OCaml Weekly News

Previous Week Up Next Week


Here is the latest OCaml Weekly News, for the week of August 18 to 25, 2020.

Table of Contents

Dune 2.7.0

Rudi Grinberg announced

On behalf of the dune team, I'm pleased to announce the release of dune 2.7.0. This release adds a couple of important features and many bug fixes. I'll just briefly summarize the two features, and let our improved documentation elaborate on the details.

The first one is first class support for instrumentation tools such as bisect_ppx and landmarks. This is one of the most requested dune features ever, and we're pleased to show you all what we've come up with

Special thanks to @stephanieyou & @nojb for implementing this feature. Many thanks to @aantron for tirelessly iterating on the end user experience with us.

While the features is looking quite good already, we consider it a "first take" on this subject. We welcome your experience reports and feature requests.

The second feature we're introducing this release is dune's cram testing framework. The cram framework is our secret weapon for making dune (relatively) bug free. I'll soon write up a dedicated post to give you a flavour of how it works. To those who can't wait, there's a [new section (https://dune.readthedocs.io/en/latest/tests.html#cram-tests) in the documentation that explains everything.

As always, the change log is replicated below for your convenience.

Happy Hacking.

2.7.0 (13/08/2020)

  • Write intermediate files in a .mdx folder for each mdx stanza to prevent the corresponding actions to be executed as part of the @all alias (#3659, @NathanReb)
  • Read Coq flags from env (#3547 , fixes #3486, @gares)
  • Allow bisect_ppx to be enabled/disabled via dune-workspace. (#3404, @stephanieyou)
  • Formatting of dune files is now done in the executing dune process instead of in a separate process. (#3536, @nojb)
  • Add a --debug-artifact-substution flag to help debug problem with version not being captured by dune-build-info (#3589, @jeremiedimino)
  • Allow the use of the context_name variable in the enabled_if fields of executable(s) and install stanzas. (#3568, fixes #3566, @voodoos)
  • Fix compatibility with OCaml 4.12.0 when compiling empty archives; no .a file is generated. (#3576, @dra27)
  • $ dune utop no longer tries to load optional libraries that are unavailable (#3612, fixes #3188, @anuragsoni)
  • Fix dune-build-info on 4.10.0+flambda (#3599, @emillon, @jeremiedimino).
  • Allow multiple libraries with inline_tests to be defined in the same directory (#3621, @rgrinberg)
  • Run exit hooks in jsoo separate compilation mode (#3626, fixes #3622, @rgrinberg)
  • Add (alias …), (mode …) fields to (copy_fields …) stanza (#3631, @nojb)
  • (copy_files …) now supports copying files from outside the workspace using absolute file names (#3639, @nojb)
  • Dune does not use ocamlc as an intermediary to call C compiler anymore. Configuration flags ocamlc_cflags and ocamlc_cppflags are always prepended to the compiler arguments. (#3565, fixes #3346, @voodoos)
  • Revert the build optimization in #2268. This optimization slows down building individual executables when they're part of an executables stanza group (#3644, @rgrinberg)
  • Use {dev} rather than {pinned} in the generated .opam file. (#3647, @kit-ty-kate)
  • Insert correct extension name when editing dune-project files. Previously, dune would just insert the stanza name. (#3649, fixes #3624, @rgrinberg)
  • Fix crash when evaluating an mdx stanza that depends on unavailable packages. (#3650, @CraigFe)
  • Fix typo in cache-check-probablity field in dune config files. This field now requires 2.7 as it wasn't usable before this version. (#3652, @edwintorok)
  • Add "odoc" {with-doc} to the dependencies in the generated .opam files. (#3667, @kit-ty-kate)
  • Do not allow user actions to capture dune's stdin (#3677, fixes #3672, @rgrinberg)
  • (subdir ...) stanzas can now appear in dune files used via (include ...). (#3676, @nojb)

Cram - Tests on Short Notice

Rudi Grinberg announced

Does it seem like writing unit tests is endlessly boring and time consuming? Does your test suite require constant attention and tweaking, and yet the stream of bugs never seems to end? I feel the same way. Moreover, if I would knew how much time I would spend writing unit tests, I most likely would have picked a different profession. There has to be a better way.

In this post, I'd like to share one better way for testing binaries. A way that allows you to add new test cases in seconds, avoid writing manual assertions, and makes it easy to write self documenting tests by non technical users. The catch? You must upgrade to dune 2.7 and enable the new cram extension:

$ cat dune-project
(lang dune 2.7)
(cram enable)

Now dune will treat every file and directory that ends with .t as a cram test.

Let's create a trivial test wc.t to test the word counting utility:

Test the behavior of wc.t. Any line that doesn't start with 2 spaces
is a comment (like this one).

Next, we'll create a sample file to feed to wc:

Note the two spaces before the command:
  $ cat >sample.txt <<EOF
  > a
  > b
  > c
  > EOF

The command above creates a file with 3 lines. Note the leading 2 spaces and the $ denoting a command. We also use the heredoc syntax to pipe multiple lines to cat.

We'll finally write a test that makes sure that wc works.

Count the lines:
  $ wc -l sample.txt

Note how we didn't mention the expected output anywhere. This is where secret sauce comes in. We just run the test with dune:

$ dune runtest
 |  $ wc -l sample.txt
+|         3 sample.txt

And dune is helpful enough to fill in the output for us after promoting:

$ dune promote
$ dune runtest # now the tests pass

If we modify the wc utility to give a different result, the test will now fail because the command produced a different output. This style of testing is called expectation (or snapshot) testing. Here this style is dressed up in a shell like syntax to give us the cram test.

Dune 2.7 offers full support for this style, and we recomend it to all users. Do cram tests scale? In the dune project, this is our main testing mechanism and we have over 200 cram tests in our test suite. We use this approach to test and document both new features and regressions. So far we've been very satisfied with this approach, and we're happy to share it with our users.

As usual, there's far too much to describe in a single blog post. The rest is thoroughly documented in our manual

I look forward to answer any questions you might have about cram.

Guillaume Bury asked and Craig Ferguson replied

What would be the canonical way to express as a cram test a run of a binary that is expected to fail (i.e. non zero exit code), while also comparing the output (e.g. for testing error reporting of a compiler-like program).

Cram has a [ .. ] syntax for expecting a particular exit code:

$ dune build cycle.exe
Error: Dependency cycle detected between the following libraries:
   "a" in _build/default
-> "b" in _build/default
-> "c" in _build/default
-> "a" in _build/default
-> required by library "c" in _build/default
-> required by executable cycle in dune:17

(Example taken from Dune here.)

If the error code isn't as expected, it's shown in the diff:

 |Cycle detection
 |  $ dune build cycle.exe
 |  Error: Dependency cycle detected between the following libraries:
 |     "a" in _build/default
 |  -> "b" in _build/default
 |  -> "c" in _build/default
 |  -> "a" in _build/default
 |  -> required by library "c" in _build/default
 |  -> required by executable cycle in dune:17
-|  [2]
+|  [1]

OCaml 4.11.0 released

octachron announced

I have the pleasure of announcing the release of OCaml version 4.11.0, dedicated to the memory of Blaise Pascal on the anniversary of his death.

Some of the highlights in this release are:

  • Statmemprof: a new statistical memory profiler
  • A new instrumented runtime that logs runtime statistics in a standard format
  • A native backend for the RISC-V architecture
  • Improved backtraces that refer to function names
  • Support for recursive and yet unboxed types
  • A quoted extension syntax for ppxs.
  • Many quality of life improvements
  • Many bug fixes.

The full list of change can be found in the changelog below:

OCaml 4.11.0 (19 August 2020)

(Changes that can break existing programs are marked with a "*breaking change" warning)

  • Runtime system:
    • #9096: Print function names in backtraces.
      Old output:
      > Called from file "foo.ml", line 16, characters 42-53

      New output:
      > Called from Foo.bar in file "foo.ml", line 16, characters 42-53

      (Stephen Dolan, review by Leo White and Mark Shinwell)

    • #9082: The instrumented runtime now records logs in the CTF format. A new API is available in the runtime to collect runtime statistics, replacing the previous instrumented runtime macros. Gc.eventlog_pause and Gc.eventlog_resume were added to allow user to control instrumentation in a running program. See the manual for more information on how to use this instrumentation mode. (Enguerrand Decorne and Stephen Dolan, with help and review from David Allsopp, Sébastien Hinderer, review by Anil Madhavapeddy, Nicolás Ojeda Bär, Shakthi Kannan, KC Sivaramakrishnan, Gabriel Scherer, Guillaume Munch-Maccagnoni, Damien Doligez, Leo White, Daniel Bünzli and Xavier Leroy)
    • #9230, #9362: Memprof support for native allocations. (Jacques-Henri Jourdan and Stephen Dolan, review by Gabriel Scherer)
    • #8920, #9238, #9239, #9254, #9458: New API for statistical memory profiling in Memprof.Gc. The new version does no longer use ephemerons and allows registering callbacks for promotion and deallocation of memory blocks. The new API no longer gives the block tags to the allocation callback. (Stephen Dolan and Jacques-Henri Jourdan, review by Damien Doligez and Gabriel Scherer)
    • #9353: Reimplement output_value and the Marshal.to_* functions using a hash table to detect sharing, instead of temporary in-place modifications. This is a prerequisite for Multicore OCaml. (Xavier Leroy and Basile Clément, review by Gabriel Scherer and Stephen Dolan)
    • #9119: Make [caml_stat_resize_noexc] compatible with the [realloc] API when the old block is NULL. (Jacques-Henri Jourdan, review by Xavier Leroy)
    • #9233: Restore the bytecode stack after an allocation. (Stephen Dolan, review by Gabriel Scherer and Jacques-Henri Jourdan)
    • #9249: restore definition of ARCH_ALIGN_INT64 in m.h if the architecture requires 64-bit integers to be double-word aligned (autoconf regression) (David Allsopp, review by Sébastien Hinderer)
    • #9259: Made Ephemeron.blit_key and Weak.blit faster. They are now linear in the size of the range being copied instead of depending on the total sizes of the ephemerons or weak arrays involved. (Arseniy Alekseyev, design advice by Leo White, review by François Bobot and Damien Doligez)
    • #9279: Memprof optimisation. (Stephen Dolan, review by Jacques-Henri Jourdan)
    • #9280: Micro-optimise allocations on amd64 to save a register. (Stephen Dolan, review by Xavier Leroy)
    • #9426: build the Mingw ports with higher levels of GCC optimization (Xavier Leroy, review by Sébastien Hinderer)
    • [ breaking change ] #9483: Remove accidental inclusion of <stdio.h> in <caml/misc.h> The only release with the inclusion of stdio.h has been 4.10.0 (Christopher Zimmermann, review by Xavier Leroy and David Allsopp)
    • #9282: Make Cconst_symbol have typ_int to fix no-naked-pointers mode. (Stephen Dolan, review by Mark Shinwell, Xavier Leroy and Vincent Laviron)
    • #9497: Harmonise behaviour between bytecode and native code for recursive module initialisation in one particular case (fixes #9494). (Mark Shinwell, David Allsopp, Vincent Laviron, Xavier Leroy, Geoff Reedy, original bug report by Arlen Cox)
    • #8791: use a variable-length encoding when marshalling bigarray dimensions, avoiding overflow. (Jeremy Yallop, Stephen Dolan, review by Xavier Leroy)
  • Code generation and optimizations:
    • #9441: Add RISC-V RV64G native-code backend. (Nicolás Ojeda Bär, review by Xavier Leroy and Gabriel Scherer)
    • #9316, #9443, #9463, #9782: Use typing information from Clambda for mutable Cmm variables. (Stephen Dolan, review by Vincent Laviron, Guillaume Bury, Xavier Leroy, and Gabriel Scherer; temporary bug report by Richard Jones)
    • #8637, #8805, #9247, #9296: Record debug info for each allocation. (Stephen Dolan and Jacques-Henri Jourdan, review by Damien Doligez, KC Sivaramakrishnan and Xavier Leroy)
    • #9193: Make tuple matching optimisation apply to Lswitch and Lstringswitch. (Stephen Dolan, review by Thomas Refis and Gabriel Scherer)
    • #9392: Visit registers at most once in Coloring.iter_preferred. (Stephen Dolan, review by Pierre Chambart and Xavier Leroy)
    • #9549, #9557: Make -flarge-toc the default for PowerPC and introduce -fsmall-toc to enable the previous behaviour. (David Allsopp, report by Nathaniel Wesley Filardo, review by Xavier Leroy)
  • Language features
    • #8820, #9166: quoted extensions: {%foo|…|} is lighter syntax for [%foo {||}], and {%foo bar|…|bar} for [%foo {bar|…|bar}]. (Gabriel Radanne, Leo White, Gabriel Scherer and Pieter Goetschalckx, request by Bikal Lem)
    • #7364, #2188, #9592, #9609: improvement of the unboxability check for types with a single constructor. Mutually-recursive type declarations can now contain unboxed types. This is based on the paper https://arxiv.org/abs/1811.02300 (Gabriel Scherer and Rodolphe Lepigre, review by Jeremy Yallop, Damien Doligez and Frédéric Bour)
    • #1154, #1706: spellchecker hints and type-directed disambiguation for extensible sum type constructors (Florian Angeletti, review by Alain Frisch, Gabriel Radanne, Gabriel Scherer and Leo White)
    • #6673, #1132, #9617: Relax the handling of explicit polymorphic types. This improves error messages in some polymorphic recursive definition, and requires less polymorphic annotations in some cases of mutually-recursive definitions involving polymorphic recursion. (Leo White, review by Jacques Garrigue and Gabriel Scherer)
    • #9232: allow any class type paths in #-types, For instance, "val f: #F(X).t -> unit" is now allowed. (Florian Angeletti, review by Gabriel Scherer, suggestion by Leo White)
  • Standard library:
    • #9077: Add Seq.cons and Seq.append (Sébastien Briais, review by Yawar Amin and Florian Angeletti)
    • #9235: Add Array.exists2 and Array.for_all2 (Bernhard Schommer, review by Armaël Guéneau)
    • #9226: Add Seq.unfold. (Jeremy Yallop, review by Hezekiah M. Carty, Gabriel Scherer and Gabriel Radanne)
    • #9059: Added List.filteri function, same as List.filter but with the index of the element. (Léo Andrès, review by Alain Frisch)
    • #8894: Added List.fold_left_map function combining map and fold. (Bernhard Schommer, review by Alain Frisch and github user @cfcs)
    • #9365: Set.filter_map and Map.filter_map (Gabriel Scherer, review by Stephen Dolan and Nicolás Ojeda Bär)
    • #9248: Add Printexc.default_uncaught_exception_handler (Raphael Sousa Santos, review by Daniel Bünzli)
    • #8771: Lexing: add set_position and set_filename to change (fake) the initial tracking position of the lexbuf. (Konstantin Romanov, Miguel Lumapat, review by Gabriel Scherer, Sébastien Hinderer, and David Allsopp)
    • #9237: Format.pp_update_geometry ppf (fun geo -> {geo with ...}) for formatter geometry changes that are robust to new geometry fields. (Gabriel Scherer, review by Josh Berdine and Florian Angeletti)
    • #7110: Added Printf.ikbprintf and Printf.ibprintf (Muskan Garg, review by Gabriel Scherer and Florian Angeletti)
    • #9266: Install pretty-printer for the exception Fun.Finally_raised. (Guillaume Munch-Maccagnoni, review by Daniel Bünzli, Gabriel Radanne, and Gabriel Scherer)
  • Other libraries:
    • #9106: Register printer for Unix_error in win32unix, as in unix. (Christopher Zimmermann, review by David Allsopp)
    • #9183: Preserve exception backtrace of exceptions raised by top-level phrases of dynlinked modules. (Nicolás Ojeda Bär, review by Xavier Clerc and Gabriel Scherer)
    • #9320, #9550: under Windows, make sure that the Unix.exec* functions properly quote their argument lists. (Xavier Leroy, report by André Maroneze, review by Nicolás Ojeda Bär and David Allsopp)
    • #9490, #9505: ensure proper rounding of file times returned by Unix.stat, Unix.lstat, Unix.fstat. (Xavier Leroy and Guillaume Melquiond, report by David Brown, review by Gabriel Scherer and David Allsopp)
  • Tools:
    • #9283, #9455, #9457: add a new toplevel directive #use_output "<command>" to run a command and evaluate its output. (Jérémie Dimino, review by David Allsopp)
    • #6969: Argument -nocwd added to ocamldep (Muskan Garg, review by Florian Angeletti)
    • #8676, #9594: turn debugger off in programs launched by the program being debugged (Xavier Leroy, report by Michael Soegtrop, review by Gabriel Scherer)
    • #9057: aid debugging the debugger by preserving backtraces of unhandled exceptions. (David Allsopp, review by Gabriel Scherer)
    • #9276: objinfo: cm[x]a print extra C options, objects and dlls in the order given on the cli. Follow up to #4949. (Daniel Bünzli, review by Gabriel Scherer)
    • #463: objinfo: better errors on object files coming from a different (older or newer), incompatible compiler version. (Gabriel Scherer, review by Gabriel Radanne and Damien Doligez)
    • #9181: make objinfo work on Cygwin and look for the caml_plugin_header symbol in both the static and the dynamic symbol tables. (Sébastien Hinderer, review by Gabriel Scherer and David Allsopp)
    • [ breaking change ] #9197: remove compatibility logic from #244 that was designed to synchronize toplevel printing margins with Format.std_formatter, but also resulted in unpredictable/fragile changes to formatter margins. Setting the margins on the desired formatters should now work. typically on Format.std_formatter. Note that there currently is no robust way to do this from the toplevel, as applications may redirect toplevel printing. In a compiler/toplevel driver, one should instead access Location.formatter_for_warnings; it is not currently exposed to the toplevel. (Gabriel Scherer, review by Armaël Guéneau)
    • #9207, #9210: fix ocamlyacc to work correctly with up to 255 entry points to the grammar. (Andreas Abel, review by Xavier Leroy)
    • #9482, #9492: use diversions (@file) to work around OS limitations on length of Sys.command argument. (Xavier Leroy, report by Jérémie Dimino, review by David Allsopp)
    • #9552: restore ocamloptp build and installation (Florian Angeletti, review by David Allsopp and Xavier Leroy)
  • Manual and documentation:
    • #9141: beginning of the ocamltest reference manual (Sébastien Hinderer, review by Gabriel Scherer and Thomas Refis)
    • #9228: Various Map documentation improvements: add missing key argument in the 'merge' example; clarify the relationship between input and output keys in 'union'; note that find and find_opt return values, not bindings. (Jeremy Yallop, review by Gabriel Scherer and Florian Angeletti)
    • #9255, #9300: reference chapter, split the expression grammar (Florian Angeletti, report by Harrison Ainsworth, review by Gabriel Scherer)
    • #9325: documented base case for List.for_all and List.exists (Glenn Slotte, review by Florian Angeletti)
    • #9410, #9422: replaced naive fibonacci example with gcd (Anukriti Kumar, review by San Vu Ngoc, Florian Angeletti, Léo Andrès)
    • #9541: Add a documentation page for the instrumented runtime; additional changes to option names in the instrumented runtime. (Enguerrand Decorne, review by Anil Madhavapeddy, Gabriel Scherer, Daniel Bünzli, David Allsopp, Florian Angeletti, and Sébastien Hinderer)
    • #9610: manual, C FFI: naked pointers are deprecated, detail the forward-compatible options for handling out-of-heap pointers. (Xavier Leroy, review by Mark Shinwell, David Allsopp and Florian Angeletti)
    • #9618: clarify the Format documentation on the margin and maximum indentation limit (Florian Angeletti, review by Josh Berdine)
    • #8644: fix formatting comment about @raise in stdlib's mli files (Élie Brami, review by David Allsopp)
    • #9327, #9401: manual, fix infix attribute examples (Florian Angeletti, report by David Cadé, review by Gabriel Scherer)
    • #9403: added a description for warning 67 and added a "." at the end of warnings for consistency. (Muskan Garg, review by Gabriel Scherer and Florian Angeletti)
    • #7708, #9580: Ensure Stdlib documentation index refers to Stdlib. (Stephen Dolan, review by Florian Angeletti, report by Hannes Mehnert)
  • Compiler user-interface and warnings:
    • #9712: Update the version format to allow "". The new format is "major.minor[.patchlevel][(+|)additional-info]", for instance "4.12.0~beta1+flambda". This is a documentation-only change for the 4.11 branch, the new format will be used starting with the 4.12 branch. (Florian Angeletti, review by Damien Doligez and Xavier Leroy)
    • #1664: make -output-complete-obj link the runtime native c libraries when building shared libraries like -output-obj. (Florian Angeletti, review by Nicolás Ojeda Bär)
    • #9349: Support [@inlined hint] attribute. (Leo White, review by Stephen Dolan)
    • #2141: generate .annot files from cmt data; deprecate -annot. (Nicolás Ojeda Bär, review by Alain Frisch, Gabriel Scherer and Damien Doligez)
    • [ breaking change ] #7678, #8631: ocamlc -c and ocamlopt -c pass same switches to the C compiler when compiling .c files (in particular, this means ocamlopt passes -fPIC on systems requiring it for shared library support). (David Allsopp, report by Daniel Bünzli, review by Sébastien Hinderer)
    • #9074: reworded error message for non-regular structural types (Florian Angeletti, review by Jacques Garrigue and Leo White, report by Chas Emerick)
    • #8938: Extend ocamlopt option "-stop-after" to handle "scheduling" argument. (Greta Yorsh, review by Florian Angeletti and Sébastien Hinderer)
    • #8945, #9086: Fix toplevel show directive to work with constructors (Simon Parry, review by Gabriel Scherer, Jeremy Yallop, Alain Frisch, Florian Angeletti)
    • #9107: improved error message for exceptions in module signature errors (Gabriel Scherer, review by Florian Angeletti)
    • #9208: -dno-locations option to hide source locations (and debug events) from intermediate-representation dumps (-dfoo). (Gabriel Scherer, review by Vincent Laviron)
    • #9393: Improve recursive module usage warnings (Leo White, review by Thomas Refis)
    • #9486: Fix configuration for the Haiku operating system (Sylvain Kerjean, review by David Allsopp and Sébastien Hinderer)
  • Internal/compiler-libs changes:
    • #9021: expose compiler Longident.t parsers (Florian Angeletti, review by Gabriel Scherer)
    • #9452: Add locations to docstring attributes (Leo White, review by Gabriel Scherer)
    • #463: a new Misc.Magic_number module for user-friendly parsing and validation of OCaml magic numbers. (Gabriel Scherer, review by Gabriel Radanne and Damien Doligez)
    • #1176: encourage better compatibility with older Microsoft C compilers by using GCC's -Wdeclaration-after-statement when available. Introduce Caml_inline to stop abuse of the inline keyword on MSVC and to help ensure that only static inline is used in the codebase (erroneous instance in runtime/win32.c removed). (David Allsopp, review by Oliver Andrieu and Xavier Leroy)
    • #8934: Stop relying on location to track usage (Thomas Refis, review by Gabriel Radanne)
    • #8970: separate value patterns (matching on values) from computation patterns (matching on the effects of a copmutation) in the typedtree. (Gabriel Scherer, review by Jacques Garrigue and Alain Frisch)
    • #9060: ensure that Misc.protect_refs preserves backtraces (Gabriel Scherer, review by Guillaume Munch-Maccagnoni and David Allsopp)
    • #9078: make all compilerlibs/ available to ocamltest. (Gabriel Scherer, review by Sébastien Hinderer)
    • #9079: typecore/parmatch: refactor ppat_of_type and refine the use of backtracking on wildcard patterns (Florian Angeletti, Jacques Garrigue, Gabriel Scherer, review by Thomas Refis)
    • #9081: typedtree, make the pat_env field of pattern data immutable (Gabriel Scherer, review by Jacques Garrigue, report by Alain Frisch)
    • #9178, #9182, #9196: refactor label-disambiguation (Typecore.NameChoice) (Gabriel Scherer, Thomas Refis, Florian Angeletti and Jacques Garrigue, reviewing each other without self-loops)
    • #9321, #9322, #9359, #9361, #9417, #9447: refactor the pattern-matching compiler (Thomas Refis and Gabriel Scherer, review by Florian Angeletti)
    • #9211, #9215, #9222: fix Makefile dependencies in compilerlibs, dynlink, ocamltest. (Gabriel Scherer, review by Vincent Laviron and David Allsopp)
    • #9305: Avoid polymorphic compare in Ident (Leo White, review by Xavier Leroy and Gabriel Scherer)
    • #7927: refactor val_env met_env par_env to class_env (Muskan Garg, review by Gabriel Scherer and Florian Angeletti)
    • #2324, #9613: Replace the caml_int_compare and caml_float_compare (C functions) with primitives. (Greta Yorsh, review by Stephen Dolan and Vincent Laviron)
    • #9246: Avoid rechecking functor applications (Leo White, review by Jacques Garrigue)
    • #9402: Remove sudo:false from .travis.yml (Hikaru Yoshimura)
    • [ breaking change ] #9411: forbid optional arguments reordering with -nolabels (Thomas Refis, review by Frédéric Bour and Jacques Garrigue)
    • #9414: testsuite, ocamltest: keep test artifacts only on failure. Use KEEP_TEST_DIR_ON_SUCCESS=1 to keep all artifacts. (Gabriel Scherer, review by Sébastien Hinderer)
  • Build system:
    • #9250: Add –disable-ocamltest to configure and disable building for non-development builds. (David Allsopp, review by Sébastien Hinderer)
  • Bug fixes:
    • #7520, #9547: Odd behaviour of refutation cases with polymorphic variants (Jacques Garrigue, report by Leo White, reviews by Gabriel Scherer and Leo)
    • #7562, #9456: ocamlopt-generated code crashed on Alpine Linux on ppc64le, arm, and i386. Fixed by turning PIE off for musl-based Linux systems except amd64 (x86_64) and s390x. (Xavier Leroy, review by Gabriel Scherer)
    • #7683, #1499: Fixes one case where the evaluation order in native-code may not match the one in bytecode. (Nicolás Ojeda Bär, report by Pierre Chambart, review by Gabriel Scherer)
    • #7696, #6608: Record expression deleted when all fields specified (Jacques Garrigue, report by Jeremy Yallop)
    • #7741, #9645: Failure to report escaping type variable (Jacques Garrigue, report by Gabriel Radanne, review by Gabriel Scherer)
    • #7817, #9546: Unsound inclusion check for polymorphic variant (Jacques Garrigue, report by Mikhail Mandrykin, review by Gabriel Scherer)
    • #7897, #9537: Fix warning 38 for rebound extension constructors (Leo White, review by Florian Angeletti)
    • #7917, #9426: Use GCC option -fexcess-precision=standard when available, avoiding a problem with x87 excess precision in Float.round. (Xavier Leroy, review by Sébastien Hinderer)
    • #9011: Allow linking .cmxa files with no units on MSVC by not requiring the .lib file to be present. (David Allsopp, report by Dimitry Bely, review by Xavier Leroy)
    • #9064: Relax the level handling when unifying row fields (Leo White, review by Jacques Garrigue)
    • #9097: Do not emit references to dead labels introduced by #2321 (spacetime). (Greta Yorsh, review by Mark Shinwell)
    • #9163: Treat loops properly in un_anf (Leo White, review by Mark Shinwell, Pierre Chambart and Vincent Laviron)
    • #9189, #9281: fix a conflict with Gentoo build system by removing an one-letter Makefile variable. (Florian Angeletti, report by Ralph Seichter, review by David Allsopp and Damien Doligez)
    • #9225: Do not drop bytecode debug info after C calls. (Stephen Dolan, review by Gabriel Scherer and Jacques-Henri Jourdan)
    • #9231: Make sure a debug event (and the corresponding debug information) is inserted after every primitive that can appear in a collected call stack, and make sure ocamlc preserves such events even if they are at tail position. (Jacques-Henri Jourdan, review by Gabriel Scherer)
    • #9244: Fix some missing usage warnings (Leo White, review by Florian Angeletti)
    • #9274, avoid reading cmi file while printing types (Florian Angeletti, review by Gabriel Scherer)
    • #9307, #9345: reproducible env summaries for reproducible compilation (Florian Angeletti, review by Leo White)
    • #9309, #9318: Fix exhaustivity checking with empty types (Florian Angeletti, Stefan Muenzel and Thomas Refis, review by Gabriel Scherer and Thomas Refis)
    • #9335: actually have –disable-stdlib-manpages not build the manpages (implementation conflicted with #8837 which wasn't

    picked up in review) (David Allsopp, review by Florian Angeletti and Sébastien Hinderer)

    • #9343: Re-enable -short-paths for some error messages (Leo White, review by Florian Angeletti)
    • #9355, #9356: ocamldebug, fix a fatal error when printing values whose type involves a functor application. (Florian Angeletti, review by Gabriel Scherer, report by Cyril Six)
    • #9367: Make bytecode and native-code backtraces agree. (Stephen Dolan, review by Gabriel Scherer)
    • #9375, #9477: add forgotten substitution when compiling anonymous modules (Thomas Refis, review by Frédéric Bour, report by Andreas Hauptmann)
    • #9384, #9385: Fix copy scope bugs in substitutions (Leo White, review by Thomas Refis, report by Nick Roberts)
    • [ breaking change ] #9388: Prohibit signature local types with constraints (Leo White, review by Jacques Garrigue)
    • #9406, #9409: fix an error with packed module types from missing cmis. (Florian Angeletti, report by Thomas Leonard, review by Gabriel Radanne and Gabriel Scherer)
    • #9415: Treat open struct as include struct in toplevel (Leo White, review by Thomas Refis)
    • #9416: Avoid warning 58 in flambda ocamlnat (Leo White, review by Florian Angeletti)
    • #9420: Fix memory leak when caml_output_value_to_block raises an exception (Xavier Leroy, review by Guillaume Munch-Maccagnoni)
    • #9428: Fix truncated exception backtrace for C->OCaml callbacks on Power and Z System (Xavier Leroy, review by Nicolás Ojeda Bär)
    • #9623, #9642: fix typing environments in Typedecl.transl_with_constraint (Gabriel Scherer, review by Jacques Garrigue and Leo White, report by Hugo Heuzard)
    • #9695, #9702: no error when opening an alias to a missing module (Jacques Garrigue, report and review by Gabriel Scherer)
    • #9714, #9724: Add a terminator to the caml_domain_state structure to better ensure that members are correctly spaced. (Antonin Décimo, review by David Allsopp and Xavier Leroy)

Set up OCaml 1.1.1

Sora Morimoto announced

This release contains these changes:

  • OCaml installation errors on Windows are now properly propagated.


Sora Morimoto later added

The v1 tag is moved to v1.1.1. So, if you are using the v1 tag, you don't need to do anything.

Lightweight HList – typed heterogeneous collections

Oleg announced

This is a (very late) announcement of the pure OCaml lightweight analogue of HList – typed heterogeneous collections. Such collections store data of various types, and offer element access/modification plus bulk operations such as mapping and iteration. In HList, the type of the collection reflects the types of the elements; therefore all type mismatches can be detected statically. Correspondingly, there is no need to store any type information at run-time.

The library was actually written two years ago. I have just realized I neglected to announce it. The implementation is fairly trivial, but does come useful from time to time, e.g., to implement polyvariadic functions. One interesting facility is the ability to replace an element somewhere in a collection with a new one of a different type. The type of the returned collection is changed accordingly.

A more detailed description, and the pointer to the code is: http://okmij.org/ftp/ML/ML.html#hlist

The code contains several examples, including multi-dimensional cartesian product: converting an HList of regular lists to a list of tuples, elements of the cartesian product.

A somewhat non-trivial application of HList (specifically, type-changing update) is the tagless-final embedding of simply-typed lambda-calculus with De Bruijn levels: http://okmij.org/ftp/tagless-final/cookbook.html#dblevels At first glance, an embedding that relies on OCaml typechecker to check and infer types of lambda-terms seems impossible.

OCaml 4.10.1 released

octachron announced

Following closely the release of OCaml 4.11.0, we have the pleasure of celebrating the birthday of Augustin-Louis Cauchy by announcing the release of OCaml version 4.10.1.

This release is a collection of safe bug fixes, cherry-picked from the 4.11.0 development cycle. If you were using OCaml 4.10.0 and cannot yet upgrade to 4.11.0, this is the release for you.

OCaml 4.10.1 (20 August 2020)

  • Runtime system:
    • #9344, #9368: Disable exception backtraces in bytecode programs built with "-output-complete-exe". At the moment, such programs do not embed debug information and exception backtraces where causing them to crash. (Jérémie Dimino, review by Nicolás Ojeda Bär)
  • Build system:
    • #9531: fix support for the BFD library on FreeBSD (Hannes Mehnert, review by Gabriel Scherer and David Allsopp)
  • Bug fixes:
    • #9068, #9437: ocamlopt -output-complete-obj failure on FreeBSD 12 (Xavier Leroy, report by Hannes Mehnert, review by Sébastien Hinderer)
    • #9165, #9840: Add missing -function-sections flag in Makefiles. (Greta Yorsh, review by David Allsopp)
    • #9495: fix a bug where bytecode binaries compiled with -output-complete-exe would not execute at_exit hooks at program termination (in particular, output channels would not be flushed). (Nicolás Ojeda Bär, review by David Allsopp)
    • #9714, #9724: Use the C++ alignas keyword when compiling in C++ in MSVC. Fixes a bug with MSVC C++ 2015 onwards. (Antonin Décimo, review by David Allsopp and Xavier Leroy)
    • #9736, #9749: Compaction must start in a heap where all free blocks are blue, which was not the case with the best-fit allocator. (Damien Doligez, report and review by Leo White)
  • Tools:
    • #9552: restore ocamloptp build and installation (Florian Angeletti, review by David Allsopp and Xavier Leroy)

OpenID connect

Ulrik Strid announced

I recently released a OpenID connect library to opam named oidc. And today I got it certified by the openid foundation which proves that it's up to par with the specification. At least for the 2(3) profiles that I certified.

I also created a higher level library oidc-client that used the excellent Piaf library under the hood to make it easy to build a client. I can't publish that on opam since it uses unreleased libraries by @anmonteiro but I try to maintain the opam files even though I mainly use esy. This should allow anyone to create websites with simple login with Facebook/Apple/Microsoft/Google.

This has been quite a ride for me, I built and published JOSE to work with JWTs and JWKs and cookie to have the implementation be less ad-hoc. I also built a web framework that I used in the certification process.


obus 1.2.3

Freyr666 announced

I'm pleased to announce a new minor obus release, porting obus ppx from ocaml-migrate-parsetree to the latest ppxlib. Kudos to @NathanReb for the contribution.

If all goes well, the next big major release will come soon, bringing comprehensive docs and modularity. So stay tuned ;)

Yawar Amin then added

For convenience, the repo: https://github.com/ocaml-community/obus

kqueue-ml 0.1.0

Anurag Soni announced

This will probably not be of interest to a lot of users, but i'd like to announce an initial release of kqueue-ml, thin ctypes binding for the Kqueue event notification interface.


So far this has only been tested on macOS and freebsd. At the moment the library only defines the filters and flags that are common to both macOS and freebsd.

Opam link: https://opam.ocaml.org/packages/kqueue/
Source: https://github.com/anuragsoni/kqueue-ml/

Draft of OCaml Scientific Computing book

jrzhao42 announced

Liang @ryanrhymes and I just finished the first draft of our book - OCaml Scientific Computing ! You can read the draft here: https://lnkd.in/dSE6hEg

This book is a summary of our long-term dedication to functional programming and numerical computing. While the entrance bar to data science, ML, and AI becomes lower and lower thanks to the fast development of various powerful frameworks and toolkits. The tool itself remains a black box and mysterious to many data scientists in reality.

This book gives you a very different angle to look at data science, by illustrating how we, hardcore computer scientists and engineering computing experts, build up a high performance numerical system from scratch. In some sense, the book will help you to find the missing link between a basic pseudo random number generator to a fancy deep neural network application.

This book is not for those who just want to "cast the spell", but rather for those who want to "make the magic more magic" :)

Original Post on LinkedIn

An online version of this book can be seen at: https://ocaml.xyz/book/


If you happen to miss a CWN, you can send me a message and I'll mail it to you, or go take a look at the archive or the RSS feed of the archives.

If you also wish to receive it every week by mail, you may subscribe online.