Dear OCaml users,

We have the pleasu= re of announcing the release of OCaml 5.2.1, dedicated to the
memory of = Niels Bohr and Paul =C9luard on the anniversary of their deaths.

OCa= ml 5.2.1 is a collection of safe but import runtime time bug fixes backport= ed
from the 5.3 branch of OCaml to improve the stability of the 5.2 runt= ime while waiting for the
upcoming release of OCaml 5.3.0.

The fu= ll list of bug fixes is available below for more details.

Happy hack= ing,
-- Florian Angeletti, for the OCaml team.

Installation Instr= uctions

-------------------------------

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

&nbs= p; opam update
opam switch create 5.2.1
The source code for the release is also directly available on:
* GitHub: https://github.com/ocaml/ocaml/archive/5.2.1.tar.gz)
* Inria = archive: https://caml.inria.fr/pub/distrib/ocaml-5.2/ocaml-5.2.1.tar.gz
=

Bug Fixes In OCaml 5.2.1 (18 November 2024)

---------= ----------------------------------------------------

- #13207(https://github.com/ocaml/ocaml/issu= es/13207): Be sure to reload the register caching the exception handler in<= br>    `caml_c_call` and `caml_c_call_stack_args`, as its va= lue may have been changed
    if the OCaml stack is expan= ded during a callback.
    (Miod Vallat, report by Vesa K= arvonen, review by Gabriel Scherer and
     Xavier L= eroy)

- #13252(https://github.com/ocaml/ocaml/issues/1= 3252): Rework register assignment in the interpreter code on m68k on Linux,=
    due to the %a5 register being used by GLIBC.
&nbs= p;   (Miod Vallat, report by St=E9phane Glondu, review by Gabriel= Scherer and
     Xavier Leroy)

= - #13268(https://github.com/ocaml/ocaml/issues/13268): Fix a call to test = in `configure.ac` that was causing errors when
    LDFLAG= S contains several words.
    (St=E9phane Glondu, review = by Miod Vallat)

- #13234(https://github.com/ocaml/ocam= l/issues/13234), #13267(https://github.com/ocaml/ocaml/issues/13267): Open = runtime events file in read-write mode on ARMel
    (ARMv= 5) systems due to atomic operations limitations on that
  &nbs= p; platform.
    (St=E9phane Glondu, review by Miod Valla= t and Vincent Laviron)

- #13188(https://github.com/oca= ml/ocaml/issues/13188): fix races in the FFI code coming from the use of `I= nt_val(...)`
    on rooted values inside blocking questio= ns / without the runtime lock.
    (Calling `Int_val(...)= ` on non-rooted immediates is fine, but any
     acc= ess to rooted values must be done outside blocking sections /
&nbs= p;   with the runtime lock.)
    (Etienne Millo= n, review by Gabriel Scherer, Jan Midtgaard, Olivier Nicole)

= - #13318(https://github.com/ocaml/ocaml/issues/13318): Fix regressio= n in GC alarms, and fix them for Flambda.
    (Guillaume = Munch-Maccagnoni, report by Benjamin Monate, review by
   = ; Vincent Laviron and Gabriel Scherer)

- #13140(= https://github.com/ocaml/ocaml/issues/13140): POWER back-end: fix issue wit= h call to `caml_call_realloc_stack`
    from a DLL
&nb= sp;   (Xavier Leroy, review by Miod Vallat)

= - #13370]https://github.com/ocaml/ocaml/issues/13370): Fix a low-probabilit= y crash when calling `Gc.counters`.
    (Demi Marie Obeno= ur, review by Gabriel Scherer)

- #13402(https://github= .com/ocaml/ocaml/issues/13402), #13512(https://github.com/ocaml/ocaml/issue= s/13512), #13549(https://github.com/ocaml/ocaml/issues/13549), #13553(https= ://github.com/ocaml/ocaml/issues/13553): Revise bytecode implementation of = callbacks
    so that it no longer produces dangling regi= stered bytecode fragments.
    (Xavier Leroy, report by J= an Midtgaard, analysis by Stephen Dolan,
     review= by Miod Vallat)

- #13502(https://github.com/ocaml/oca= ml/issues/13502): Fix misindexing related to `Gc.finalise_last` that could = prevent
    finalisers from being run.
  &nb= sp; (Nick Roberts, review by Mark Shinwell)

- #13520(h= ttps://github.com/ocaml/ocaml/issues/13520): Fix compilation of native-code= version of systhreads. Bytecode fields
    were being in= cluded in the thread descriptors.
    (David Allsopp, rev= iew by S=E9bastien Hinderer and Miod Vallat)