[ANN] Core Suite 112.06.02

[Ben Millwood]

Well, this has been a long time in the pipeline. 112.06 was delayed first by competing with other priorities, then by the winter holidays, then by some mixups with the release process and some build issues on OS X and 32-bit Linux. It's now finally ready for general consumption! (shortly to be superseded by the 112.17 release anyway, oh well.)

The following packages were upgraded:

- async

- async_extended

- async_extra

- async_kernel

- async_unix

- bignum

- bin_prot

- core_bench

- core

- core_extended

- core_kernel

- custom_printf

- jenga

- ocaml_plugin

- pa_bench

- patdiff

- re2

- sexplib

- textutils

- typerep

Files for this release are available on our website and all packages

are in opam:

  https://ocaml.janestreet.com/ocaml-core/112.06/individual/

Unfortunately, the documentation generator has still not been fixed. I'm hopeful it will be soon.

The full changelog:

### async_extended

- Unwound a recent change to `Mailbox` where one invocation of `receive`

  would put aside items, preventing other invocations from noticing

  them.

- Added `Delimited.Row.nth_conv_exn`, as a counterpart to

  `get_conv_exn`.

- Fixed `File_updates` handling of identical mtimes.

### async_extra

- In `Log`, exposed the raw message.

- Changed `Rpc` creators' `connection_state` to be a function that takes

  the connection and returns the state.

    This makes it possible for the connection state to actually get

    a handle on the connection itself, which simplifies a number of

    idioms for using RPC.  In particular, it makes it easier to respond

    with an RPC back to a client over client's own connection.

- Fixed some nondeterministically failing tests.

- In `Log`, made logs discard messages when their output list is empty.

    Also, removed redundant tracking of current level.

- Moved `Udp.bind_to_interface_exn` to `Unix` module in `async_unix`.

- Added `Versioned_typed_tcp.Repeater`.

    Repeater is used in the cases where we want to inspect and possibly

    alter the flow between a client and a server without having to

    change either the client or the server or the protocol between them.

### async_kernel

- Added `Deferred.Sequence` module, analogous to `Deferred.List` but for

  `Core_kernel.Std.Sequence`.

- Modernized code style.

### async_unix

- Added `Writer.behave_nicely_in_pipeline`, which makes a program behave

  nicely when used in a shell pipeline where the consumer goes away.

- Modernized code style.

- Removed spurious overrides in `Signal`: `set` and `signal`.

    These overrides are no longer necessary because the functions moved

    from `Core.Signal` to `Core.Signal.Expert`.

- Moved `async_extra`'s `Udp.bind_to_interface_exn` to `Unix`.

### bignum

- Added functions to round from `Bignum.t` to `Bigint.t`, and to convert

  `Bigint.t` into `Bignum.t`.

### bin_prot

- Sped up `float` and `float array` operations.

- Removed a use of `Obj.magic` in code generated by `pa_bin_prot` for

  polymorphic variants that led to memory unsafety.

    Previously, `pa_bin_prot` generated this kind of code for

    polymorphic variants:

        match Obj.magic (read_int buf pos) with

        | `A as x -> x

        | `B as x -> x

        | `C -> `C (read_float buf pos)

        | _ -> fail

    and this caused the compiler to assume the result is an immediate

    value.  To fix this we removed the `as x -> x` and used the computed

    integer hash.

### core

- Renamed `Linux_ext.gettid` as `Unix.gettid`, and added OpenBSD support.

    `SYS_gettid` is not available on OpenBSD, but is used in

    `Core_extended`. See the mailing list discussion about this here:

    https://groups.google.com/forum/#!topic/ocaml-core/51knlnuJ8MM

    Seems like the OpenBSD alternative is:

        pid_t        getthrid(void);

    although it's not defined in any header file, which is a bit unfortunate.

- Added `Piecewise_linear.precache`, which computes a lookup table that

  speeds up subsequent calls to `Piecewise_linear.get`.

- Added `Time_ns` module, representing times as 63-bit integers of

  nanoseconds since the epoch.

- Fixed build of `unix_stubs.c` on OpenBSD.

- In `Daemon`, fixed an error message regarding `WSTOPPED` (fixes #47).

- Added `Time.Span.Stable.V2`, with sexps that use new suffixes for

  microseconds (`us`) and nanoseconds (`ns`).

    `Time.Span.of_string` supports the new format, but

    `Time.Span.to_string` doesn't yet produce it -- we plan to change

    that later, after the new `of_string` has made it out more widely.

- Added `Time.Span.to_string_hum`, which gives more options for

  rendering time spans.

- Merged the `recvmmsg` stubs in `Bigstring` and `Iobuf`.

    Factored out a shared underlying `recvmmsg` call that both

    stubs use.

    Restored `-pedantic` by avoiding a C99 feature (variable-length

    stack arrays).

- Made `Date.t` abstract, and changed its representation from a 4-word

  record to an immediate int (packing year, month, day).

- In `Daemon`, changed the permissions of the `std{err,out}` files

  generated during daemonization from `0o777` to `0o644`.

- Moved `Thread_safe_queue` from `core` to `core_kernel`.

    This was done so that `Async_kernel` can use it, eliminating one of

    `Async_kernel`'s dependencies on `Core`.

    `Thread_safe_queue_unit_tests` remains `Core`, at least for now,

    because it has some dependencies on other stuff in `Core`.

### core_bench

- Solved a problem in which OCaml 4.02 was optimizing away benchmarks,

  making them meaningless.

### core_extended

- Sped up `String.is_substring` by replacing the OCaml implementation

with a call to libc `memmem`.

    `memmem` runs in 20% of the time, incurs minimal GC pressure, is

    portable among UNIXen that we target, AND it's clearer than the ML

    version.

- Made `Float_ref` support `bin_io` and `sexp`.

- Removed `gettid`, which is now available in `Core.Unix`.

- Added `Fast_int_div` module, which speeds up integer division by

  a fixed divisor.

- Moved `Sexp.of_sexp_allow_extra_fields` to core_kernel.

### core_kernel

- Made `String_id` have `Stable_containers.Comparable`.

- Changed `Gc.disable_compaction` to require an `allocation_policy`.

- Made `Option` match `Invariant.S1`.

- Added `Sequence.filter`, `compare`, and `sexp_of_t`.

- Added `With_return.with_return_option`, abstracting a common pattern

  of `with_return`.

        val with_return        : ('a return -> 'a  ) -> 'a

        val with_return_option : ('a return -> unit) -> 'a option

- Install a handler for uncaught exceptions, using

  `Printexc.set_uncaught_exception_handler`, new in OCaml 4.02.

- Changed `Day_of_week` representation to a normal variant.

- Changed `Exn.handle_uncaught` so that if it is unable to print, it

  still does `exit 1`.

- Added `Sexp.of_sexp_allow_extra_fields`, previously in

  `Core_extended.Sexp`.

- Changed the implementation of `Exn.raise_without_backtrace` to use

  `raise_notrace`, new in OCaml 4.02.

- Added `Float` functions for converting to and from IEEE

  sign/exponent/mantissa.

- Added `String.Caseless` module, which compares and hashes strings

  ignoring case.

- Reimplemented `Type_equal.Id` using extensible types (new in OCaml

  4.02), removing a use of `Obj.magic`.

    Changed `Type_equal.Id.same_witness` to return `option` rather than

    `Or_error`, which allows it to be implemented without allocation.

- Removed a reference to the `Unix` module. Applications using

  `core_kernel` should be able to link without `unix.cma` again.

- Made `Char.is_whitespace` accept `\f` and `\v` as whitespace,

  matching C.

### jenga

- Support for user control of stale-artifact deletion, by allowing

  specification of an artifact-determination policy.

- Expose jenga's internal (and better - only quotes when necessary)

  definition of `Shell.escape` in `Api`

- Removed `Action.shell` from the API, superseded by `Action.process`.

- Changed RPC interface as needed for build manager to switch from

  scraping error messages to RPCs.

- Fixed jenga's per-rule memo table, which mistakenly kept stale values.

- Show what target is being demanded, useful for debugging rules.

- Run user action when persistent format changes.

- When filtering buildable targets by globs, pay attention to the kinds

  allowed by the glob.

    Specifically, if the kinds don't include `` `File `` (i.e. only

    include `` `Directory ``) then we should not see any

    `buildable_targets` in the filtered list.

### ocaml_plugin

- Stopped using the `~exclusive` with `Reader`, because it doesn't work

  on read-only file systems.

    It's not even needed because these files are written atomically.

- Used a generative functor in the generated code, so the user code can

  apply generative functors at toplevel, or unpack first class modules

  that contain type components.

- Fixed bug when mli file references something defined only in

  another ml.

- Made it possible to compile a plugin in one process, and dynload the

  compiled `cmxs` file without starting async in another process.

    This was done with two new APIs in `Ocaml_dynloader.S`:

        val compile_ocaml_src_files_into_cmxs_file

          : dynloader

          -> string list

          -> output_file:string

          -> unit Deferred.Or_error.t

        val blocking_load_cmxs_file : string -> t Or_error.t

- Allowed plugins to optionally have a shebang line.

- Made `Ocaml_dynloader.find_dependencies` also support files with

  shebang lines.

### pa_bench

- Made the code generated by `pa_bench` for `BENCH` not use `ignore`,

  because OCaml 4.02 will remove dead code in some cases, meaning the

  benchmarks are no longer measuring what they should.  Instead the ignore

  is deep inside `Core_bench`, which is likely out of reach of the

  compiler.

    The result of the user functions given to `BENCH_FUN` and

    `BENCH_INDEXED` are changed so they don't have to return unit and

    people are encouraged not to use `ignore` when these functions don't

    return `unit` (you will get the same warning though, i.e. a warning

    if the result of your function is a function too, thus preventing

    unintended partial applications).

    For example, here are a few benchmarks and their output before the

    fix:

        let x = if Random.bool () then 100 else 1001

        let r = ref 0

        BENCH "ig-1" = 10 / x

        BENCH "ig-2" = ()

        BENCH "ig-3" = phys_equal (10 / x) (Obj.magic 0)

        BENCH "ig-4" = r := (10 / x)

        BENCH "ig-5" = r := x

        +----------------+----------+------------+

        | Name           | Time/Run | Percentage |

        +----------------+----------+------------+

        | [misc.ml] ig-1 |   3.92ns |     29.30% |

        | [misc.ml] ig-2 |   3.34ns |     24.95% |

        | [misc.ml] ig-3 |   3.91ns |     29.23% |

        | [misc.ml] ig-4 |  13.37ns |    100.00% |

        | [misc.ml] ig-5 |   3.24ns |     24.20% |

        +----------------+----------+------------+

    Many of the the numbers above are much lower than they should be

    because of the implicit ignores inserted by the benchmark caused the

    division to to eliminated by the compiler. After the fix, the same

    benchmarks produced more meaningful numbers:

        +----------------+----------+------------+

        | Name           | Time/Run | Percentage |

        +----------------+----------+------------+

        | [misc.ml] ig-1 |  12.78ns |     94.55% |

        | [misc.ml] ig-2 |   3.23ns |     23.90% |

        | [misc.ml] ig-3 |  13.51ns |     99.94% |

        | [misc.ml] ig-4 |  13.52ns |    100.00% |

        | [misc.ml] ig-5 |   3.30ns |     24.40% |

        +----------------+----------+------------+

### sexplib

- Improved the implementation of `Exn.sexp_of_t`, using the unique id in

  exceptions in OCaml 4.02.

    We use the identifier to map exception constructors to converters.

---

We hope you find it useful!

-- Ben, on behalf of the Core team