OCaml 5.3.0 released

octachron · January 8, 2025, 10:21pm

We have the pleasure of announcing the release of OCaml version 5.3.0. dedicated
to the memory of John William Mauchly and Paul Verlaine on the anniversary of
their death.

De la musique avant toute chose,
Et pour cela préfère l’Impair

(Music first and foremost of all!
Choose your measure of odd not even)

Some of the highlights in OCaml 5.3.0 are:

Syntax for deep effect handlers

There is now a dedicated syntax for installing deep effect handler
```
match f () with
| x -> x
| effect Random_float, k -> Effect.Deep.continue k (Random.float 1.0)
```
This new syntax adds a new effect keyword, which may break existing code.
To improve backward compatibility, this new keyword can be disabled with
the new -keywords flags if needed for backward compatibility.
Restored MSVC port

It is now possible to use the MSVC toolchain on Windows, restoring the last
missing port from OCaml 4 (except for the native compiler support for 32-bit
architectures which is not planned)
Re-introduced statistical memory profiling (statmemprof)

The submodule Gc.memprof is restored with a slightly different API. This
submodule can be used to monitor memory allocation statistics inside a
program. In OCaml 5, each domain can be monitored independently while child
domains inherit the parent domain profiling profile (if there is one active).
utf-8 encoded Unicode source files and modest support of Unicode identifiers
```
type saison = Hiver | Été | Printemps | Automne 
```
The OCaml lexer has been extended to support a modest subset of Unicode
characters in identifiers. This is mostly intended for pedagogical use. This
extended support requires source files to be utf-8 encoded Unicode text.
More space-efficient implementation of Dynarray

The internal implementation of Dynarray now uses an unboxed representation
which avoids the need of storing items wrapped in a Some x block and thus
save some spaces and indirections.
Improved metadata on the pairs of declarations and definitions for merlin.

The metadata stored inside cmt files has been improved to better distinguish the provenance of identifiers (previous versions could confuse an interface and implementation identifier).
Similarly, the metadata now tracks more precisely the association between declarations and definitions. For instance, in
```
module X = struct let x = 0 end
module M: sig
  val x: int
end = struct
  let x = 1
  include X
end
```
Merlin can now determine that the definition of the M.x value lies inside the module X.

And a lot of incremental changes:

Around 20 new functions in the standard library (in the Domain, Dynarray
Format, List, Queue, Sys, and Uchar modules).
Many fixes and improvements in the runtime
Improved error messages for first-class modules, functors, labelled arguments,
and type clashes.
Numerous bug fixes

Please report any unexpected behaviours on the OCaml issue tracker.

The full list of changes can be found in the changelog below.

Happy hacking,
Florian Angeletti for the OCaml team.

Installation Instructions

The base compiler can be installed as an opam switch with the following commands:

opam update
opam switch create 5.3.0

The source code for the release is also directly available on:

Fine-Tuned Compiler Configuration

If you want to tweak the configuration of the compiler, you can switch to the option variant with:

opam update
opam switch create <switch_name> ocaml-variants.5.3.0+options <option_list>

where <option_list> is a space separated list of ocaml-option-* packages. For instance, for a flambda and no-flat-float-array switch:

opam switch create 5.3.0+flambda+nffa ocaml-variants.5.3.0+options ocaml-option-flambda ocaml-option-no-flat-float-array

Changelog

Restored backend:

#12954: Restore the MSVC port
(David Allsopp, Antonin Décimo, Samuel Hym, and Miod Vallat, review by Nicolás
Ojeda Bär)
#13093: Allow building the MSVC port with clang-cl.
(Antonin Décimo, review by Nicolás Ojeda Bär, Samuel Hym,
David Allsopp and Sébastien Hinderer)

Language features:

#12309, #13158: Add syntax support for deep effect handlers
(Leo White, Tom Kelly, Anil Madhavapeddy, KC Sivaramakrishnan, Xavier Leroy
and Florian Angeletti, review by the same, Hugo Heuzard, and Ulysse Gérard)
#11736, #12664, #13628: Support utf-8 encoded source files and latin-9
compatible identifiers.
(Xavier Leroy and Florian Angeletti, review by Daniel Bünzli and
Jules Aguillon)
#12828, #13283: Add short syntax for dependent functor types (X:A) -> ...
(Jeremy Yallop, review by Nicolás Ojeda Bär and Gabriel Scherer)

Type system

#11891, #12507: Allow to name new locally abstract types in constructor type
annotations.
(Jacques Garrigue, report and review by Gabriel Scherer and Florian Angeletti)

Runtime system:

#11911, #12923: Multicore statistical memory profiling.
This restores a notable OCaml 4 feature that was missing
in OCaml 5.
(Nick Barnes, review by Stephen Dolan, Jacques-Henri Jourdan
and Guillaume Munch-Maccagnoni).
#13419: Fix memory bugs in runtime events system.
(B. Szilvasy and Nick Barnes, review by Miod Vallat, Nick Barnes,
Tim McGilchrist, and Gabriel Scherer)
#13364: Emit major slice counters in the runtime events.
(KC Sivaramakrishnan and Sadiq Jaffer, review by Gabriel Scherer)
#13382: Add more documentation for Runtime_events types
(Sadiq Jaffer, review by Tim McGilchrist, Miod Vallat and KC Sivaramakrishnan)
#13370: Fix a low-probability crash when calling Gc.counters.
(Demi Marie Obenour, review by Gabriel Scherer)
#13272: Allow maximum number of domains to be specified as a OCAMLRUNPARAM
parameter.
(KC Sivaramakrishnan, review by Guillaume Munch-Maccagnoni, Miod Vallat,
Gabriel Scherer, David Allsopp, request by Zachary Yedidia).
#12579: OS-based Synchronisation for Stop-the-World Sections
(B. Szilvasy, review by Miod Vallat, Nick Barnes, Olivier Nicole,
Gabriel Scherer and Damien Doligez)
#12789: Restore caml_unregister_frametable from OCaml 4
(Frédéric Recoules, review by Gabriel Scherer)
#13003: new, more consistent names for array-creation C functions
(Gabriel Scherer, review by Olivier Nicole)
#13013: introduce a caml_result type to supersede the
use of ‘encoded exception values’ in the FFI.
(Gabriel Scherer, review by Damien Doligez,
Guillaume Munch-Maccagnoni and Xavier Leroy,
suggested by Guillaume Munch-Maccagnoni)
#12407, #13226: Resource-handling improvements: add
exception-returning variants for all exception-raising functions in
caml/fail.h, for the purpose of cleaning-up state and resources
before raising.
(Guillaume Munch-Maccagnoni, review by Damien Doligez, Xavier Leroy
and Gabriel Scherer)
#13086: Avoid spurious major GC slices.
(Damien Doligez, report by Stephen Dolan, review by Gabriel Scherer
and Stephen Dolan)
#11779, #13117: Improve logic for fiber stack alignment.
(Miod Vallat, report by Damien Doligez, review by Gabriel Scherer)
#12839: Remove ATOMIC_UINTNAT_INIT from camlatomic.h (as part of a larger
cleanup of camlatomic.h)
(David Allsopp, review by Antonin Décimo, Sébastien Hinderer, Samuel Hym,
Guillaume Munch-Maccagnoni and Miod Vallat)
#13163: Enable frame pointers on macOS x86_64
(Tim McGilchrist, review by Sébastien Hinderer and Fabrice Buoro)
#12951: Constify constructors and flags tables in C code (take 2).
Now these tables will go in the readonly segment, where they belong.
(Antonin Décimo, review by David Allsopp)
#10696: Introduce __has_attribute and __has_c_attributes in
<caml/misc.h> to test the support of specific atributes in C
code. Introduce fallthrough as a wrapper around the fallthrough
attribute.
(Antonin Décimo, review by Nicolás Ojeda Bär, Xavier Leroy, and
Gabriel Scherer)
#13083: Use macros from limits.h to avoid signed-integer wrap-around.
Introduce CAML_{U,}INTNAT_{MIN,MAX} macros to expose {u,}intnat limits.
(Antonin Décimo, review by Nick Barnes, Xavier Leroy, Gabriel Scherer,
and Miod Vallat)
#13239: Check whether the compiler supports the labels as values
extension to enable threaded code interpretation.
(Antonin Décimo, review by Miod Vallat)
#13238: Enable software prefetching on x86 and x86_64 when building
with MSVC or clang-cl.
(Antonin Décimo, review by Miod Vallat)
#13241, #13261, #13271: Add CFI_SIGNAL_FRAME to ARM64 and RiscV runtimes,
for the purpose of displaying backtraces correctly in GDB.
(Tim McGilchrist, review by Miod Vallat, Gabriel Scherer and
KC Sivaramakrishnan)
#13139: Simplify CAMLalign to always use C23/C++11 alignas or C11
_Alignas. Ensures that stat data is always aligned to the best
boundary.
(Antonin Décimo, review by Miod Vallat and Xavier Leroy)
#13280: Check for support of compiler attributes. Allows using
compiler attributes with clang-cl.
(Antonin Décimo, review by Miod Vallat)
#13243: Enable C compiler warnings internally when building with
clang-cl or MSVC. Provide fixes too.
(Antonin Décimo, review by Miod Vallat and Xavier Leroy)
#13242: Define and use unreachable and trap annotation, and clean-up some
runtime assertions.
(Antonin Décimo, review by Miod Vallat, Gabriel Scherer, and David Allsopp)
#13402, #13512, #13549, #13553: Revise bytecode implementation of callbacks
so that it no longer produces dangling registered bytecode fragments.
(Xavier Leroy, report by Jan Midtgaard, analysis by Stephen Dolan,
review by Miod Vallat)
#13407: Add Runtime_events.EV_EMPTY_MINOR
(Thomas Leonard)
#13522: Confirm runtime events ring is still active after callback.
(KC Sivaramakrishnan, review by Sadiq Jaffer and Miod Vallat)
#13529: Do not write to event ring after going out of stw participant set.
(KC Sivaramakrishnan, review by Sadiq Jaffer)

Code generation and optimizations:

#13014: Enable compile-time option -function-sections on all previously
unsupported native backends (POWER, riscv64 and s390x)
(Miod Vallat, review by Nicolás Ojeda Bär)
#7241, #12555, #13076, #13138, #13338, #13152, #13153, #13154:
fix a soundness bug in the pattern-matching compiler
when side-effects mutate the scrutinee during matching.
(Gabriel Scherer, review by Nick Roberts)
#13341: a warning when the pattern-matching compiler pessimizes code
because side-effects may mutate the scrutinee during
matching. (This warning is disabled by default, as this rarely
happens and its performance impact is typically not noticeable.)
(Gabriel Scherer, review by Nick Roberts, Florian Angeletti
and David Allsopp)
#13179: Fix evaluation of toplevel lets in classes containing
local opens
(Vincent Laviron, review by Hugo Heuzard, Nathanaëlle Courant
and Gabriel Scherer)
#13543: Remove some String-Bytes conversion from the stdlib to behave better
with js_of_ocaml
(Hugo Heuzard, review by Gabriel Scherer)

Standard library:

#12885: move Dynarray to an unboxed representation
(Gabriel Scherer, suggestions by Vincent Laviron,
review by Olivier Nicole and Simon Cruanes, Yann Leray, Alain Frisch)
#12884: Add Queue.drop
(Léo Andrès, review by Nicolás Ojeda Bär and Gabriel Scherer)
#13168: In Array.shuffle, clarify the code that validates the
result of the user-supplied function rand, and improve the
error message that is produced when this result is invalid.
(François Pottier, review by Florian Angeletti, Daniel Bünzli
and Gabriel Scherer)
#12133: Expose support for printing substrings in Format
(Florian Angeletti, review by Daniel Bünzli, Gabriel Scherer
and Nicolás Ojeda Bär)
#12869: Add List.take, List.drop, List.take_while and List.drop_while
(Kate Deplaix and Oscar Butler-Aldridge, review by Nicolás Ojeda Bär,
Craig Ferguson and Gabriel Scherer)
#13047: Add Sys.poll_actions to (only) run pending runtime actions.
(Nick Barnes, review by Gabriel Scherer, Guillaume Munch-Maccagnoni, and
Vincent Laviron)
#13144: Dynarray.{equal, compare}
(Gabriel Scherer,
review by Jeremy Yallop, Daniel Bünzli and Olivier Nicole,
request by Olivier Nicole)
#13171: expose Domain.self_index : unit -> int (a somewhat-dense
indexing of currently-running domains) for advanced use-cases of
domain-indexed concurrent data structures.
(Gabriel Scherer,
review by KC Sivaramakrishnan, Miod Vallat and Nicolás Ojeda Bär,
report by Vesa Karvonen)
#13197: Dynarray.blit, which allows to extend the destination
dynarray (0 <= dst_pos <= dst_length).
(Gabriel Scherer, report by Hazem Elmasry,
review by Olivier Nicole, Hazem Elmasry and Nicolás Ojeda Bär)

(breaking change) #13240: Add Uchar.seeded_hash, Change Uchar.hash implementation.
Previously, Uchar.hash was aliased to Uchar.to_int. If you need that behavior,
change your module instantiation from eg module HT = Hashtbl.Make(Uchar) to
```
  module HT = Hashtbl.Make(struct
    ...
    let hash = Uchar.to_int
  end)
```
If the current implementation is desired, and you have a hashtable module HT
(produced with the Make functor) in persistent storage, use HT.rebuild to
ensure it doesn’t break when reading from or writing to buckets.
(Hazem ElMasry, review by Gabriel Scherer and Nicolás Ojeda Bär)

#13318: Fix regression in GC alarms, and fix them for flambda.
(Guillaume Munch-Maccagnoni, report by Benjamin Monate, review by
Vincent Laviron and Gabriel Scherer)
#13296: Add mem, memq, find_opt, find_index, find_map and find_mapi
to Dynarray.
(Jake H, review by Gabriel Scherer and Florian Angeletti)

Other libraries:

#11996: release the dependency of dynlink on compilerlibs.
(Sébastien Hinderer and Stephen Dolan, review by Damien Doligez and
Hugo Heuzard)
#13326: Implement Unix.O_APPEND on windows.
(Romain Beauxis, review by Miod Vallat, Gabriel Scherer and Antonin Décimo)

Tools:

#11716: ocamllex: mismatched parentheses and curly brackets are now caught
by ocamllex, instead of causing invalid OCaml code to be generated.
(Demi Marie Obenour, review by Damien Doligez and Xavier Leroy)
#12904: Run the testsuite with ThreadSanitizer on a PR when label
run-thread-sanitizer is added
(Olivier Nicole, suggested by Sébastien Hinderer and David Allsopp, review by
Gabriel Scherer)

(breaking change) #13114: Support ocamldebug remote debugging over IPv6 on all
platforms, and over Unix domain sockets on Windows.
(Antonin Décimo, review by Gabriel Scherer and Miod Vallat)

#13136: Rewrite GDB extensions and macros in debugger-agnostic Python, and add
LLDB support for them.
(Nick Barnes, review by Tim McGilchrist and Gabriel Scherer)

Toplevel:

#12891: Improved styling for initial prompt
(Florian Angeletti, review by Gabriel Scherer)
#13053: Improved display of builtin types such as _ list when aliased.
(Samuel Vivien, review by Florian Angeletti)

Manual and documentation:

#13370: Document that that temporary variables holding GCd pointers must
not be live across a GC.
(Demi Marie Obenour)
#12298: Manual: emphasize that Bigarray.int refers to an OCaml integer,
which does not match the C int type.
(Edwin Török, review by Florian Angeletti)
#12868: Manual: simplify style colours of the post-processed manual and API
HTML pages, and fix the search button icon
(Yawar Amin, review by Simon Grondin, Gabriel Scherer, and Florian Angeletti)
#12949: document OCaml release cycles and version strings in
release-info/introduction.md.
(Florian Angeletti, review by Fabrice Buoro, Kate Deplaix, Damien Doligez, and
Gabriel Scherer)
#12976: Manual: use webman//*.html and
webman//api/ for OCaml.org HTML manual generation
(Shakthi Kannan, review by Hannes Mehnert, and Florian Angeletti)
#13045: Emphasize caution about behaviour of custom block finalizers.
(Nick Barnes)
#13216: document the new caml_result type in the FFI chapter of the manual.
(Gabriel Scherer, review by Miod Vallat, Daniel Bünzli, Nick Barnes,
Guillaume Munch-Maccagnoni and Antonin Décimo)
#13287: stdlib/sys.mli: Update documentation on Sys.opaque_identity
following #9412.
(Matt Walker, review by Guillaume Munch-Maccagnoni and Vincent Laviron)
#13295: Use syntax for deep effect handlers in the effect handlers manual
page.
(KC Sivaramakrishnan, review by Anil Madhavapeddy, Florian Angeletti and Miod
Vallat)
#13424: Fix Gc.quick_stat documentation to clarify that returned fields
live_words, live_blocks, free_words, and fragments are not zero.
(Jan Midtgaard, review by Damien Doligez and KC Sivaramakrishnan)
#13440: Update documentation of Gc.{control,get,set} to reflect fields
not currently supported on OCaml 5.
(Jan Midtgaard, review by Gabriel Scherer)
#13469, #13474, #13535: Document that [Hashtbl.create n] creates a hash table
with a default minimal size, even if [n] is very small or negative.
(Antonin Décimo, Nick Bares, report by Nikolaus Huber and Jan Midtgaard,
review by Florian Angeletti, Anil Madhavapeddy, Gabriel Scherer,
and Miod Vallat)
#13666: Rewrite parts of the example code around nested lists in Chapter 6
(Polymorphism and its limitations → Polymorphic recursion) giving the
“depth” function [in the non-polymorphically-recursive part of the example]
a much more sensible behavior; also fix a typo and some formatting.
(Frank Steffahn, review by Florian Angeletti)
#13668: Document the basic support for unicode identifiers and the switch to
UTF-8 encoded Unicode text for OCaml source file
(Florian Angeletti, review by Nicolás Ojeda Bär and Daniel Bünzli)

Compiler user-interface and warnings:

(breaking change) #12084, #13669, #13673: Check link order when creating archive and when using
ocamlopt.
(Hugo Heuzard, review by Stefan Muenzel and Sébastien Hinderer)

#12980: Explain type mismatch involving first-class modules by including
the module level error message
(Florian Angeletti, review by Vincent Laviron)
#12985, #12988: Better error messages for partially applied functors.
(Florian Angeletti, report by Arthur Wendling, review by Gabriel Scherer)
#13034, #13260: Better error messages for mismatched function labels
(Florian Angeletti, report by Daniel Bünzli, review by Gabriel Scherer and
Samuel Vivien)
#13051: Add a “Syntax error” to error messages for invalid package signatures.
(Samuel Vivien, review by Gabriel Scherer)
#13099: Fix erroneous loading of cmis for some module type errors.
(Nick Roberts, review by Florian Angeletti)
#13151, name conflicts explanation as a footnote
(Florian Angeletti, review by Gabriel Scherer)
#13228: Re-export Cmt2annot.{iterator,binary_part} which had become hidden
since #11288 and broke ocamlbrowser.
(David Allsopp, report by Jacques Garrigue, review by Sébastien Hinderer)
#13251: Register printer for errors in Emitaux
(Vincent Laviron, review by Miod Vallat and Florian Angeletti)
#13255: Re-enable warning 34 for unused locally abstract types
(Nick Roberts, review by Chris Casinghino and Florian Angeletti)
#12182: Improve the type clash error message.
For example, this message:
```
  This expression has type ...
```
is changed into:
```
  The constant "42" has type ...
```
(Jules Aguillon, review by Gabriel Scherer and Florian Angeletti)
#13471: add -keywords <version?+list> flag to define the list of keywords
recognized by the lexer, for instance -keywords 5.2 disable the effect
keyword.
(Florian Angeletti, review by Gabriel Scherer)

Internal/compiler-libs changes:

#13286: Distinguish unique identifiers Shape.Uid.t according to their
provenance: either an implementation or an interface.
(Ulysse Gérard, review by Florian Angeletti and Leo White)
#13308: keep track of relations between declaration in the cmt files. This is
useful information for external tools for navigation and analysis purposis.
(Ulysse Gérard, Florian Angeletti, review by Florian Angeletti and Gabriel
Scherer)
#11129, #11148: enforce that ppxs do not produce parsetrees with
an empty list of universally quantified type variables
(. int -> int instead of 'a . int -> int')
(Florian Angeletti, report by Simmo Saan, review by Gabriel Scherer)
#12534: document and refactor Matching.mk_failaction_pos
(Gabriel Scherer, review by Vincent Laviron and Nick Roberts)
#13076: change the handling of Match_failure exits in the pattern-matching
compiler, to prepare for a complete fix for #7241
(Gabriel Scherer, review by Thomas Refis and Nick Roberts)
#12896: Simplify the compilation of custom bytecode runtimes by explicitly
compiling the primitives file before calling the linker. Tidy-up both the
generating code and the output itself for C code being generated by the
bytecode linker in -custom and -output-* modes.
(David Allsopp, Antonin Décimo and Samuel Hym, review by Vincent Laviron)
#12932: Remove useless code in Typecore.type_label_exp (was a fix for #4862)
(Jacques Garrigue, review by Gabriel Scherer)
#12943: Make transient_expr.scope a bitfield, and use it to store marks.
Marks are automatically allocated, and removed when leaving their scope.
Falls back to using TransientTypeSet when marks are exhausted.
(Jacques Garrigue and Takafumi Saikawa, review by Basile Clément)
#12946: Make generalization automatic when leaving scope.
As a result, the Ctype.generalize* and Ctype.correct_levels functions
were removed. The latter is now called Ctype.duplicate_type.
(Jacques Garrigue and Takafumi Saikawa, review by Richard Eisenberg)
#12968: Attach location to constants in the parsetree
(Jules Aguillon, review by Gabriel Scherer)
#12959, #13055: Avoid an internal error on recursive module type inconsistency
(Florian Angeletti, review by Jacques Garrigue and Gabriel Scherer)
#13049: graphical debugging printer for types
(Florian Angeletti, review by Gabriel Scherer)
#13074, #13082, #13084: refactoring in the pattern-matching compiler
(Gabriel Scherer, review by Thomas Refis, Vincent Laviron and Nick Roberts)
#13067: rework volatile memory access rules under TSan to consider properly
aligned smaller-than-register read operations as atomic, which gets rid of
false positives on s390x
(Miod Vallat, review by Fabien Buoro)
#13162: Use quoted strings to clarify code being generated.
(Antonin Décimo, review by Miod Vallat and Gabriel Scherer)
#13015: Emit floating-point literals in .rodata section on ELF arm64
platforms (Linux, *BSD).
(Miod Vallat, review by Nicolás Ojeda Bär)
#13169, #13311: Introduce a document data type for compiler messages
rather than relying on Format.formatter -> unit closures.
(Florian Angeletti, review by Gabriel Scherer)
#13193: Remove the unused env_init field from class blocks
(Vincent Laviron, review by Jacques Garrigue)
#13257: integrate MetaOCaml in the Menhir grammar to ease MetaOCaml
maintenance. This is a purely internal change: there is no support
in the lexer, so no change to the surface OCaml grammar.
(Oleg Kiselyov, Gabriel Scherer and Florian Angeletti,
review by Jeremy Yallop)
#13289: Use C99 for loop to reduce the scope of the for loop iterator.
(Antonin Décimo, review by Miod Vallat and Gabriel Scherer)
#13336: compiler-libs, split the Printtyp in three to only keep
“user-friendly” functions in the Printtyp module.
(Florian Angeletti, review by Gabriel Scherer)
#13361: split runtime/array.c functions to consistently expose
uniform_array and floatarray versions, use floatarray versions
in Float.Array.
(Gabriel Scherer, review by Nicolás Ojeda Bär)
#13507: A small refactoring to [free_vars] to make it a bit faster
by not allocating a list when the list is not necessary.
(Richard Eisenberg, review by Jacques Garrigue)

Build system:

#12909: Reorganise how MKEXE_VIA_CC is built to make it correct for MSVC by
grouping all the linker flags at the end of the C compiler commandline
(David Allsopp and Samuel Hym, review by Nicolás Ojeda Bär)
#12992, #13009: Check that flexlink can be executed only when building in a
native windows environment.
(Romain Beauxis, review by David Allsopp and Sébastien Hinderer)
#12996: Only link with -lgcc_eh when available.
(Romain Beauxis, review by David Allsopp and Miod Vallat)

(breaking change) #13200: Do not use CFLAGS for linking.
(Sébastien Hinderer, review by Gabriel Scherer, Antonin Décimo,
Miod Vallat and Samuel Hym)

#13201, #13244: Fix and speedup builds with TSan.
(Sébastien Hinderer, review by Miod Vallat, Gabriel Scherer and
Olivier Nicole)

(breaking change) #12578, #12589, #13322, #13519: Use configured CFLAGS and CPPFLAGS only
during the build of the compiler itself. Do not use them when
compiling third-party C sources through the compiler. Flags for
compiling third-party C sources can still be specified at configure
time in the COMPILER_{BYTECODE,NATIVE}_{CFLAGS,CPPFLAGS}
configuration variables.
(Sébastien Hinderer, report by William Hu, review by David Allsopp)

#13285: continue the merge of the sub-makefiles into the root
Makefile started with #11243, #11248, #11268, #11420, #11675,
#12198, #12321, #12586, #12616, #12706 and #13048.
(Sébastien Hinderer, review by David Allsopp and Florian Angeletti)

Bug fixes:

#12854: Add a test in the regression suite that flags the bug #12825.
(Luc Maranget)
#12888: fix printing of uncaught exceptions in .cmo files passed on the
command-line of the toplevel.
(Nicolás Ojeda Bär, review by Florian Angeletti, report by Daniel Bünzli)
#12910, #12920: Fix an unsound interaction between first-class modules
and polymorphic records by saving and restoring univar_pairs.
(Stephen Dolan, review by Gabriel Scherer, report by Jeremy Yallop)
#12994: Remove un-used and unsafe caml_drop_continuation
(Tim McGilchrist, reviewed by Gabriel Scherer and Miod Vallat)
#12963: Restore caml_runtime_parameters implementation. This primitive allows
programs to query the runtime parameters supplied to an OCaml program.
Implementation missing since OCaml 5.0.
(Tim McGilchrist, reviewed by David Allsopp and Miod Vallat)
#13012: parsing: Fix dropped attributes after a - or +
The syntax -(1 [@foo]) was incorrectly parsed as -1.
(Jules Aguillon, reviewed by Gabriel Scherer, report by Gabriel Scherer)

(breaking change) #13070: On Windows, when configured with bootstrapped flexdll, don’t add
+flexdll to the search path when -nostdlib is specified (which then means
-L no longer gets passed to the system linker).
(David Allsopp, review by Florian Angeletti)

#13089: Fix bug in runtime_events library which could result in garbled
output under Windows.
(B. Szilvasy, review by Nicolás Ojeda Bär and Miod Vallat)
#13088: A few type-checker behaviors look at a type to see if there are
any labeled arguments in it. This sometimes required expansion, which
could, in obscure scenarios, result in superfluous type errors.
(Richard Eisenberg, review by Gabriel Scherer and Jacques Garrigue)
#13103: FreeBSD/amd64: properly annotate .o files with non-executable stack
notes (Konstantin Belousov, review by Nicolás Ojeda Bär)
#13150: improve a transitive-closure computation algorithm in the flambda
middle-end to avoid a compilation time blowup on Menhir-generated code
(Florian Weimer, review by Gabriel Scherer and Pierre Chambart,
report by Richard Jones)
#13166: Fix a MinGW/MSVC Sys.rename regression on renaming a parent directory
to an empty child directory.
(Jan Midtgaard, review by Antonin Décimo, Sébastien Hinderer, and
David Allsopp)
#13185, #13192: Reject type-level module aliases on functor parameter
inside signatures.
(Jacques Garrigue, report by Richard Eisenberg, review by Florian Angeletti)
#13170: Fix a bug that would result in some floating alerts [@@@alert ...]
incorrectly triggering Warning 53.
(Nicolás Ojeda Bär, review by Chris Casinghino and Florian Angeletti)
#13203: Do not issue warning 53 if the compiler is stopping before attributes
have been accurately marked.
(Chris Casinghino, review by Florian Angeletti)
#13207: Be sure to reload the register caching the exception handler in
caml_c_call and caml_c_call_stack_args, as its value may have been changed
if the OCaml stack is expanded during a callback.
(Miod Vallat, report by Vesa Karvonen, review by Gabriel Scherer and
Xavier Leroy)
#13209: Fix configure test that checks whether ar supports @FILE
arguments.
(Nicolás Ojeda Bär, report by Boris D.)
#13221: Compute more accurate instruction sizes for branch relocation on
POWER.
(Miod Vallat, review by Gabriel Scherer)
#13252: Rework register assignment in the interpreter code on m68k on Linux,
due to the %a5 register being used by Glibc.
(Miod Vallat, report by Stéphane Glondu, review by Gabriel Scherer and
Xavier Leroy)
#13247: Disable lib_unix/kill test for MacOS AMD64 with TSan, linking
to llvm bug report causing infinite signal loops.
(Tim McGilchrist, review by Olivier Nicole, Miod Vallat, Sébastien Hinderer
and Gabriel Scherer)
#13234, #13267: Open runtime events file in read-write mode on armel
(armv5) systems due to atomic operations limitations on that
platform.
(Stéphane Glondu, review by Miod Vallat and Vincent Laviron)
#13273: Fix a call to test in configure.ac that was causing errors when
LDFLAGS contains several words.
(Stéphane Glondu, review by Miod Vallat)
#13290: Fix uninitialized and out of bounds reads in runtime_events_consumer.c
(Edwin Török, review by Miod Vallat and Antonin Décimo)
#13306: An algorithm in the type-checker that checks two types for equality
could sometimes, in theory, return the wrong answer. This patch fixes the
oversight. No known program triggers the bug.
(Richard Eisenberg, review by Florian Angeletti)
#13400: Initialize th->signal_stack to avoid free of uninitialized data
if the user calls caml_c_thread_unregister on the main thread.
(Richard W.M. Jones, review by Guillaume Munch-Maccagnoni and
Gabriel Scherer)
#13140: POWER back-end: fix issue with call to caml_call_realloc_stack
from a DLL
(Xavier Leroy, review by Miod Vallat)
#13263, #13560: fix printing true and false in toplevel and error
messages (no more unexpected #true)
(Florian Angeletti, report by Samuel Vivien, review by Gabriel Scherer)
#13388, #13540: raises an error message (and not an internal compiler error)
when two local substitutions are incompatible (for instance module type S:=sig end type t:=(module S))
(Florian Angeletti, report by Nailen Matschke, review by Gabriel Scherer, and
Leo White)
#13408: Fix misplaced debug runtime assertion triggerable by a race
between domain exit and backup thread
(Miod Vallat and Gabriel Scherer, report by Jan Midtgaard)
#13417: Filename.quote_command: fix handling of forward slashes in program
path under Win32.
(Nicolás Ojeda Bär, review by David Allsopp and Damien Doligez)
#13501: Regression on mutually recursive types caused by #12180.
Resuscitate Typedecl.update_type.
(Jacques Garrigue and Takafumi Saikawa, review by Florian Angeletti, Richard
Eisenberg and Gabriel Scherer)
#13502: Fix misindexing related to Gc.finalise_last that could prevent
finalisers from being run.
(Nick Roberts, review by Mark Shinwell)
#13495, #13514: Fix typechecker crash while typing objects
(Jacques Garrigue, report by Nicolás Ojeda Bär, review by
Nicolas Ojeda Bär, Gabriel Scherer, Stephen Dolan, Florian Angeletti)
#13391, #13551: fix a printing bug with -dsource when using
raw literal inside a locally abstract type constraint
(i.e. let f: type \#for. ... )
(Florian Angeletti, report by Nick Roberts, review by Richard Eisenberg)
#13520: Fix compilation of native-code version of systhreads. Bytecode fields
were being included in the thread descriptors.
(David Allsopp, review by Sébastien Hinderer and Miod Vallat)
#13541, #13591: Fix headers for C++ inclusion.
(Antonin Décimo, review by Nick Barnes, report by Kate Deplaix)
#13579, #13583: Unsoundness involving non-injective types + gadts
(Jacques Garrigue, report by @v-gb,
review by Richard Eisenberg and Florian Angeletti)
#13598: Falsely triggered warning 56 [unreachable-case]
This was caused by unproper protection of the retyping function.
(Jacques Garrigue, report by Tõivo Leedjärv, review by Florian Angeletti)
#13603, #13604: fix source printing in the presence of the escaped raw
identifier \#mod.
(Florian Angeletti, report by Chris Casinghino, review by Gabriel Scherer)

dbuenzli · January 8, 2025, 11:24pm

Thank you for the release work!

This is good news. Do you happen to have any insight at how it remains important, from the library eco-system perspective, to support the 4.14.* line ? (I see the word of caution about 5.0 has not been removed from the README)

dra27 · January 9, 2025, 5:43am

The warning has been “softened” (Soften the 5.0 cautionary README note by dra27 · Pull Request #13394 · ocaml/ocaml · GitHub). My (personal) intuition is that supporting 4.14.* remains important for now, because there are users of large codebases which are still in the process of evaluating OCaml 5.x, but that nothing reported is unsolvable (so far ). I would hope that by this time next year the cautionary note is able to be removed entirely. Note that existing known issues are in OCaml’s issue tracker, in particular the Performance and gc labels.

As ever, if one has a system using OCaml 4.x which for whatever reason cannot be switched to OCaml 5.x (especially if this is owing to a performance degradation), please do file an issue or start a discussion here, even if the system in question is closed-source (it’s obviously somewhat harder to investigate those regressions, but it’s still immensely useful to know the problems exist).

semarie · January 9, 2025, 11:27am

I would like to signal that The OCaml system, release 5.3 page has several dead links (HTTP 404) for the href in the line “This manual is also available in PDF, plain text, as a bundle of HTML files, and as a bundle of Emacs Info files”.

Thanks for this release.

virgile · January 9, 2025, 12:01pm

Moreover, the links to the documentation of the stdlib’s module are incorrect. They refer to a non-existing libref directory, while the pages seem to be in the api directory (e.g. https://ocaml.org/manual/5.3/libref/Arg.html → https://ocaml.org/manual/5.3/api/Arg.html)

octachron · January 9, 2025, 12:13pm

This is being fixed, I somehow sent the stand-alone html version of the manual to ocaml.org rather than the ocaml.org version.

octachron · January 9, 2025, 4:17pm

The manual and standard library documentation has been fixed now, thanks for the report !

nojaf · January 10, 2025, 10:24am

So, after reading the section “Debugging Improvements” from Tarides | OCaml 5.3: Features and Fixes! would this mean that debugging in say Vscode with OCaml earlybird is magically better on Mac after updating to 5.3?

EmileTrotignon · January 10, 2025, 10:55am

No, earlybird uses ocamldebug, it debugs the bytecode (and because of that is not platform specific). The improvement are for lldb which debugs native code.

giltho · January 10, 2025, 12:40pm

Absolute banger of a release
Thank you

lambda_foo · January 12, 2025, 10:22pm

The debugging improvements are all for native code but apply equally to both LLDB and GDB, where that makes sense. At the moment my focus is on improving native debugging since I think it will have the largest impact. If anyone wants to help me please get in touch.

Improving bytecode debugging would be useful but I don’t have much information about how people are using it and whether it is more / less important than native debugging.

toots · January 13, 2025, 6:02am

While super supportive and stoked for the pace of improvement in the OCaml releases, I have to note that this Gc regression has been documented for a while: Regression with default GC settings between `4.14.2` and `5.1.1` · Issue #13123 · ocaml/ocaml · GitHub.

The last release of our software has reduced memory footprint at runtime so I think that we might be able to work around it but, so far, this issue is the reason we still recommend OCaml 4.14 for our production users.

bluddy · January 13, 2025, 6:26am

Native debugging is far more important in general. Because it’s still limited though, especially with regard to value inspection, people tend to see bytecode debugging as more useful, but bytecode debugging is also not available in some contexts.

nojaf · January 14, 2025, 5:43pm

Pardon my ignorance, but does anyone have any links on how to set this up with Mac and vscode?
Is that a thing for native debugging?

lambda_foo · January 14, 2025, 9:33pm

Such a thing does exist, I documented my Emacs setup Debugging OCaml with Emacs · Perpetually Curious Blog

The short version for VSCode is:

Install CodeLLDB and OCaml Platform
Setup a launch.json at the top level of the project (example in that blog post).
Profit!

At step 3 you’ll be able to debug a binary, step through it and get accurate backtraces. Unfortunately at that point you’ll run into the current limitations of native debugging on macOS. Setting breakpoints based on symbol names requires ocaml#13050 and setting breakpoints based on filename / line number requires some minimal amount of DWARF information to be emitted by the compiler (my current work).

adrien · February 9, 2025, 1:59pm

If I’ve understood correctly, 5.3 makes memtrace possible again. Is memtrace usable again right away or does it need changes to adapt to 5.3?

It’s possible my program is affected in the same way as liquidsoap; I’ve copied the GC settings that @toots changed there and testing now but it will take more time to test (memory usage happens after network queries so it’s not easy to speed up reproducing the situation).

kit-ty-kate · February 9, 2025, 2:14pm

memtrace needs changes to be usable again. You can follow the progress at janestreet/memtrace#22

lambda_foo · February 10, 2025, 2:43am

What @kit-ty-kate said

The changes to memtrace are minimal (mostly locking around writing to a file) but sufficient to profile various applications using OCaml 5 domains and effects. Please test it out and report any issues or improvements you would find useful.

There is lots of scope for improving both the memtrace library itself and providing new ways of visualising the data. In particular supporting the directed graph view that pprof uses or Tree-maps like firefox, which I’m starting to look at with the help of an intern at Tarides.

adrien · February 10, 2025, 9:44pm

Thanks @kit-ty-kate and @lambda_foo . I’ll give it a try this week. Treemaps also sound wonderful, can’t wait to see the result.

Topic		Replies	Views
OCaml 4.11.0 released Ecosystem	5	8342	September 25, 2020
OCaml 5.0.0 is out! Ecosystem	8	35229	January 10, 2023
OCaml 5.1.0 released Ecosystem announce	11	3158	September 18, 2023
First beta release for OCaml 5.3.0 Ecosystem	0	339	November 4, 2024
Multicore OCaml: October 2021 Community multicore , multicore-monthly	0	8967	November 16, 2021