Identify mismatched class type parameters and class parameters by ordinal in error messages.

Compiler's ocaml-variants.opam always includes the VERSION number, so
there's no need to edit it further.

* Building the compiler with ThreadSanitizer and running the testsuite caused too
many reports in OCaml 5 and was disabled (see ocaml#11040).

Since then, the work on TSan support for OCaml programs has led to fix a number
of those data races and temporarily silence the ones that are waiting to be
investigated (see ocaml#11040 again). As a result, running the testsuite with
`--enable-tsan` is now a cheap and effective way of detecting new data races
that may be introduced in the runtime.

A second good reason to restore the TSan CI is that it will detect early if a
recent change has accidentally broken TSan instrumentation (as has happened
before as an accidental consequence of removing a symbol
ocaml#12383 (review)), or
other issues (e.g. a new test revealed a TSan limitation with signals
ocaml#12561 (comment)).

Adding this test to the Github Actions CI arguably lengthens the runs (a GHA
run on amd64 Linux with TSan takes about 50 minutes). This PR therefore
suggests the compromise of enabling it on the Inria CI which is run on every
merge.

* Disable tests parallel/catch_break with tsan

* CI sanitizers: Use clang 14

clang 13 thread sanitizer produces different, less precise traces.
Also, clang 14 is the default version in Ubuntu 22.04 LTS.

---------

Co-authored-by: Xavier Leroy <xavier.leroy@college-de-france.fr>

Simplify `opam pin` command in `HACKING.adoc`

Silence ocamldoc warning on iter2

…rity-in-simplif

Stop `Simplif` from causing disagreement between runtime arity and syntatic arity of a function

parsing: Attach a location to the RHS of Ptyp_alias

Merge compilerlibs/Makefile.compilerlibs into the root Makefile

This is captured by the generic framework and does thus not need
to be written here.

(Follow-up to PR ocaml#12321 merging ocamltest/Makefile into the root Makefile)

This variable was used in ocamldoc/Makefile, by the test targets that got
removed in PR ocaml#12615.

This commit introduces two private build variables: OCAMLDOC_TARGET
and OCAMLDOC_OPT_TARGET.

The ./ prefix in front of $(OCAMLDOC) and $(OCAMLDOC_OPT) is not useful
since the definitions of these variables are already prefixed with
$(ROOTDIR).

Makefile.best_ocamldoc was included in two files:

1. In ocamldoc/Makefile, which also includes

  $(ROOTDIR)/Makefile.best_binaries

So that the definitions which were in Makefile.best_ocamldoc remain available.

2. In api_docgen/ocamldoc/Makefile which also includes
Makefile.best_binaries via api_docgen/Makefile.common

STDLIB_MANPAGES is still defined in Makefile.config.in for backwards
compatibility, whereas build_libraries_manpages is defined in
Maekfile.build_config.in and thus remains private.

Co-authored-by: David Allsopp <david.allsopp@metastack.com>

Before this commit, the build_libraries_manpages variable (aka STDLIB_MANPAGES)
was set to true if the build of the manpages for libraries was
enabled at configure time, whether ocamldoc (which is required
 to build those manpages) was enabled or not. It was thus the build
system's responsibility to determine whether the manpages for
libraries should be built / installed,
by testing both the
build_ocamldoc and the build_libraries_manpages variables.

However, it is known at configure time whether ocamldoc is available
or not, which makes it possible to set build_libraries_manpages to
true only if the manpages for libraries have been requested AND
ocamldoc has been enabled.

This is what is done in this commit, leading to a simplification
on the build system's side since it becomes enough to test only
one variable, namely build_libraries_manpages, rather than two
like before.

Co-authored-by: David Allsopp <david.allsopp@metastack.com>

Merge ocamldoc/Makefile into the root Makefile

The data race shows up in TSan-enabled runs, and can be removed for a
small cost.

… behavior

…ater back to work units using a different heap size. When the heap is growing, this unduly accelerates the major GC, slowing down the program.

I think that it is natural for users to expect that 'blit' can be used
on ranges that go past the current end of the array. We could warn
about it in the documentation and have a clear error, or we could
decide to support this use-case.

I suffer from decision fatigue on this question, and I think that it
is better to remove this rarely-used function.

just use Dynarray.init

…ation

… most cases

…ay_length

Co-authored-by: Daniel Bünzli <daniel.buenzli@erratique.ch>

Reviewed-by: Clément Allain <clef-men@orange.fr>

Suggested-by: Daniel Bünzli <daniel.buenzli@erratique.ch>
Suggested-by: Damien Doligez <damien.doligez@inria.fr>

Suggested-by: Guillaume Munch-Maccagnoni <Guillaume.Munch-Maccagnoni@inria.fr>
Suggested-by: Simon Cruanes <simon.cruanes.2007@m4x.org>

Dynarrays, boxed

Annotate Domain.self with noalloc

…sambiguation-of-optional-arg-defaults

Fix type directed disambiguation of optional arg defaults

Simplify calls to caml_int_compare (and similar functions) (was ocaml#1809)

Test:

    $ echo "let x = 1" > test.ml
    $ ocamlc -dparsetree -pp cat test.ml

Result before:

    [
      structure_item (/tmp/ocamlpp1775ad[1,0+0]..[1,0+9])
        Pstr_value Nonrec
        [
          <def>
            pattern (/tmp/ocamlpp1775ad[1,0+4]..[1,0+5])
              Ppat_var "x" (/tmp/ocamlpp1775ad[1,0+4]..[1,0+5])
            expression (/tmp/ocamlpp1775ad[1,0+8]..[1,0+9])
              Pexp_constant PConst_int (1,None)
        ]
    ]

Problem: the locations in the AST refer to the temporary
file created for preprocessing, which has since been removed.

Result after:

    [
      structure_item (test.ml[1,0+0]..[1,0+9])
        Pstr_value Nonrec
        [
          <def>
            pattern (test.ml[1,0+4]..[1,0+5])
              Ppat_var "x" (test.ml[1,0+4]..[1,0+5])
            expression (test.ml[1,0+8]..[1,0+9])
              Pexp_constant PConst_int (1,None)
        ]
    ]

Pre-C99 compilers used to support FAM as an extension, with a
different syntax and sizeof semantics.

This commit replaces the double quotes that start magic numbers by
{magic| and those that end them by |magic}.

Such quotes make it easier to automate the bumping process for magic numbers.