsing constraints on a first class module, attributes
found after the
<br>
module type were parsed but ignored. Now they are attached to
the
<br>
corresponding core_type.
<br>
(Etienne Millon, review by Thomas Refis)
<br>
<br>
- #6633, #9673: Add hint when a module is used instead of a module
type or
<br>
when a module type is used instead of a module or when a class
type is used
<br>
instead of a class.
<br>
(Xavier Van de Woestyne, report by whitequark, review by Florian
Angeletti
<br>
and Gabriel Scherer)
<br>
<br>
- #9754: allow [@tailcall true] (equivalent to [@tailcall]) and
<br>
[@tailcall false] (warns if on a tailcall)
<br>
(Gabriel Scherer, review by Nicolás Ojeda Bär)
<br>
<br>
- #9751: Add warning 68. Pattern-matching depending on mutable
state
<br>
prevents the remaining arguments from being uncurried.
<br>
(Hugo Heuzard, review by Leo White)
<br>
<br>
- #9783: Widen warning 16 (Unerasable optional argument) to more
cases.
<br>
(Leo White, review by Florian Angeletti)
<br>
<br>
- #10008: Improve error message for aliases to the current
compilation unit.
<br>
(Leo White, review by Gabriel Scherer)
<br>
<br>
- #10046: Link all DLLs with -static-libgcc on mingw32 to prevent
dependency
<br>
on libgcc_s_sjlj-1.dll with mingw-w64 runtime 8.0.0 (previously
this was
<br>
only needed for dllunix.dll).
<br>
(David Allsopp, report by Andreas Hauptmann, review by Xavier
Leroy)
<br>
<br>
- #9634: Allow initial and repeated commas in `OCAMLRUNPARAM`.
<br>
(Nicolás Ojeda Bär, review by Gabriel Scherer)
<br>
<br>
### Internal/compiler-libs changes:
<br>
<br>
- #8987: Make some locations more accurate
<br>
(Thomas Refis, review by Gabriel Scherer)
<br>
<br>
- #9216: add Lambda.duplicate which refreshes bound identifiers
<br>
(Gabriel Scherer, review by Pierre Chambart and Vincent Laviron)
<br>
<br>
- #9376: Remove spurious Ptop_defs from #use
<br>
(Leo White, review by Damien Doligez)
<br>
<br>
- #9604: refactoring of the ocamltest codebase.
<br>
(Nicolás Ojeda Bär, review by Gabriel Scherer and Sébastien
Hinderer)
<br>
<br>
- #9498, #9511: make the pattern-matching analyzer more robust to
<br>
or-pattern explosion, by stopping after the first
counter-example to
<br>
exhaustivity
<br>
(Gabriel Scherer, review by Luc Maranget, Thomas Refis and
Florian Angeletti,
<br>
report by Alex Fedoseev through Hongbo Zhang)
<br>
<br>
- #9514: optimize pattern-matching exhaustivity analysis in the
single-row case
<br>
(Gabriel Scherer, review by Stephen DOlan)
<br>
<br>
- #9442: refactor the implementation of the [@tailcall] attribute
<br>
to allow for a structured attribute payload
<br>
(Gabriel Scherer, review by Vladimir Keleshev and Nicolás Ojeda
Bär)
<br>
<br>
- #9688: Expose the main entrypoint in compilerlibs
<br>
(Stephen Dolan, review by Nicolás Ojeda Bär, Greta Yorsh and
David Allsopp)
<br>
<br>
- #9715: recheck scope escapes after normalising paths
<br>
(Matthew Ryan, review by Gabriel Scherer and Thomas Refis)
<br>
<br>
- #9778: Fix printing for bindings where polymorphic type
annotations and
<br>
attributes are present.
<br>
(Matthew Ryan, review by Nicolás Ojeda Bär)
<br>
<br>
- #9797, #9849: Eliminate the routine use of external commands in
ocamltest.
<br>
ocamltest no longer calls the mkdir, rm and ln external commands
(at present,
<br>
the only external command ocamltest uses is diff).
<br>
(David Allsopp, review by Nicolás Ojeda Bär, Sébastien Hinderer
and
<br>
Xavier Leroy)
<br>
<br>
- #9801: Don't ignore EOL-at-EOF differences in ocamltest.
<br>
(David Allsopp, review by Damien Doligez, much input and thought
from
<br>
Daniel Bünzli, Damien Doligez, Sébastien Hinderer, and Xavier
Leroy)
<br>
<br>
- #9889: more caching when printing types with -short-path.
<br>
(Florian Angeletti, review by Gabriel Scherer)
<br>
<br>
- #9591: fix pprint of polyvariants that start with a core_type,
closed,
<br>
not low (Chet Murthy, review by Florian Angeletti)
<br>
<br>
- #9590: fix pprint of extension constructors (and exceptions)
that rebind
<br>
(Chet Murthy, review by octachron@)
<br>
<br>
- #9963: Centralized tracking of frontend's global state
<br>
(Frédéric Bour and Thomas Refis, review by Gabriel Scherer)
<br>
<br>
- #9631: Named text sections for caml_system__code_begin/end
symbols
<br>
(Greta Yorsh, review by Frédéric Bour)
<br>
<br>
- #9896: Share the strings representing scopes, fixing some
regression
<br>
on .cmo/.cma sizes
<br>
(Alain Frisch and Xavier Clerc, review by Gabriel Scherer)
<br>
<br>
### Build system:
<br>
<br>
- #9332, #9518, #9529: Cease storing C dependencies in the
codebase. C
<br>
dependencies are generated on-the-fly in development mode. For
incremental
<br>
compilation, the MSVC ports require GCC to be present.
<br>
(David Allsopp, review by Sébastien Hinderer, YAML-fu by Stephen
Dolan)
<br>
<br>
- #7121, #9558: Always have the autoconf-discovered ld in PACKLD,
with
<br>
extra flags in new variable PACKLD_FLAGS. For
<br>
cross-compilation, this means the triplet-prefixed version will
always be
<br>
used.
<br>
(David Allsopp, report by Adrian Nader, review by Sébastien
Hinderer)
<br>
<br>
- #9527: stop including configuration when running 'clean' rules
<br>
to avoid C dependency recomputation.
<br>
(Gabriel Scherer, review by David Allsopp)
<br>
<br>
- #9804: Build C stubs of libraries in otherlibs/ with debug info.
<br>
(Stephen Dolan, review by Sébastien Hinderer and David Allsopp)
<br>
<br>
- #9938, #9939: Define __USE_MINGW_ANSI_STDIO=0 for the mingw-w64
ports to
<br>
prevent their C99-compliant snprintf conflicting with ours.
<br>
(David Allsopp, report by Michael Soegtrop, review by Xavier
Leroy)
<br>
<br>
- #9895, #9523: Avoid conflict with C++20 by not installing
VERSION to the OCaml
<br>
Standard Library directory.
<br>
(Bernhard Schommer, review by David Allsopp)
<br>
<br>
- #10044: Always report the detected ARCH, MODEL and SYSTEM, even
for bytecode-
<br>
only builds (fixes a "configuration regression" from 4.08 for
the Windows
<br>
builds)
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #10071: Fix bug in tests/misc/weaklifetime.ml that was reported
in #10055
<br>
(Damien Doligez and Gabriel Scherer, report by David Allsopp)
<br>
<br>
### Bug fixes:
<br>
<br>
- #7538, #9669: Check for misplaced attributes on module aliases
<br>
(Leo White, report by Thomas Leonard, review by Florian
Angeletti)
<br>
<br>
- #7813, #9955: make sure the major GC cycle doesn't get stuck in
Idle state
<br>
(Damien Doligez, report by Anders Fugmann, review by
Jacques-Henri Jourdan)
<br>
<br>
- #7902, #9556: Type-checker infers recursive type, even though
-rectypes is
<br>
off.
<br>
(Jacques Garrigue, report by Francois Pottier, review by Leo
White)
<br>
<br>
- #8746: Hashtbl: Restore ongoing traversal status after
filter_map_inplace
<br>
(Mehdi Bouaziz, review by Alain Frisch)
<br>
<br>
- #8747, #9709: incorrect principality warning on functional
updates of records
<br>
(Jacques Garrigue, report and review by Thomas Refis)
<br>
<br>
* #8907, #9878: `Typemod.normalize_signature` uses wrong
environment
<br>
(Jacques Garrigue, report and review by Leo White)
<br>
<br>
- #9421, #9427: fix printing of (::) in ocamldoc
<br>
(Florian Angeletti, report by Yawar Amin, review by Damien
Doligez)
<br>
<br>
- #9440: for a type extension constructor with parameterised
arguments,
<br>
REPL displayed <poly> for each as opposed to the concrete
values used.
<br>
(Christian Quinn, review by Gabriel Scherer)
<br>
<br>
- #9433: Fix package constraints for module aliases
<br>
(Leo White, review by Jacques Garrigue)
<br>
<br>
- #9469: Better backtraces for lazy values
<br>
(Leo White, review by Nicolás Ojeda Bär)
<br>
<br>
- #9521, #9522: correctly fail when comparing functions
<br>
with Closure and Infix tags.
<br>
(Gabriel Scherer and Jeremy Yallop and Xavier Leroy,
<br>
report by Twitter user @st_toHKR through Jun Furuse)
<br>
<br>
- #9611: maintain order of load path entries in various
situations: when passing
<br>
them to system linker, ppx contexts, etc.
<br>
(Nicolás Ojeda Bär, review by Jérémie Dimino and Gabriel
Scherer)
<br>
<br>
- #9633: ocamltest: fix a bug when certain variables set in test
scripts would
<br>
be ignored (eg `ocamlrunparam`).
<br>
(Nicolás Ojeda Bär, review by Sébastien Hinderer)
<br>
<br>
- #9681, #9690, #9693: small runtime changes
<br>
for the new closure representation (#9619)
<br>
(Xavier Leroy, Sadiq Jaffer, Gabriel Scherer,
<br>
review by Xavier Leroy and Jacques-Henri Jourdan)
<br>
<br>
- #9739, #9747: Avoid calling type variables, types that are not
variables in
<br>
recursive occurrence error messages
<br>
(for instance, "Type variable int occurs inside int list")
<br>
(Florian Angeletti, report by Stephen Dolan, review by Armaël
Guéneau)
<br>
<br>
- #9759, #9767: Spurious GADT ambiguity without -principal
<br>
(Jacques Garrigue, report by Thomas Refis,
<br>
review by Thomas Refis and Gabriel Scherer)
<br>
<br>
- #9799, #9803: make pat_env point to the correct environment
<br>
(Thomas Refis, report by Alex Fedoseev, review by Gabriel
Scherer)
<br>
<br>
- #9825, #9830: the C global variable caml_fl_merge and the C
function
<br>
caml_spacetime_my_profinfo (bytecode version) were declared and
<br>
defined with different types. This is undefined behavior and
<br>
cancause link-time errors with link-time optimization (LTO).
<br>
(Xavier Leroy, report by Richard Jones, review by Nicolás Ojeda
Bär)
<br>
<br>
- #9753: fix build for Android
<br>
(Eduardo Rafael, review by Xavier Leroy)
<br>
<br>
- #9848, #9855: Fix double free of bytecode in toplevel
<br>
(Stephen Dolan, report by Sampsa Kiiskinen, review by Gabriel
Scherer)
<br>
<br>
- #9858, #9861: Compiler fails with Ctype.Nondep_cannot_erase
exception
<br>
(Thomas Refis, report by Philippe Veber, review by Florian
Angeletti)
<br>
<br>
- #9860: wrong range constraint for subtract immediate on zSystems
/ s390x
<br>
(Xavier Leroy, review by Stephen Dolan)
<br>
<br>
- #9868, #9872, #9892: bugs in {in,out}_channel_length and seek_in
<br>
for files opened in text mode under Windows
<br>
(Xavier Leroy, report by Alain Frisch, review by Nicolás Ojeda
Bär
<br>
and Alain Frisch)
<br>
<br>
- #9925: Correct passing -fdebug-prefix-map to flexlink on Cygwin
by prefixing
<br>
it with -link.
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #9927: Restore Cygwin64 support.
<br>
(David Allsopp, review by Xavier Leroy)
<br>
<br>
- #9940: Fix unboxing of allocated constants from other
compilation units
<br>
(Vincent Laviron, report by Stephen Dolan, review by Xavier
Leroy and
<br>
Stephen Dolan)
<br>
<br>
- #9991: Fix reproducibility for `-no-alias-deps`
<br>
(Leo White, review by Gabriel Scherer and Florian Angeletti)
<br>
<br>
- #9998: Use Sys.opaque_identity in CamlinternalLazy.force
<br>
This removes extra warning 59 messages when compiling
afl-instrumented
<br>
code with flambda -O3.
<br>
(Vincent Laviron, report by Louis Gesbert, review by Gabriel
Scherer and
<br>
Pierre Chambart)
<br>
<br>
- #9999: fix -dsource printing of the pattern (`A as x | (`B as
x)).
<br>
(Gabriel Scherer, report by Anton Bachin, review by Florian
Angeletti)
<br>
<br>
- #9970, #10010: fix the declaration scope of extensible-datatype
constructors.
<br>
A regression that dates back to 4.08 makes extensible-datatype
constructors
<br>
with inline records very fragile, for example:
<br>
type 'a t += X of {x : 'a}
<br>
(Gabriel Scherer, review by Thomas Refis and Leo White,
<br>
report by Nicolás Ojeda Bär)
<br>
<br>
- #10048: Fix bug with generalized local opens.
<br>
(Leo White, review by Thomas Refis)
<br>
<br>
- #10106, #10112: some expected-type explanations were forgotten
<br>
after some let-bindings
<br>
(Gabriel Scherer, review by Thomas Refis and Florian Angeletti,
<br>
report by Daniil Baturin)
<br>
<br>
<br>
<br>
<br>
</div>
</body>
</html>
--------------43F7E16313203F4ED826936E--