Backport from sid to buster

[hcoop/debian/mlton.git] / CHANGELOG.adoc

= CHANGELOG

== Version 20180206

Here are the changes from version 20130715 to version 20180206.

=== Summary

* Compiler.
  ** Added an experimental LLVM codegen (`-codegen llvm`); requires LLVM tools
  (`llvm-as`, `opt`, `llc`) version ≥ 3.7.
  ** Made many substantial cosmetic improvements to front-end diagnostic
  messages, especially with respect to source location regions, type inference
  for `fun` and `val rec` declarations, signature constraints applied to a
  structure, `sharing type` specifications and `where type` signature
  expressions, type constructor or type variable escaping scope, and
  nonexhaustive pattern matching.
  ** Fixed minor bugs with exception replication, precedence parsing of function
  clauses, and simultaneous `sharing` of multiple structures.
  ** Made compilation deterministic (eliminate output executable name from
  compile-time specified `@MLton` runtime arguments; deterministically generate
  magic constant for executable).
  ** Updated `-show-basis` (recursively expand structures in environments,
  displaying components with long identifiers; append `(* @ region *)`
  annotations to items shown in environment).
  ** Forced amd64 codegen to generate PIC on amd64-linux targets.
* Runtime.
  ** Added `gc-summary-file file` runtime option.
  ** Reorganized runtime support for `IntInf` operations so that programs that
  do not use `IntInf` compile to executables with no residual dependency on GMP.
  ** Changed heap representation to store forwarding pointer for an object in
  the object header (rather than in the object data and setting the header to a
  sentinel value).
* Language.
  ** Added support for selected SuccessorML features; see
  http://mlton.org/SuccessorML for details.
  ** Added `(*#showBasis "file" *)` directive; see
  http://mlton.org/ShowBasisDirective for details.
  ** FFI:
    *** Added `pure`, `impure`, and `reentrant` attributes to `_import`.  An
    unattributed `_import` is treated as `impure`.  A `pure` `_import` may be
    subject to more aggressive optimizations (common subexpression elimination,
    dead-code elimination).  An `_import`-ed C function that (directly or
    indirectly) calls an `_export`-ed SML function should be attributed
    `reentrant`.
  ** ML Basis annotations.
    *** Added `allowSuccessorML {false|true}` to enable all SuccessorML features
    and other annotations to enable specific SuccessorML features; see
    http://mlton.org/SuccessorML for details.
    *** Split `nonexhaustiveMatch {warn|error|igore}` and `redundantMatch
    {warn|error|ignore}` into `nonexhaustiveMatch` and `redundantMatch`
    (controls diagnostics for `case` expressions, `fn` expressions, and `fun`
    declarations (which may raise `Match` on failure)) and `nonexhaustiveBind`
    and `redundantBind` (controls diagnostics for `val` declarations (which may
    raise `Bind` on failure)).
    *** Added `valrecConstr {warn|error|ignore}` to report when a `val rec` (or
    `fun`) declaration redefines an identifier that previously had constructor
    status.
* Libraries.
  ** Basis Library.
    *** Improved performance of `Array.copy`, `Array.copyVec`, `Vector.append`,
    `String.^`, `String.concat`, `String.concatWith`, and other related
    functions by using `memmove` rather than element-by-element constructions.
  ** `Unsafe` structure.
    *** Added unsafe operations for array uninitialization and raw arrays; see
    https://github.com/MLton/mlton/pull/207 for details.
  ** Other libraries.
    *** Updated: ckit library, MLLPT library, MLRISC library, SML/NJ library
* Tools.
  ** mlnlffigen
    *** Updated to warn and skip (rather than abort) when encountering functions
    with `struct`/`union` argument or return type.

=== Details

* 2018-02-6
  ** Remove ancient and unused `cmcat` tool.

* 2018-02-03
  ** Upgrade `gdtoa.tgz`.

* 2018-02-02
  ** Remove docs from `all` target of `./Makefile`; this eliminates the
  `all-no-docs` target (which was frequently used in favor of `all`).

* 2018-01-31
  ** Use C compiler with `-std=gnu11` (rather than `-std=gnu99`).
  ** Revert rudimentary support for `./configure`; the support was so minimal
  that it seems unhelpful to pretend that there are exhaustive compatibility
  checks being performed.  All of the basic configuration can be accomplished
  with simple `make` variable definitions.

* 2018-01-25
  ** Remove (expert, undocumented) `-debug-format` option; the same effect can
  be achieved with `-as-opt` and `-cc-opt`.
  ** Propagate C compiler from `./configure` to `mlton` script.

* 2018-01-24
  ** Extend `-target-*-opt` options to support `arch-os` pairs.
  ** Remove `./package/rpm/*` and corresponding targets in `./Makefile`;
  upstream MLton has not produced RPMs for years.

* 2018-01-24
  ** Slightly improve performance of `Vector.concat` and
  `String.{concat,concatWith,tokens,fields}` by avoiding `List.map`-s.

* 2018-01-23
  ** Restore, but deprecate, `-drop-pass` compile-time expert option.

* 2018-01-19
  ** Update SML/NJ libraries to SML/NJ 110.82.

* 2017-12-29
  ** Add support for `(*#showBasis "file" *)` directives.  This feature is
  meant to facilitate auto-completion via
  https://github.com/MatthewFluet/company-mlton[`company-mlton`] and similar
  tools.

* 2017-12-20
  ** Update performance comparison on website.  Thanks to Curtis Dunham for the
  pull request.

* 2017-12-17
  ** Updates to `-show-basis`:
    *** `-show-basis-flat`: Recursively expand structures in environments,
    displaying components with long identifiers.
    *** `-show-basis-def`: Appends `(* @ region *)` annotations to items shown
    in environment.
    *** `-show-basis-compact`: Tries to optimize vertical space (at the expense
    of long lines).

* 2017-12-11
  ** Drop `_BSD_SOURCE` and `_POSIX_C_SOURCE` feature macros in
  `./runtime/cenv.h`.

* 2017-12-10
  ** Add a `Dockerfile` to build/test MLton.  Thanks to Richard Laughlin for the
  pull request.

* 2017-12-06
  ** Remove `$PREFIX` and `$prefix` from top-level `Makefile.in`; use
  `./configure --prefix path`.  Thanks to Richard Laughlin for the pull
  request.

* 2017-12-03
  ** Fix heap invariant predicates.

* 2017-11-15
  ** Eliminate the use of (some) global mutable state for signal handling.

* 2017-11-14
  ** Store forwarding pointer for an object in the object header (rather than in
  the object data and setting the header to a sentinel value).

* 2017-11-02
  ** Updates to stack management in backend:
    *** Improve `Allocation.Stack.get`.
    *** Do not force `Cont` block arguments to stack.

* 2017-10-30
  ** In `signature SSA_TO_RSSA_STRUCTS` share by `Rssa.Atoms = Ssa.Atoms`.  This
  is the idiom used elsewhere in the compiler, rather than sharing individual
  sub-structures of `Atoms`.
  ** Minor updates to `DirectedGraph` and `Tree` in MLton library.

* 2017-10-23
  ** Add `-seed-rand w` compile-time option, to seed the pseudo-random number
  generator.
  ** Add a new MachineShuffle pass (disabled by default) that shuffles the
  collection of chunks within the program and shuffles the collection of blocks
  within a chunk.  With the `-seed-rand w` compile-time option, can be used to
  generate executables with distinct code placements.

* 2017-10-23
  ** Use a relative path in the `mlton` script, rather than an absolute path.
  The absolute path needed to be set to the intended installation directory,
  which made it difficult to install a binary release in a local directory.
  Undertaken by Maksim Yegorov at RIT supported by NSF CISE Research
  Infrastructure (CRI) award.

* 2017-10-21
  ** Add unsafe operations for array uninitialization and raw arrays.
    *** Rename `Array_uninit: SeqIndex.int -> 'a array` primitive to
    `Array_alloc: SeqIndex.int -> 'a array`.
    *** Add `Array_uninit: 'a array * SeqIndex.int -> unit` primitive to set all
    objptrs in the element at the given index to a bogus non-objptr value
    (`0wx1`).  One motivation for this primitive is to support space-efficient
    polymorphic resizeable arrays.  When shrinking a resizeable array, we would
    like to "`NULL`" out the elements that are no longer part of the logical
    array, in order to avoid a (logical) space leak.
    *** Add `Array_uninitIsNop: 'a array -> bool` primitive to answer if the
    `Array_uninit` primitive applied to the same array would be a nop (i.e., if
    the array has no objptrs in the elements).  This can be used to skip a
    bulk-`Array_uninit` loop when it is known that the `Array_uninit` operations
    would be nops.
    *** Add `Array_allocRaw: SeqIndex.int -> 'a array` primitive to allocate an
    array, but with a header that indicates that the array has no objptrs.  Add
    `Array_toArray: 'a array -> 'a array` primitive to update the header of an
    `Array_allocRaw` allocated array to reveal the objptrs.  One motiviation for
    this primitive is that, in a parallel setting, the uninitialization of an
    array can be a sequential bottleneck.  The `Array_allocRaw` is a constant
    time operation and the subsequent `Array_uninit` operations can be performed
    in parallel.
    *** Extend `structure Unsafe.Array` with additional operations.  See
    `./basis-library/sml-nj/unsafe.sig`.

* 2017-10-20
  ** Introduce ShareZeroVec SSA optimization to share zero-length vectors after
  coercion-based optimizations.  Undertaken by Maksim Yegorov at RIT supported
  by NSF CISE Research Infrastructure (CRI) award.

* 2017-10-18
  ** New canonicalization strategy for CommonSubexp SSA optimization.
  Previously, the canonicalization of commutative arithmetic primitives was
  sensitive to variable hashes (created by an unseeded pseudo-random number
  generator); now, the canonicalization of commutative arithmetic primitives is
  sensitive to relative definition order of variables.

* 2017-10-12
  ** Fix bug in runtime argument option parsing.

* 2017-10-05
  ** Many updates and improvements to diagnostic messages.  See
  https://github.com/MLton/mlton/pull/195 for details.

* 2017-09-27
  ** Add rudimentary support for `./configure`; in particular, support
  `--with-gmp-lib` and `--with-gmp-include` to set location of GMP and
  `--prefix` to specify an install prefix.  Undertaken by Maksim Yegorov at RIT
  supported by NSF CISE Research Infrastructure (CRI) award.

* 2017-08-21
  ** Introduce `Array_copyArray: 'a array * SeqIndex.int * 'a array *
  SeqIndex.int * SeqIndex.int -> unit` and `Array_copyVector: 'a array *
  SeqIndex.int * 'a vector * SeqIndex.int * SeqIndex.int -> unit` primitives
  which are used to implement a number of array and vector construction
  functions, particularly `append`, `concat`, and `concatWith`.  The primitives
  compile to `memmove` operations, which (significantly) outperforms MLton's
  element-by-element construction for large sequences.  Undertaken by Bryan Camp
  at RIT supported by NSF CISE Research Infrastructure (CRI) award.

* 2017-07-25
  ** Force PIC generation on amd64-linux targets.  Thanks to Kuen-Bang Hou
  (Favonia) for the pull request.

* 2017-07-11
  ** Generalize the `subWord` primitives to
+
----
      | WordArray_subWord of {seqSize:WordSize.t, eleSize: WordSize.t}
      | WordArray_updateWord of {seqSize: WordSize.t, eleSize: WordSize.t}
      | WordVector_subWord of {seqSize: WordSize.t, eleSize: WordSize.t}
----
+
Undertaken by Bryan Camp at RIT supported by NSF CISE Research Infrastructure
(CRI) award.

* 2017-07-11
  ** Add a parser combinator library (`structure StreamParser`) to the MLton
  Library.  Undertaken by Jason Carr at RIT supported by NSF CISE Research
  Infrastructure (CRI) award.
  ** Add a parser for the SXML IR (`structure ParseSxml`).  Undertaken by Jason
  Carr at RIT supported by NSF CISE Research Infrastructure (CRI) award.
  ** Allow compilation to start with a `.sxml` file.  Undertaken by Jason Carr
  at RIT supported by NSF CISE Research Infrastructure (CRI) award.

* 2017-06-29
  ** Replace `-drop-pass regex` compile-time option with `-disable-pass regex`
  compile option and add `-enable-pass regex` compile option.  Various XML,
  SXML, SSA, SSA2, RSSA, and Machine IR optimization passes are initialized with
  a default status, which can be overriden by `-{disable,enable}-pass`.  In
  particular, it is now easy to add a work-in-progress (and potentially buggy)
  pass to the simplification pipeline with `execute = false` default status, to
  be selectively executed with `-enable-pass`.  Undertaken by Bryan Camp at RIT
  supported by NSF CISE Research Infrastructure (CRI) award.
  ** Add LoopUnswitch and LoopUnroll SSA optimizations (undertaken by Matthew
  Surawski as an RIT CS MS Capstone Project).  Initial evaluation demonstrates
  some non-trivial performance gains, no non-trivial performance losses, and
  only minor code size increases, but currently disabled pending a more thorough
  evaluation.

* 2017-05-23
  ** Expand the set of MLB annotations:
    *** `nonexhaustiveBind`, `nonexhaustiveExnBind`, `redundantBind`: controls
    diagnostics for `val` declarations (which may raise `Bind` on failure).
    *** `nonexhaustiveMatch`, `nonexhaustiveExnMatch`, `redundantMatch`:
    controls diagnostics for `case` expressions, `fn` expressions, and `fun`
    declarations (which may raise `Match` on failure).
    *** `nonexhaustiveRaise`, `nonexhaustiveExnRaise`, `redundantRaise`:
    controls diagnostics for `handle` expressions (which implicitly re-raise on
    failure).  Note that `nonexhaustiveRaise` and `nonexhaustiveExnRaise`
    default to `ignore`.  The combination of `nonexhaustiveRaise warn` and
    `nonexhaustiveExnRaise ignore` can be useful for finding handlers that
    handle some, but not all, values of an exception variant.
  ** Make a number of improvements to diagnostic messages:
    *** Display nonexhaustive exception patterns as `_ : exn`, rather than
    `e`.
    *** Normalize nonexhaustive patterns by sorting (e.g., by `ConApp` name).
    *** Report complete enumeration of unhandled constants, rather than a single
    example.
    *** Report nonexhaustive patterns of record type as records, rather than as
    tuples.

* 2017-04-20
  ** Updates to SSA, SSA2, and RSSA IR support infrastructure
    *** Display more context when reporting SSA and SSA2 IR type errors.
    *** Add `-layout-width n` compile expert option to control the target width
    for the pretty printer.
    *** Make cosmetic improvments to SSA and SSA2 IR display (uses of global
    variables bound to small constants and conapps are commented with the
    corresponding value; include loop forest for functions with `-keep dot`).
    *** Improve RSSA constant folding and copy propagation.
    *** Limit Machine IR `Globals` to variables used outside of the `main`
    function.

* 2017-04-15
  ** Add `gc-summary-file file` runtime option.

* 2017-04-15
  ** Rename and add `smlnj-mlton-x{2,4,8,16}` top-level `Makefile` targets.
  ** Update SML/NJ librarys to SML/NJ 110.80 (making use of supported
  SuccessorML features).
  ** Not support for SML/NJ extensions via SuccessorML MLB annotations on
  website.

* 2017-04-14
  ** Add support for vector expressions (`#[e1, e2, ..., en]`) and vector
  patterns (`#[p1, p2, ..., pn]`) and add `Vector_vector` n-ary primitive.
  Initial support for vector expressions and the `Vector_vector` primitive were
  undertaken by Krishna Ravikumar as an RIT CS MS Capstone Project.

* 2017-03-29
  ** Update DOS eol handling and tweak error messages in lexer.

* 2017-03-27
  ** Correct off-by-one error in column numbers.  Thanks to Jacob Zimmerman for
  the error report and pull request.

* 2017-03-15
  ** Updates to SuccessorML support:
    *** Add an `allowSuccessorML {false|true}` MLB annotation to enable all
    Successor ML features with a single annotation.
    *** Fix parsing of numeric labels to only accept an INT token that does not
    begin with 0, is not an extended literal, is not negative, and is decimal.
    *** Drop the alternate word prefixes (`0xw` and `0bw`).
    *** Unconditionally allow line comments in MLB files.
    *** Allow UTF-8 byte sequences in text constants.
    *** Refactor `ml.lex` and `mlb.lex` to be more maintainable.
    *** Rename `allowRecPunning` annotation to `allowRecordPunExps`.

* 2017-02-27
  ** Update ML-Yacc examples (`calc`, `fol`, `pascal`) to comply with MLton
  build process.  Thanks to Hai Nguyen Van for the pull request.

* 2017-01-25
  ** Update PortingMLton documentation and `./bin/add-cross` script.  Thanks to
  Daniel Moerner for the pull request.

* 2016-09-29
  ** Constant fold `CPointer_equal(NULL, NULL)` to `true`.

* 2016-09-29
  ** Introduce `NEEDS_SIGALTSTACK_EXEC` config in runtime system.

* 2016-09-27
  ** Construct a devel build version string from last commit time and last
  commit hash.
  ** Omit build date and build node from version banner; makes self-compiles
  deterministic.
  ** Remove `upgrade-basis.sml` from build.  The generated `upgrade-basis.sml`
  was introduced to handle incompatibilities in the Basis Library provided by an
  old version of MLton and the Basis Library assumed by the current sources.
  However, there are no incompatibilities with MLton 20130715, MLton 20100608,
  or MLton 20070826.  Nonetheless, the feature testing performed by
  `./bin/upgrade-basis` to generate `upgrade-basis.sml` is time consuming,
  especially when trying to simply type check the compiler sources.

* 2016-06-20
  ** Do not `gzip` man pages on OpenBSD.  Thanks to Alexander Abushkevich for
  the pull request.

* 2016-06-20
  ** Generate position independent code for OpenBSD.  Thanks to Alexander
  Abushkevich for the pull request.

* 2016-06-20
  ** Fix profiling for amd64-openbsd and x86-openbsd.  Thanks to Alexander
  Abushkevich for the pull request.

* 2016-04-06
  ** Update SML/NJ librarys to SML/NJ 110.79.

* 2016-03-22
  ** Update LLVM codegen to support (and require) >= llvm-3.7.  Thanks to Eugene
  Akentyev for the pull request.

* 2016-02-26
  ** Configure GMP location via `Makefile`.

* 2016-01-10
  ** Fix typo in `mlb-formal.tex`.  Thanks to Jon Sterling for the pull request.

* 2015-11-10
  ** Update SML/NJ librarys to SML/NJ 110.78.  Use `allowOrPats` and
  `allowSigWithtype` to minimize diffs.

* 2015-10-20
  ** Fix elaboration of `withtype` in signature.

* 2015-10-06
  ** Add support for setting CM anchor bindings in `cm2mlb` tool.

* 2015-10-06
  ** Fix non-exhaustive match warnings with or-patterns.  Thanks to Rob Simmons
  for the bug report.
  ** Distinguish between partial and fully redundant matches.
  ** Report partial redundancy in `val` declarations.
  ** Lower precedence of or-patterns in parser.
  ** Make a variety of cosmetic improvements to non-exhaustive and redundant
  error/warning messages, primarily to be consistent in formatting between
  quoted AST and generated messages.

* 2015-07-10
  ** Extend support for arm64 (aarch64).  Thanks to Edmund Evans for the patch.

* 2015-06-22
  ** Introduce `valrecConstr {warn|error|ignore}` MLB annotation to report when
  a `val rec` (or `fun`) declaration redefines an identifier that previously had
  constructor status.

* 2015-06-19
  ** Add support for selected SuccessorML features (undertaken by Kevin Bradley
  as an RIT CS MS Capstone Project).
    *** `do`-declarations (`allowDoDecls`)
    *** extended literals (`allowExtendedLiterals`)
    *** line comments (`allowLineComments`)
    *** optional leading bar in matches, fun decls, and datatype decls
    (`allowOptBar`)
    *** optional trailing semicolon in sequence expressions (`allowOptSemicolon`)
    *** or patterns (`allowOrPats`)
    *** record expression punning (`allowRecPunning`)
    *** withtype in signatures (`allowSigWithtype`)

* 2015-06-10
  ** Hide equality status of poly (and mono) vector and array slices.
  ** Hide type equality of mono and poly `Word8.word` arrays and vectors.

* 2015-06-08
  ** Added `reentrant` attribute to `_import`.  An `_import`-ed C function that
  (directly or indirectly) calls an `_export`-ed SML function should be
  attributed `reentrant`.

* 2015-06-08
  ** Make compilation deterministic:
    *** Eliminate output executable name from compile-time specified `@MLton`
    arguments.
    *** Deterministically generate magic constant for executable.

* 2015-06-08
  ** Add `-keep ast` compile option.  Undertaken by Ross Bayer at RIT supported
  by NSF CISE Research Infrastructure (CRI) award.

* 2015-06-02
  ** Updates to Debian packaging.  Thanks to Christopher Cramer for the pull
  request.

* 2015-03-30
  ** Use `LANG=en_us` when computing version and build date.  Thanks to Eugene
  Akentyev for the pull request.

* 2015-02-17
  ** Update `mlnlffigen` to warn and skip functions with `struct`/`union`
  arguments.  Thanks to Armando Doval for the pull request.

* 2014-12-22
  ** Move pervasive constructs from `./mlton/ast` to `./mlton/atoms`, so that
  `./mlton/ast/sources.mlb` depends on `./mlton/atoms/sources.mlb` (and not the
  other way around).  Undertaken by Vedant Raiththa at RIT supported by NSF CISE
  Research Infrastructure (CRI) award.

* 2014-12-17
  ** Cache a worker thread to service calls of `_export`-ed functions.  Thanks
  to Bernard Berthomieu for the bug report.

* 2014-12-02
  ** Post-process generated front-end files for compatibility with SML/NJ's
  recent `ml-lex` and `ml-yacc` tools that generate log identifiers rather than
  unqualified (top-level environment) identifiers.
  ** Corrected documentation for SML/NJ `Makefile` target and fixed
  `bootstrap-nj` target.  Thanks to Daniel Rosenwasser for the pull request.

* 2014-11-21
  ** Reorganized runtime support for `IntInf` operations so that programs that
  do not use `IntInf` compile to executables with no residual dependency on GMP.
  ** Fixed bug in `MLton.IntInf.fromRep` that could yield values that violate
  the `IntInf` representation invariants. Thanks to Rob Simmons for the bug
  report.

* 2014-10-24
  ** Added `pure` and `impure` attributes to `_import`.  An unattributed
  `_import` is treated as `impure`.  A `pure` `_import` may be subject to more
  aggressive optimizations (common subexpression elimination, dead-code
  elimination).  Undertaken by Vedant Raiththa at RIT supported by NSF CISE
  Research Infrastructure (CRI) award.

* 2014-10-22
  ** Various updates to treatment of `IntInf` constants in the compiler.
    *** Recognize both `Big` and `Small` representations of `IntInf`-s.
    *** Translate `IntInf` consts to `Big` and `Small` representations in
    conversion from SSA to RSSA. This is consistent with the treatment of other
    `IntInf` operations in the conversion. After the conversion, `IntInf` is no
    longer treated as a primitive.
    *** Remove `initIntInfs` from program initialization.
    *** Constant fold `IntInf_toVector` and `WordVector_toIntInf` primitives.

* 2014-10-20
  ** Various updates to `structure WordXVector` in compiler proper.
    *** Update the `WordXVector.layout` function. If the `elementSize` is
   `WordX.word8` and more than 90% of the characters satisfy `Char.isGraph
   orelse Char.isSpace`, then display as an SML string constant (with
   non-printable characters SML-escaped). Otherwise, display as an SML/NJ-style
   `#[0x0, 0xF]` vector literal.
    *** Update initialization of `static struct GC_vectorInit vectorInits[]`
   constants in runtime. If the `WordXVector`'s (primitive) `elementSize` is
   `WordSize.W8`, then emit a C-escaped string constant. Otherwise, emit a
   C-array initialization.

* 2014-08-15
  ** More updates to benchmark infrastructure.
    *** Make `update-counts.sh` script more robust.
    *** Update `hamlet.sml` benchmark program to close input file after each
    loop.
    *** Update `fft.sml` benchmark program to only invoke `test` function with
    power-of-2 arguments.
    *** Update `model-elimination.sml` benchmark program to iterate `main ()`
    according to `doit` size parameter.

* 2014-08-11
  ** Include `winsock2.h` before `windows.h` in MinGW port.  Thanks to Shu-Hung
  You for the pull request.

* 2014-07-31
  ** Refactor array and vector implementation in Basis Library into a primitive
  implementation (using `SeqInt.int` for indexing) and a wrapper implementation
  (using the default `Int.int` for indexing).  Thanks to Rob Simmons for the
  pull request.
  ** Correct description of `MLton.{Vector,Array}.unfoldi` on website.  Thanks
  to Rob Simmons for the pull request.

* 2014-07-14
  ** Updates to benchmark infrastructure.
    *** Add `even-odd.sml` benchmark that exercises mutual tail recursion.
    *** Add `update-counts.sh` script to calculate appropriate benchmark
    iteration counts and update benchmark iteration counts so that all
    benchmarks run for at least 30 seconds.
    *** Updates to benchmark driver program.

* 2014-07-07
  ** Change `./basis-library/integer/int-inf.sml` to reduce dependency on
  GMP-specific details of `./basis-library/integer/int-inf0.sml`.  Thanks to Rob
  Simmons for the pull request.
  ** Correct type and description of `MLton.IntInf.fromRep` on website.  Thanks
  to Rob Simmons for the pull request.

* 2014-07-01
  ** Add experimental LLVM codegen (undertaken by Brian Leibig as an RIT CS MS
  Project).

* 2014-06-09
  ** Update `CallingFromSMLToC` page on website.  Thanks to Bikal Gurung for the
  pull request.

* 2014-03-18
  ** Updates for MinGW port.

* 2014-02-07
  ** Update AsciiDoc sources for website.

* 2013-10-31
  ** Various updates to website.  Thanks to Mauricio C Antunes for the pull
  request.
    *** Add Tofte's tutorial and Rossberg's grammar.
    *** Fix links to implementations.

* 2013-10-10
  ** Update links from `References` page on website.  Thanks to Mauricio C
  Antunes for the pull request.

* 2013-09-02
  ** Fix example for `Lazy` page on website.  Thanks to Daniel Rosenwasser for
  the pull request.

== Version 20130715

Here are the changes from version 20100608 to version 20130715.

=== Summary

* Compiler.
  ** Cosmetic improvements to type-error messages.
  ** Removed features:
    *** Bytecode codegen: The bytecode codegen had not seen significant use and
    it was not well understood by any of the active developers.
    *** Support for `.cm` files as input: The ML Basis system provides much
    better infrastructure for "programming in the very large" than the (very)
    limited support for CM.  The `cm2mlb` tool (available in the source
    distribution) can be used to convert CM projects to MLB projects, preserving
    the CM scoping of module identifiers.
  ** Bug fixes: see changelog
* Runtime.
  ** Bug fixes: see changelog
* Language.
  ** Interpret `(*#line line:col "file" *)` directives as relative
  file names.
  ** ML Basis annotations.
    *** Added: `resolveScope`
* Libraries.
  ** Basis Library.
    *** Improved performance of `String.concatWith`.
    *** Use bit operations for `REAL.class` and other low-level operations.
    *** Support additional variables with `Posix.ProcEnv.sysconf`.
    *** Bug fixes: see changelog
  ** `MLton` structure.
    *** Removed: `MLton.Socket`
  ** Other libraries.
    *** Updated: ckit library, MLRISC library, SML/NJ library
    *** Added: MLLPT library
* Tools.
  ** mllex
    *** Generate `(*#line line:col "file.lex" *)` directives with simple
    (relative) file names, rather than absolute paths.
  ** mlyacc
    *** Generate `(*#line line:col "file.grm" *)` directives with simple
    (relative) file names, rather than absolute paths.
    *** Fixed bug in comment-handling in lexer.

=== Details

* 2013-07-06
  ** Update SML/NJ libraries to SML/NJ 110.76.

* 2013-06-19
  ** Upgrade `gdtoa.tgz`; fixed bug in `Real32.{fmt,toDecimal,toString}`, which
  in some cases produced too many digits

* 2013-06-18
  ** Removed `MLton.Socket` structure (deprecated in last release).

* 2013-06-10
  ** Improved performance of `String.concatWith`.

* 2013-05-22
  ** Update SML/NJ libraries to SML/NJ 110.75.

* 2013-04-30
  ** Detect PowerPC 64 architecture.

* 2012-10-09
  ** Fixed bug in elaboration that erroneously accepted the following:

      signature S = sig structure A : sig type t end
                              and B : sig type t end where type t = A.t end

* 2012-09-04
  ** Introduce an MLB annotation to control overload and flex record resolution
  scope: `resolveScope {strdec|dec|topdec|program}`.

* 2012-07-04
  ** Simplify use of `getsockopt` and `setsockopt` in Basis Library.
  ** Direct implementation of `Socket.Ctl.{getATMARK,getNREAD}` in runtime
  system, rather than indirect implementation in Basis Library via `ioctl`.
  ** Replace use of casting through a union with `memcpy` in runtime.

* 2012-06-11
  ** Use bit operations for `REAL.class` and other low-level operations.
  ** Fixed bugs in `REAL.copySign`, `REAL.signBit`, and `REAL.{to,from}Decimal`.

* 2012-06-01
  ** Cosmetic improvements to type-error messages.
  ** Fixed bug in elaboration that erroneously rejected the following:

      datatype ('a, ''a) t = T
      type ('a, ''a) u = unit

  and erroneously accepted the following:

      fun f (x: 'a) : ''a = x
      fun g (x: 'a) : ''a = if x = x then x else x

* 2012-02-24
  ** Fixed bug in redundant SSA optimization.

* 2011-06-20
  ** Support additional variables with `Posix.ProcEnv.sysconf`.

* 2011-06-17
  ** Change `mllex` and `mlyacc` to generate `#line` directives with simple file
  names, rather than absolute paths.
  ** Interpret `#line` directives as relative file names.

* 2011-06-14
  ** Fixed bug in SSA/SSA2 shrinker that could erroneously turn a non-tail
  function call with a `Bug` transfer as its continuation into a tail function
  call.

* 2011-06-11
  ** Update SML/NJ libraries to SML/NJ 110.73 and add ML-LPT library.

* 2011-06-10
  ** Fixed bug in translation from SSA2 to RSSA with case expressions over
  non-primitive-sized words.
  ** Fixed bug in SSA/SSA2 type checking of case expressions over words.

* 2011-06-04
  ** Upgrade `gdtoa.tgz`.
  ** Remove bytecode codegen.
  ** Remove support for `.cm` files as input.

* 2011-05-03
  ** Fixed a bug with the treatment of `as`-patterns, which should not allow the
  redefinition of constructor status.

* 2011-02-18
  ** Fixed bug with treatment of nan in common subexpression elimination SSA
  optimization.

* 2011-02-18
  ** Fixed bug in translation from SSA2 to RSSA with weak pointers.

* 2011-02-05
  ** Fixed bug in amd64 codegen calling convention for varargs C calls.

* 2011-01-17
  ** Fixed bug in comment-handling in lexer for `mlyacc`'s input language.

* 2010-06-22
  ** Fixed bug in elaboration of function clauses with different numbers of
  arguments that would raise an uncaught `Subscript` exception.


== Version 20100608

Here are the changes from version 20070826 to version 20100608.

=== Summary

* New platforms.
  ** ia64-hpux
  ** powerpc64-aix
* Compiler.
  ** Command-line switches.
    *** Added: `-mlb-path-var '<name> <value>'`
    *** Removed: `-keep sml`, `-stop sml`
  ** Improved constant folding of floating-point operations.
  ** Experimental: Support for compiling to a C library; see wiki documentation.
  ** Extended `-show-def-use` output to include types of variable definitions.
  ** Deprecated features (to be removed in a future release)
    *** Bytecode codegen: The bytecode codegen has not seen significant use and
    it is not well understood by any of the active developers.
    *** Support for `.cm` files as input: The ML Basis system provides much
    better infrastructure for "programming in the very large" than the (very)
    limited support for CM.  The `cm2mlb` tool (available in the source
    distribution) can be used to convert CM projects to MLB projects, preserving
    the CM scoping of module identifiers.
  ** Bug fixes: see changelog
* Runtime.
  ** `@MLton` switches.
    *** Added: `may-page-heap {false|true}`
  ** `may-page-heap`: By default, MLton will not page the heap to disk when
  unable to grow the heap to accomodate an allocation.  (Previously, this
  behavior was the default, with no means to disable, with security an
  least-surprise issues.)
  ** Bug fixes: see changelog
* Language.
  ** Allow numeric characters in ML Basis path variables.
* Libraries.
  ** Basis Library.
    *** Bug fixes: see changelog.
  ** `MLton` structure.
    *** Added: `MLton.equal`, `MLton.hash`, `MLton.Cont.isolate`,
    `MLton.GC.Statistics, `MLton.Pointer.sizeofPointer`,
    `MLton.Socket.Address.toVector`
    *** Changed:
    *** Deprecated: `MLton.Socket`
  ** `Unsafe` structure.
    *** Added versions of all of the monomorphic array and vector structures.
  ** Other libraries.
    *** Updated: ckit library, MLRISC library, SML/NJ library.
* Tools.
  ** `mllex`
    *** Eliminated top-level `type int = Int.int` in output.
    *** Include `(*#line line:col "file.lex" *)` directives in output.
    *** Added `%posint` command, to set the `yypos` type and allow the lexing of
    multi-gigabyte files.
   ** `mlnlffigen`
    *** Added command-line switches `-linkage archive` and `-linkage shared`.
    *** Deprecated command-line switch `-linkage static`.
    *** Added support for ia64 and hppa targets.
   ** `mlyacc`
    *** Eliminated top-level `type int = Int.int` in output.
    *** Include `(*#line line:col "file.grm" *)` directives in output.

=== Details

* 2010-05-12
  ** Fixed bug in the mark-compact garbage collector where the C library's
  `memcpy` was used to move objects during the compaction phase; this could lead
  to heap corruption and segmentation faults with newer versions of `gcc` and/or
  `glibc`, which assume that src and dst in a `memcpy` do not overlap.

* 2010-03-12
  ** Fixed bug in elaboration of `datatype` declarations with `withtype`
  bindings.

* 2009-12-11
  ** Fixed performance bug in RefFlatten SSA2 optimization.

* 2009-12-09
  ** Fixed performance bug in SimplifyTypes SSA optimization.

* 2009-12-02
  ** Fixed bug in amd64 codegen register allocation of indirect C calls.

* 2009-09-17
  ** Fixed bug in `IntInf.scan` and `IntInf.fromString` where leading spaces
  were only accepted if the stream had an explicit sign character.

* 2009-07-10
  ** Added CombineConversions SSA optimization.

* 2009-06-09
  ** Removed deprecated command line switch `-show-anns {false, true}`.

* 2009-04-18
  ** Removed command line switches `-keep sml` and `-stop sml`.  Their meaning
  was unclear with `.mlb` files; their effect with `.cm` files can be achieved
  with `-stop f`.

* 2009-04-16
  ** Fixed bug in `IntInf.~>>` that could cause a `glibc` assertion failure.

* 2009-04-01
  ** Fixed exported type of `MLton.Process.reap`.

* 2009-01-27
  ** Added `MLton.Socket.Address.toVector` to get the network-byte-order
  representation of an IP address.

* 2008-11-10
  ** Fixed bug in `MLton.size` and `MLton.share` when tracing the current stack.

* 2008-10-27
  ** Fixed phantom typing of sockets by hiding the representation of socket
  types.  Previously the representation of sockets was revealed rendering the
  phantom types useless.

* 2008-10-10
  ** Fixed bug in nested `_export`/`_import` functions.

* 2008-09-12
  ** Improved constant folding of floating point operations.

* 2008-08-20
  ** Store the card/cross map at the end of the allocated ML heap; avoids
  possible out of memory errors when resizing the ML heap cannot be followed by
  a card/cross map allocation.

* 2008-07-24
  ** Added support for compiling to a C library. The relevant new compiler
  options are `-ar` and `-format`. Libraries are named based on the name of the
  `-export-header` file. Libraries have two extra methods:
    *** `NAME_open(argc, argv)` initializes the library and runs the SML code
    until it reaches the end of the program. If the SML code exits or raises an
    uncaught exception, the entire program will terminate.
    *** `NAME_close()` will execute any registered atExit functions, any
    outstanding finalizers, and frees the ML heap.

* 2008-07-16
  ** Fixed bug in the name mangling of `_import`-ed functions with the `stdcall`
  convention.

* 2008-06-12
  ** Added `MLton.Pointer.sizeofPointer`.

* 2008-06-06
  ** Added expert command line switch `-emit-main {true|false}`.

* 2008-05-17
  ** Fixed bug in Windows code to page the heap to disk when unable to grow the
  heap to a desired size. Thanks to Sami Evangelista for the bug report.

* 2008-05-10
  ** Implemented `MLton.Cont.isolate`.

* 2008-04-20
  ** Fixed bug in *NIX code to page the heap to disk when unable to grow the
  heap to a desired size. Thanks to Nicolas Bertolotti for the bug report and
  patch.

* 2008-04-07
  ** More flexible active/paused stack resizing policy. +
  Removed `thread-shrink-ratio` runtime option. + Added
  `stack-current-grow-ratio`, `stack-current-max-reserved-ratio`,
  `stack-current-permit-ratio`, `stack-current-shrink-ratio`,
  `stack-max-reserved-ratio`, and `stack-shrink-ratio` runtime options.

* 2008-04-07
  ** Fixed bugs in Basis Library where the representations of `OS.IO.iodesc`,
  `Posix.IO.file_desc`, `Posix.Signal.signal`, `Socket.sock`,
  `Socket.SOGK.sock_type` as integers were exposed.

* 2008-03-14
  ** Added unsafe versions of all of the monomorphic array and vector
  structures.

* 2008-03-02
  ** Fixed bug in Basis Library where the representation of `OS.Process.status`
  as an integer was exposed.

* 2008-02-13
  ** Fixed space-safety bug in RefFlatten optimization (to flatten refs into
  containing data structure). Thanks to Daniel Spoonhower for the bug report and
  initial diagnosis and patch.

* 2008-01-25
  ** Various updates to GC statistics gathering.  Some basic GC statistics can
  be accessed from SML by `MLton.GC.Statistics.*` functions.

* 2008-01-24
  ** Added primitive (structural) polymorphic hash.

* 2008-01-21
  ** Fixed frontend to accept `op _longvid_` patterns and expressions. Thanks to
     Florian Weimer for the bug report.

* 2008-01-17
  ** Extended `-show-def-use` output to include types of variable definitions.

* 2008-01-09
  ** Extended `MLton_equal` to be a structural equality on all types, including
  `real` and `->` types.

* 2007-12-18
  ** Changed ML-Yacc and ML-Lex to output line directives so that MLton's
  def-use information points to the source files (`.grm` and `.lex`) instead of
  the generated implementations (`.grm.sml` and `.lex.sml`).

* 2007-12-14
  ** Added runtime option `may-page-heap {false|true}`.  By default, MLton will
  not page the heap to disk when unable to grow the heap to a desired size.
  (Previously, this behavior was the default, with no means to disable, with
  security and least-surprise concerns.) Thanks to Wesley Terpstra for the
  patch.
  ** Fixed bug the FFI visible representation of `Int16.int ref` (and references
  of other primitive types smaller than 32-bits) on big-endian platforms. Thanks
  to Dave Herman for the bug report.

* 2007-12-13
  ** Fixed bug in `ImperativeIOExtra.canInput` (`TextIO.canInput`). Thanks to
     Ville Laurikari for the bug report.

* 2007-12-09
  ** Better constant folding of `IntInf` operations.

* 2007-12-07
  ** Fixed bug in algebraic simplification of `RealX` primitives.  `Real.<= (x,
  x)` is `false` when `x` is `NaN`.

* 2007-11-29
  ** Fixed bug in type inference of flexible records.  This would later cause
  the compiler to raise the `TypeError` exception. Thanks to Wesley Terpstra for
  the bug report.

* 2007-11-28
  ** Fixed bug in cross-compilation of `gdtoa` library. Thanks to Wesley
  Terpstra for the bug report and patch.

* 2007-11-20
  ** Fixed bug in RefFlatten optimization (pass to flatten refs into containing
  data structure). Thanks to Ruy LeyWild for the bug report.

* 2007-11-19
  ** Fixed bug in the handling of weak pointers by the mark-compact garbage
  collector. Thanks to Sean McLaughlin for the bug report and Florian Weimer for
  the initial diagnosis.

* 2007-11-07
  ** Added `%posint` command to `ml-lex`, to set the `yypos` type and allow the
  lexing of multi-gigabyte input files. Thanks to Florian Weimer for the feature
  concept and original patch.

* 2007-11-07
  ** Added command-line switch `-mlb-path-var '<name> <value>'` for specifying
  MLB path variables.

* 2007-11-06
  ** Allow numeric characters in MLB path variables.

* 2007-09-20
  ** Fixed bug in elaboration of structures with signature constraints.  This
  would later cause the compiler to raise the `TypeError` exception. Thanks to
  Vesa Karvonen for the bug report.

* 2007-09-11
  ** Fixed bug in interaction of `_export`-ed functions and signal
  handlers. Thanks to Sean McLaughlin for the bug report.

* 2007-09-03
  ** Fixed bug in implementation of `_export`-ed functions using `char`
  type. Thanks to Katsuhiro Ueno for the bug report.


== Version 20070826

Here are the changes from version 20051202 to version 20070826.

=== Summary

* New platforms:
  ** amd64-linux, amd64-freebsd
  ** hppa-hpux
  ** powerpc-aix
  ** x86-darwin (Mac OS X)
* Compiler.
  ** Support for 64-bit platforms.
    *** Native amd64 codegen.
  ** Command-line switches.
    *** Added: `-codegen amd64`, `-codegen x86`, `-default-type <type>`,
    `-profile-val {false|true}`.
    *** Changed: `-stop f` (file listing now includes `.mlb` files)
  ** Bytecode codegen.
    *** Support for profiling.
    *** Support for exception history.
* Language.
  ** ML Basis annotations.
    *** Removed: `allowExport`, `allowImport`, `sequenceUnit`, `warnMatch`.
* Libraries.
  ** Basis Library.
    *** Added: `PackWord16Big, `PackWord16Little`, `PackWord64Big`,
    `PackWord64Little`.
    *** Bug Fixes: see changelog.
  ** `MLton` structure.
    *** Added: `MLTON_MONO_ARRAY`, `MLTON_MONO_VECTOR`, `MLTON_REAL`,
    `MLton.BinIO.tempPrefix`, `MLton.CharArray`, `MLton.CharVector`,
    `MLton.IntInf.BigWord`, `MLton.IntInf.SmallInt`,
    `MLton.Exn.defaultTopLevelHandler`, `MLton.Exn.getTopLevelHandler`,
    `MLton.Exn.setTopLevelHandler`, `MLton.LargeReal`, `MLton.LargeWord`,
    `MLton.Real`, `MLton.Real32`, `MLton.Real64`, `MLton.Rlimit.Rlim`,
    `MLton.TextIO.tempPrefix`, `MLton.Vector.create`, `MLton.Word.bswap`,
    `MLton.Word8.bswap`, `MLton.Word16`, `MLton.Word32`, `MLton.Word64`,
    `MLton.Word8Array`, `MLton.Word8Vector`.
    *** Changed: `MLton.Array.unfoldi`, `MLton.IntInf.rep`, `MLton.Rlimit`,
    `MLton.Vector.unfoldi`.
    *** Deprecated: `MLton.Socket`
  ** Other libraries.
    *** Added: MLRISC libary.
    *** Updated: ckit library, SML/NJ library.
* Tools.

=== Details

* 2007-08-12
  ** Removed deprecated ML Basis annotations.

* 2007-08-06
  ** Fixed bug in treatment of `Real<N>.{scan,fromString}` operations.
  `Real<N>.{scan,fromString}` were using `TO_NEAREST` semantics, but should obey
  current rounding mode.  (Only `Real<N>.fromDecimal` is specified to always
  have `TO_NEAREST` semantics.)  Thanks to Sean McLaughlin for the bug report.

* 2007-07-27
  ** Fixed bugs in constant-folding of floating-point operations with C codegen.

* 2007-07-26
  ** Fixed bug in treatment of floating-point operations.  Floating-point
  operations depend on the current rounding mode, but were being treated as
  pure. Thanks to Sean McLaughlin for the bug report.

* 2007-07-13
  ** Added `MLton.Exn.{default,get,set}TopLevelHandler`.

* 2007-07-12
  ** Restored `native` option to `-codegen` flag.

* 2007-07-11
  ** Fixed bug in `Real32.toInt`: conversion of real values close to
  `Int.maxInt` could be incorrect.

* 2007-07-07
  ** Updates to bytecode code generator: support for amd64-* targets, support
  for profiling (including exception history).
  ** Fixed bug in `Socket` module of Basis Library; unmarshalling of socket
  options (for `get*` functions) used `andb` rather than `orb`. Thanks to Anders
  Petersson for the bug report (and patch).

* 2007-07-06
  ** Fixed bug in `Date` module of Basis Library; some functions would
  erroneously raise `Date` when given a year <= 1900. Thanks to Joe Hurd for the
  bug report.
  ** Fixed a long-standing bug in monomorphisation pass. Thanks to Vesa Karvonen
  for the bug report.

* 2007-05-18
  ** Native amd64 code generator for amd64-* targets.
  ** Eliminate `native` option from `-codegen` flag.
  ** Add `x86` and `amd64` options to `-codegen` flag.

* 2007-04-29
  ** Improved type checking of RSSA and Machine ILs.

* 2007-04-14
  ** Fixed aliasing issues with `basis/Real/*.c` files.
  ** Added real/word casts in `MLton` structure.

* 2007-04-12
  ** Added primitives for bit cast of word to/from real.
  ** Implement `PackReal<N>{Big,Little}` using `PackWord<N>{Big,Little}` and bit
  casts.

* 2007-04-11
  ** Move all system header `#include`-s to `platform/` os headers.
  ** Use C99 `<assert.h>`, rather than custom `"assert.{h,c}"`.

* 2007-03-13
  ** Implement `PackWord<N>{Big,Little}` entirely in ML, using an ML byte swap
  function.

* 2007-02-25
  ** Change amd64-* target platforms from 32-bit compatibility mode (i.e.,
  `-m32`) to 64-bit mode (i.e., `-m64`).  Currently, only the C codegen is able
  to generate 64-bit executables.

* 2007-02-23
  ** Removed expert command line switch `-coalesce <n>`.
  ** Added expert command line switch `-chunkify {coalesce<n>|func|one}`.

* 2007-02-20
  ** Fixed bug in `PackReal<N>.toBytes`. Thanks to Eric McCorkle for the bug
  report (and patch).

* 2007-02-18
  ** Added command line switch `-profile-val`, to profile the evaluation of
  `val` bindings; this is particularly useful with exception history for
  debugging uncaught exceptions at the top-level.

* 2006-12-29
  ** Added command line switch `-show {anns|path-map}` and deprecated command
  line switch `-show-anns {false|true}`.  Use `-show path-map` to see the
  complete MLB path map as seen by the compiler.

* 2006-12-20
  ** Changed the output of command line switch `-stop f` to include `.mlb`
  files.  This is useful for generating Makefile dependencies.  The old output
  is easy to recover if necessary (e.g. `grep -v '\.mlb$'`).

* 2006-12-08
  ** Added command line switches `-{,target}-{as,cc,link}-opt-quote`, which pass
  their argument as a single argument to `gcc` (i.e., without tokenization at
  spaces).  These options support using headers and libraries (including the
  MLton runtime headers and libraries) from a path with spaces.

* 2006-12-02
  ** Extensive reorganization of garbage collector, runtime system, and Basis
  Library implementation. (This is in preparation for future 64bit support.)
  They should be more C standards compliant and easier to port to new systems.
  ** FFI revisions
    *** Disallow nested indirect types (e.g., `int array array`).

* 2006-11-30
  ** Fixed a bug in elaboration of FFI forms; unary FFI types (e.g., `array`,
  `ref`, `vector`) could be used in places where `MLton.Pointer.t` was required.
  This would later cause the compiler to raise the `TypeError` exception, along
  with a lot of XML IL.

* 2006-11-19
  ** On *-darwin, work with GnuMP installed via Fink or MacPorts.

* 2006-10-30
  ** Ported to x86-darwin.

* 2006-09-23
  ** Added missing specification of `find` to the `MONO_VECTOR` signature.

* 2006-08-03
  ** Fixed a bug in Useless SSA optimization, caused by calling an imported C
  function and then ignoring the result.

* 2006-06-24
  ** Fixed a bug in pass to flatten data structures. Thanks to Joe Hurd for the
  bug report.

* 2006-06-08
  ** Fixed a bug in the native codegen's implementation of the C-calling
  convention.

* 2006-05-11
  ** Ported to PowerPC-AIX.
  ** Fixed a bug in the runtime for the cases where nonblocking IO with sockets
  was implemented using `MSG_DONTWAIT`.  This flag does not exist on AIX,
  Cygwin, HPUX, and MinGW and was previously just ignored.  Now the runtime
  simulates the flag for these platforms (except MinGW, yet, where it's still
  ignored).

* 2006-05-06
  ** Added `-default-type '<ty><N>'` for specifying the binding of default types
  in the Basis Library (e.g., `Int.int`).

* 2006-04-25
  ** Ported to HPPA-HPUX.
  ** Fixed `PackReal{,32,64}{Big,Little}` to follow the Basis Library
  specification.

* 2006-04-19
  ** Fixed a bug in `MLton.share` that could cause a segfault.

* 2006-03-30
  ** Changed `MLton.Vector.unfoldi` to return the state in addition to the
  result vector.

* 2006-03-30
  ** Added `MLton.Vector.create`, a more powerful vector-creation function than
  is available in the basis library.

* 2006-03-04
  ** Added MLRISC from SML/NJ 110.57 to standard distribution.

* 2006-03-03
  ** Fixed bug in SSA simplifier that could eliminate an irredundant test.

* 2006-03-02
  ** Ported a bugfix from SML/NJ for a bug with the combination of `withNack`
  and `never` in CML.

* 2006-02-09
  ** Support compiler specific annotations in ML Basis files.  If an annotation
  contains `:`, then the text preceding the `:` is meant to denote a compiler.
  For MLton, if the text preceding the `:` is equal to `mlton`, then the
  remaining annotation is scanned as a normal annotation.  If the text preceding
  the `:` is not-equal to `mlton`, then the annotation is ignored, and no
  warning is issued.

* 2006-02-04
  ** Fixed bug in elaboration of functors; a program with a very large number of
  functors could exhibit the error `ElaborateEnv.functorClosure: firstTycons`.


== Version 20051202

Here are the changes from version 20041109 to version 20051202.

=== Summary

* New license: BSD-style instead of GPL.
* New platforms:
  ** hppa: Debian Linux.
  ** x86: MinGW.
* Compiler.
  ** improved exception history.
  ** Command-line switches.
    *** Added: `-as-opt`, `-mlb-path-map`, `-target-as-opt`, `-target-cc-opt`.
    *** Deprecated: none.
    *** Removed: `-native`, `-sequence-unit`, `-warn-match`, `-warn-unused`.
* Language.
  ** FFI syntax changes and extensions.
    *** Added: `_symbol`.
    *** Changed: `_export`, `_import`.
    *** Removed: `_ffi`.
  ** ML Basis annotations.
    *** Added: `allowFFI`, `nonexhaustiveExnMatch`, `nonexhaustiveMatch`,
    `redundantMatch`, `sequenceNonUnit`.
    *** Deprecated: `allowExport`, `allowImport`, `sequenceUnit`, `warnMatch`.
* Libraries.
  ** Basis Library.
    *** Added: `Int1`, `Word1`.
  ** `MLton` structure.
    *** Added: `Process.create`, `ProcEnv.setgroups`, `Rusage.measureGC`,
        `Socket.fdToSock`, `Socket.Ctl.getError`.
    *** Changed: `MLton.Platform.Arch`.
  ** Other libraries.
    *** Added: ckit library, ML-NLFFI library, SML/NJ library.
* Tools.
  ** updates of `mllex` and `mlyacc` from SML/NJ.
  ** added `mlnlffigen`.
  ** profiling supports better inclusion/exclusion of code.

=== Details

* 2005-11-19
  ** Updated SML/NJ Library and CKit Library from SML/NJ 110.57.

* 2005-11-15
  ** Fixed a bug in `MLton.ProcEnv.setgroups`.

* 2005-11-11
  ** Fixed a bug in the interleaving of lexing/parsing and elaborating of ML
  Basis files, which would raise an unhandled `Force` exception on cyclic basis
  references. Thanks to John Dias for the bug report.

* 2005-11-10
  ** Fixed two bugs in `Time.scan`.  One would raise `Time` on a string with a
  large fractional component.  Thanks to Carsten Varming for the bug report.
  The other failed to scan strings with an explicit sign followed by a decimal
  point.

* 2005-11-03
  ** Removed `MLton.GC.setRusage`.
  ** Added `MLton.Rusage.measureGC`.

* 2005-09-11
  ** Fixed bug in display of types with large numbers of type variables, which
  could cause unhandled exception `Chr`.

* 2005-09-08
  ** Fixed bug in type inference of flexible records that would show up as
  `"Type error: variable applied to wrong number of type args"`.

* 2005-09-06
  ** Fixed bug in `Real.signBit`, which had assumed that the underlying C
  signbit returned 0 or 1, when in fact any nonzero value is allowed to indicate
  the signbit is set.

* 2005-09-05
  ** Added `-mlb-path-map` switch.

* 2005-08-25
  ** Fixed bug in `MLton.Finalizable.touch`, which was not keeping alive
  finalizable values in all cases.

* 2005-08-18
  ** Added SML/NJ Library and CKit Library from SML/NJ 110.55 to standard
  distribution.
  ** Fixed bug in `Socket.Ctl.*`, which got the endianness wrong on big-endian
  machines. Thanks to Wesley Terpstra for the bug report and fix.
  ** Added `MLton.GC.setRusage`.
  ** Fixed bug in `mllex`, which had file positions starting at 2.  They now
  start at zero.

* 2005-08-15
  ** Fixed bug in `LargeInt.scan`, which should skip leading `"0x"` and `"0X"`.
  Thanks to Wesley Terpstra for the bug report and fix.

* 2005-08-06
  ** Additional revisions of FFI:
    *** Deprecated `_export` with incomplete annotation.
    *** Added `_address` for address of C objects.
    *** Eliminated address component of `_symbol`.
    *** Changed the type of the `_symbol*` expression.
    *** See documentation for more detail.

* 2005-08-06
  ** Annotation changes:
    *** Deprecated: `sequenceUnit`
    *** Added: `sequenceNonUnit`

* 2005-08-03
  ** Annotation changes:
    *** Deprecated: `allowExport`, `allowImport`, `warnMatch`
    *** Added: `allowFFI`, `nonexhaustiveExnMatch`, `nonexhaustiveMatch`,
    `redundantMatch`

* 2005-08-01
  ** Update `mllex` and `mlyacc` with SML/NJ 110.55+ versions.  This
  incorporates a small number of minor bug fixes.

* 2005-07-23
  ** Fixed bug in pass to flatten refs into containing data structure.

* 2005-07-23
  ** Overhaul of FFI:
    *** Deprecated `_import` of C base types.
    *** Added `_symbol` for address, getter, and setter of C base types.
    *** See documentation for more detail.

* 2005-07-21
  ** Update `mllex` and `mlyacc` with SML/NJ 110.55 versions.  This incorporates
  a small number of minor bug fixes.

* 2005-07-20
  ** Fixed bug in front end that allowed unary constructors to be used without
  an argument in patterns.

* 2005-07-19
  ** Eliminated `_ffi`, which has been deprecated for some time.

* 2005-07-14
  ** Fixed bug in runtime that caused getrusage to be called on every GC, even
  if timing info isn't needed.

* 2005-07-13
  ** Fixed bug in closure conversion tickled by making a weak pointer to a
  closure.

* 2005-07-12
  ** Changed `{OS,Posix}.Process.sleep` to call `nanosleep()` instead of
  `sleep()`.
  ** Added `MLton.ProcEnv.setgroups`.

* 2005-07-11
  ** `InetSock.{any,toAddr}` raise `SysErr` if port is not in [0, 2^16^).

* 2005-07-02
  ** Fixed bug in `Socket.recvVecFrom{,',NB,NB'}`.  The type was too polymorphic
  and allowed the creation of a bogus `sock_addr`.

* 2005-06-28
  ** The front end now reports errors on encountering undefined or cyclicly
  defined MLB path variables.

* 2005-05-22
  ** Fixed bug in `Posix.IO.{getlk,setlk,setlkw}` that caused a link-time error:
  undefined reference to `Posix_IO_FLock_typ`.
  ** Improved exception history so that the first entry in the history is the
  source position of the raise, and the rest is the call stack.

* 2005-05-19
  ** Improved exception history for `Overflow` exceptions.

* 2005-04-20
  ** Fixed a bug in pass to flatten refs into containing data structure.

* 2005-04-14
  ** Fixed a front-end bug that could cause an internal bug message of the form
  `"missing flexInst"`.

* 2005-04-13
  ** Fixed a bug in the representation of flat arrays/vectors that caused
  incorrect behavior when the element size was 2 or 4 bytes and there were
  multiple components to the element (e.g. `(char * char) vector`).

* 2005-04-01
  ** Fixed a bug in `GC_arrayAllocate` that could cause a segfault.

* 2005-03-22
  ** Added structures `Int1`, `Word1`.

* 2005-03-19
  ** Fixed a bug that caused `Socket.Ctl.{get,set}LINGER` to raise `Subscript`.
  The problem was in the use of `PackWord32Little.update`, which scales the
  supplied index by `bytesPerElem`.

* 2005-03-13
  ** Fixed a bug in CML mailboxes.

* 2005-02-26
  ** Fixed an off-by-one error in `mkstemp` defined in `mingw.c`.

* 2005-02-13
  ** Added `mlnlffigen` tool (heavily adapted from SML/NJ).

* 2005-02-12
  ** Added MLNLFFI Library (heavily adapted from SML/NJ) to standard
  distribution.

* 2005-02-04
  ** Fixed a bug in `OS.path.toString`, which did not raise `InvalidArc` when
  needed.

* 2005-02-03
  ** Fixed a bug in `OS.Path.joinDirFile`, which did not raise `InvalidArc` when
  passed a file that was not an arc.

* 2005-01-26
  ** Fixed a front end bug that incorrectly rejected expansive __valbind__s with
  useless bound type variables.

* 2005-01-22
  ** Fixed x86 codegen bug which failed to account for the possibility that a
  64-bit move could interfere with itself (as simulated by 32-bit moves).

* 2004-12-22
  ** Fixed `Real32.fmt StringCvt.EXACT`, which had been producing too many
  digits of precision because it was converting to a `Real64.real`.

* 2004-12-15
  ** Replaced MLB path variable `MLTON_ROOT` with `SML_LIB`, to use a more
  compiler-independent name.  We will keep `MLTON_ROOT` aliased to `SML_LIB`
  until after the next release.

* 2004-12-02
  ** `Unix.create` now works on all platforms (including Cygwin and MinGW).

* 2004-11-24
  ** Added support for `MLton.Process.create`, which works on all platforms
  (including Windows-based ones like Cygwin and MinGW) and allows better control
  over `std{in,out,err}` for child process.


== Version 20041109

Here are the changes from version 20040227 to 20041109.

=== Summary

* New platforms:
  ** x86: FreeBSD 5.x, OpenBSD
  ** PowerPC: Darwin (MacOSX)
* Support for MLBasis files.
* Support for dynamic libraries.
* Support for Concurrent ML (CML).
* New structures: `Int2`, `Int3`, ..., `Int31` and `Word2`, `Word3`, ..., `Word31`.
* A new form of profiling: `-profile count`.
* A bytecode generator.
* Data representation improvements.
* `MLton` structure changes.
   ** Added: `share`, `shareAll`
   ** Changed: `Exn`, `IntInf`, `Signal`, `Thread`.
* Command-line switch changes.
   ** Deprecated:
     *** `-native` (use `-codegen`)
     *** `-sequence-unit` (use `-default-ann`)
     *** `-warn-match` (use `-default-ann`)
     *** `-warn-unused` (use `-default-ann`)
   ** Removed:
     *** `-detect-overflow`
     *** `-exn-history` (use `-const`)
     *** `-safe`
     *** `-show-basis-used`
   ** Added:
     *** `-codegen`
     *** `-const`
     *** `-default-ann`
     *** `-disable-ann`
     *** `-profile-branch`
     *** `-target-link-opt`

=== Details

* 2004-09-22
  ** Extended `_import` to support indirect function calls.

* 2004-09-13
  ** Made `Date.{fromString,scan}` accept a space (treated as zero) in the first
  character of the day of the month.

* 2004-09-12
  ** Fixed bug in `IntInf` that could cause a segfault.
  ** Remove `MLton.IntInf.size`.

* 2004-09-05
  ** Made `-detect-overflow` and `-safe` expert options.

* 2004-08-30
  ** Added `val MLton.share: 'a -> unit`, which maximizes sharing in a heap
  object.

* 2004-08-27
  ** Fixed bug in `Real.toLargeInt`.  It would incorrectly raise `Option`
  instead of `Overflow` in the case when the real was not an `INF`, but rounding
  produced an `INF`.
  ** Fixed bugs in `Date.{fmt,fromString,scan,toString}`.  They incorrectly
  allowed a space for the first character in the day of the month.

* 2004-08-18
  ** Changed `MLton.{Thread,Signal,World}` to distinguish between implicitly and
  explicitly paused threads.

* 2004-07-28
  ** Added support for programming in the large using the ML Basis system.

* 2004-07-11
  ** Fixed bugs in `ListPair.*Eq` functions, which incorrectly raised the
  `UnequalLengths` exception.

* 2004-07-01
  ** Added `val MLton.Exn.addExnMessager: (exn -> string option) -> unit`.

* 2004-06-23
  ** Runtime system options that take memory sizes now accept a "`g`" suffix
  indicating gigabytes.  They also now take a real instead of an integer,
  e.g. `fixed-heap 0.5g`.  They also now accept uppercase, e.g. `150M`.

* 2004-06-12
  ** Added support for OpenBSD.

* 2004-06-10
  ** Added support for FreeBSD 5.x.

* 2004-05-28
  ** Deprecated the `-native` flag.  Instead, use the new flag `-codegen
  {native|bytecode|C}`.  This is in anticipation of adding a bytecode compiler.

* 2004-05-26
  ** Fixed a front-end bug that could cause cascading error to print a very
  large and unreadable internal bug message of the form `"datatype ... realized
  with scheme Unknown"`.

* 2004-05-17
  ** Automatically restart functions in the Basis Library that correspond
  directly to interruptable system calls.

* 2004-05-13
  ** Added `-profile count`, for dynamic counts of function calls and branches.
  ** Equate the types `Posix.Signal.signal` and `Unix.signal`.

* 2004-05-11
  ** Fixed a bug with `-basis 1997` that would cause type errors due to
  differences between types in the MLton structure and types in the rest of the
  basis library.

* 2004-05-01
  ** Fixed a bug with sharing constraints in signatures that would sometimes
  mistakenly treat two structures as identical when they shouldn't have been.
  This would cause some programs to be mistakenly rejected.

* 2004-04-30
  ** Added `MLton.Signal.{handled,restart}`.

* 2004-04-23
  ** Added `Timer.checkCPUTimes`, and updated the `Timer` structure to match the
    latest basis spec.  Also fixed `totalCPUTimer` and `totalRealTimer`, which
    were wrong.

* 2004-04-13
  ** Added `MLton.Signal.Mask.{getBlocked,isMember}`.

* 2004-04-12
  ** Fix bug that mistakenly generalized variable types containing unknown types
  when matching against a signature.
  ** Reasonable front-end error message when unification causes recursive
  (circular) type.

* 2004-04-03
  ** Fixed bug in sharing constraints so that `sharing A = B = C` means that all
  pairs `A = B`, `A = C`, `B = C` are shared, not just `A = B` and `B = C`.
  This matters in some situations.

* 2004-03-20
  ** Fixed `Time.now` which was treating microseconds as nanoseconds.

* 2004-03-14
  ** Fixed SSA optimizer bug that could cause the error `"<type> has no
  tyconInfo property"`.

* 2004-03-11
  ** Fixed `Time.fromReal` to raise `Time`, not `Overflow`, on unrepresentable
  times.

* 2004-03-04
  ** Added structures `Word2`, `Word3`, ..., `Word31`.

* 2004-03-03
  ** Added structures `Int2`, `Int3`, ..., `Int31`.
  ** Fixed bug in elaboration of `and` with signatures, structures, and functors
  so that it now evaluates all right-hand sides before binding any left-hand
  sides.


== Version 20040227

Here are the changes from version 20030716 to 20040227.

=== Summary

* The front end now follows the Definition of SML and produces readable error
messages.
* Added support for NetBSD.
* Basis library changes tracking revisions to the specification.
* Added structures: `Int64`, `Real32`, `Word64`.
* File positions use `Int64`.
* Major improvements to `-show-basis`, which now displays the basis in a very
readable way with full type information.
* Command-line switch changes.
   ** Deprecated: `-basis`.
   ** Removed: `-lib-search`, `-link`, `-may-load-world`, `-static`.
   ** Added: `-link-opt`, `-runtime`, `-sequence-unit`, `-show-def-use`,
   `-stop tc`, `-warn-match`, `-warn-unused`.
   ** Changed: `-export-header`, `-show-basis`, `-show-basis-used`.
   ** Renamed: `-host` to `-target`.
* FFI changes.
   ** Renamed `_ffi` as `_import`.
   ** Added `cdecl` and `stdcall` attributes to `_import` and `_export`
   expressions.
* MLton structure changes.
   ** Added: Pointer.
   ** Removed: Ptrace.
   ** Changed: `Finalizable`, `IntInf`, `Platform`, `Random`, `Signal`, `Word`.

=== Details

* 2004-02-16
  ** Changed `-export-header`, `-show-basis`, `-show-basis-used` to take a file
  name argument, and they no longer force compilation to halt.
  ** Added `-show-def-use` and `-warn-unused`, which deal with def-use
  information.

* 2004-02-13
  ** Added flag `-sequence-unit`, which imposes the constraint that in the
  sequence expression `(e1; e2)`, `e1` must be of type `unit`.

* 2004-02-10
  ** Lots of changes to `MLton.Signal`: name changes, removal of superfluous
  functions, additional functions.

* 2004-02-09
  ** Extended `-show-basis` so that when used with an input program, it shows
  the basis defined by the input program.
  ** Added `stop` runtime argument.
  ** Made `-call-graph {false|true}` an option to `mlprof` that determines
  whether or not a call graph file is written.

* 2004-01-20
  ** Fixed a bug in `IEEEReal.{fromString,scan}`, which would improperly return
  `INF` instead of `ZERO` for things like `"0.0000e123456789012345"`.
  ** Fixed a bug in `Real.{fromDecimal,fromString,scan}`, which didn't return an
  appropriately signed zero for `~0.0`.
  ** Fixed a bug in `Real.{toDecimal,fmt}`, which didn't correctly handle
  `~0.0`.
  ** Report a compile-time error on unrepresentable real constants.

* 2004-01-05
  ** Removed option `-may-load-world`.  You can now use `-runtime no-load-world`
  instead.
  ** Removed option `-static`.  You can now use `-link-opt -static` instead.
  ** Changed `MLton.IntInf.size` to return 0 instead of 1 on small ints.

* 2003-12-28
  ** Fixed horrible bug in `MLton.Random.alphaNumString` that caused it to
  return 0 for all characters beyond position 11.

* 2003-12-17
  ** Removed `-basis` as a normal flag.  It is still available as an expert
  flag, but its use is deprecated.  It will almost certainly disappear after the
  next release.

* 2003-12-10
  ** Allow multiple `@MLton --` runtime args in sequnce.  This makes it easier
  for scripts to prefix `@MLton` args without having to splice them with other
  ones.

* 2003-12-04
  ** Added support for files larger than 2G.  This included changing
  `Position` from `Int32` to `Int64`.

* 2003-12-01
  ** Added `structure MLton.Pointer`, which includes a `type t` for pointers
  (memory addresses, not SML heap pointers) and operations for loading from and
  storing to memory.

* 2003-11-03
  ** Fixed `Timer.checkGCTime` so that only the GC user time is included, not GC
  system time.

* 2003-10-13
  ** Added `-warn-match` to control display nonexhaustive and redundant
  match warnings.
  ** Fixed space leak in `StreamIO` causing the entire stream to be retained.
  Thanks to Jared Showalter for the bug report and fix.

* 2003-10-10
  ** Added `-stop tc` switch to stop after type checking.

* 2003-09-25
  ** Fixed `Posix.IO.getfl`, which had mistakenly called `fcntl` with `F_GETFD`
  instead of `F_GETFL`.
  ** Tracking basis library changes:
    *** `Socket` module datagram functions no longer return amount written,
    since they always write the entire amount or fail.  So,
    `send{Arr,Vec}To{,'}` now return `unit` instead of `int`.
    *** Added nonblocking versions of all the send and recv functions, as well
    as accept and connect.  So, we now have: `acceptNB`, `connectNB`,
    `recv{Arr,Vec}{,From}NB{,'}`, `send{Arr,Vec}{,To}NB{,'}`.

* 2003-09-24
  ** Tracking basis library changes:
    *** `TextIO.inputLine` now returns a `string option`.
    *** Slices used in `Byte`, `PRIM_IO`, `PrimIO`, `Posix.IO`, `StreamIO`.
    *** `Posix.IO.readVec` raises `Size`, not `Subscript`, with negative
    argument.

* 2003-09-22
  ** Fixed `Real.toManExp` so that the mantissa is in [0.5, 1), not [1, 2).  The
  spec says that 1.0 <= man * radix < radix, which since radix is 2, implies
  that the mantissa is in [0.5, 1).
  ** Added `Time.{from,to}Nanoseconds`.

* 2003-09-11
  ** Added `Real.realRound`.
  ** Added `Char{Array,Vector}Slice` to `Text`.

* 2003-09-11
  ** `OS.IO.poll` and `Socket.select` now raise errors on negative timeouts.
  ** `Time.time` is now implemented using `IntInf` instead of `Int`, which means
  that a much larger range of time values is representable.

* 2003-09-10
  ** `Word64` is now there.

* 2003-09-09
  ** Replaced `Pack32{Big,Little}` with `PackWord32{Big,Little}`.
  ** Fixed bug in `OS.FileSys.fullPath`, which mistakenly stopped as soon as it
  hit a symbolic link.

* 2003-09-08
  ** Fixed `@MLton max-heap`, which was mistakenly ignored.  Cleaned up `@MLton
  fixed-heap`.  Both `fixed-heap` and `max-heap` can use copying or mark-compact
  collection.

* 2003-09-06
  ** `Int64` is completely there.
  ** Fixed `OS.FileSys.tmpName` so that it creates the file, and doesn't use
  `tmpnam`.  This eliminates an annoying linker warning message.

* 2003-09-05
  ** Added structures `{LargeInt,LargeReal,LargeWord,Word}{Array,Array2,ArraySlice,Vector,VectorSlice}`
  ** Fixed bug in `Real.toDecimal`, which return class `NORMAL` for subnormals.
  ** Fixed bug in `Real.toLargeInt`, which didn't return as precise an integer
  as possible.

* 2003-09-03
  ** Lots of fixes to `REAL` functions.
    *** `Real32` is now completely in place, except for `Real32.nextAfter` on
    SunOS.
    *** Fixed `Real.Math.exp` on x86 to return the right value when applied to
    `posInf` and `negInf`.
    *** Changed `Real.Math.{cos,sin,tan}` on x86 to always use a call to the C
    math library instead of using the x86 instruction.  This eliminates some
    anomalies between compiling `-native false` and `-native true`.
    *** Change `Real.Math.pow` to handle exceptional cases in the SML code.
    *** Fixed `Real.signBit` on Sparcs.

* 2003-08-28
  ** Fixed `PackReal{,64}Little` to work correctly on Sparc.
  ** Added `PackReal{,64}Big`, `PackReal32{Big,Little}`.
  ** Added `-runtime` switch, which passes arguments to the runtime via
  `@MLton`.  These arguments are processed before command line switches.
  ** Eliminated MLton switch `-may-load-world`.  Can use `-runtime` combined
  with new runtime switch `-no-load-world` to disable load world in an
  executable.

* 2003-08-26
  ** Changed `-host` to `-target`.
  ** Split `MLton.Platform.{arch,os}` into `MLton.Platform.{Arch,OS}.t`.

* 2003-08-21
  ** Fixed bug in C codegen that would cause undefined references to
  `Real_{fetch,move,store}` when compiling on Sparcs with `-align 4`.

* 2003-08-17
  ** Eliminated `-link` and `-lib-search`, which are no longer needed.
  Eliminated support for passing `-l*`, `-L*`, and `*.a` on the command line.
  Use `-link-opt` instead.

* 2003-08-16
  ** Added `-link-opt`, for passing options to `gcc` when linking.

* 2003-07-19
  ** Renamed `_ffi` as `_import`.  The old `_ffi` will remain for a while, but
  is deprecated and should be replaced with `_import`.
  ** Added attributes to `_export` and `_import`.  For now, the only attributes
  are `cdecl` and `stdcall`.


== Version 20030716

Here are the changes from version 20030711 to 20030716.

== Summary

* Fixed several serious bugs with the 20030711 release.

== Details

* 2003-07-15
  ** Fixed bug that caused a segfault when attempting to create an
  array that was too large, e.g

      1 + Array.sub (Array.tabulate (valOf Int.maxInt, fn i => i), 0)

  ** mlton now checks the command line arguments following the file to compile
  that are passed to the linker to make sure they are reasonable.

* 2003-07-14
  ** Fixed packaging for Cygwin and Sparc to include `libgmp.a`.
  ** Eliminated bootstrap target.  The `Makefile` automatically determines
  whether to bootstrap or not.
  ** Fixed XML type checker bug that could cause error: `"empty tyvars in
  PolyVal dec"`.

* 2003-07-12
  ** Turned off `FORCE_GENERATIONAL` in gc.  It had been set, which caused the
  gc to always use generational collection.  This could seriously slow apps down
  that don't need it.


== Version 20030711

Here are the changes from version 20030312 to 20030711.

=== Summary

* Added support for Sparc/SunOS using the C code generator.
* Completed the basis library implementation.  At this point, the only missing
basis library function is `use`.
* Added `_export`, which allows one to call SML functions from C.
* Added weak pointers (via `MLton.Weak`) and finalization (via
`MLton.Finalizable`).
* Added new integer modules: `Int8`, `Int16`.
* Better profiling call graphs
* Fixed conversions between reals and their decimal representations to be
correct using the gdtoa library.

=== Details

* 2003-07-07
  ** Profiling improvements:
    *** Eliminated `mlton -profile-split`.  Added `mlprof -split`.  Now the
    profiling infrastructure keeps track of the splits and allows one to decide
    which splits to make (if any) when `mlprof` is run, which is much better
    than having to decide at compile time.
    *** Changed `mlprof -graph` to `mlprof -keep`, and changed the behavior so
    that `-keep` also controls which functions are displayed in the table.
    *** Eliminated `mlprof -ignore`: it's behavior is now subsumed by `-keep`,
    whose meaning has changed to be more like -ignore on nodes that are not
    kept.
  ** When calling `gcc` for linking, put `-link` args in same order as they
  appeared on the MLton command line (they used to be reversed).

* 2003-07-03
  ** Making `OS.Process.{atExit,exit}` conform to the basis library spec in that
  exceptions raised during cleaners are caught and ignored.  Also, calls to
  `exit` from cleaners cause the rest of cleaners to run.

* 2003-07-02
  ** Fixed bug with negative `IntInf` constants that could cause compile time
  error message: `"x86Translate.translateChunk ... strange Offset: base: ..."`
  ** Changed argument type of `MLton.IntInf.Small` from `word` to `int`.
  ** Added fix to profiling so that the `mlmon.out` file is written even when
  the program terminates due to running out of memory.

* 2003-06-25
  ** Added `{Int{8,16},Word8}{,Array,ArraySlice,Vector,VectorSlice,Array2}`
  structures.

* 2003-06-25
  ** Fixed bug in `IntInf.sign`, which returned the wrong value for zero.

* 2003-06-24
  ** Added `_export`, for calling from C to SML.

* 2003-06-18
  ** Regularization of options:
    *** `-diag` --> `-diag-pass`
    *** `-drop-pass` takes a regexp

* 2003-06-06
  ** Fixed bug in `OS.IO.poll` that caused it to return the input event types
  polled for instead of what was actually available.

* 2003-06-04
  ** Fixed bug in KnownCase SSA optimization that could case incorrect results
  in compiled programs.

* 2003-06-03
  ** Fixed bug in SSA optimizer that could cause the error message:

      Type error: Type.equals
      {from = char vector, to = unit vector}
      Type error: analyze raised exception loopStatement: ...
      unhandled exception: TypeError

* 2003-06-02
  ** Fixed `Real.rem` to work correctly on `inf`-s and `nan`-s.
  ** Fixed bug in profiling that caused the function name to be omitted on
  functions defined by `val rec`.

* 2003-05-31
  ** `Fixed Real.{fmt,fromString,scan,toString}` to match the basis library
  spec.
  ** Added `IEEEReal.{fromString,scan}`.
  ** Added `Real.{from,to}Decimal`.

* 2003-05-25
  ** Added `Real.nextAfter`.
  ** Added `OS.Path.{from,to}UnixPath`, which are the identity function on Unix.

* 2003-05-20
  ** Added type `MLton.pointer`, the type of C pointers, for use with the FFI.

* 2003-05-18
  ** Fixed two bugs in type inference that could cause the compiler to raise the
  `TypeError` exception, along with a lot of XML IL.  The `type-check.sml`
  regression contains simple examples of what failed.
  ** Fixed a bug in the simplifier that could cause the message: `"shrinker
  raised Prim.apply raised assertion failure: SmallIntInf.fromWord"`.

* 2003-05-15
  ** Fixed bug in `Real.class` introduced on 04-28 that cause many regression
  failures with reals when using newer `gcc`-s.
  ** Replaced `MLton.Finalize` with `MLton.Finalizable`, which has a more robust
  approach to finalization.

* 2003-05-13
  ** Fixed bug in `MLton.FFI` on Cygwin that caused `Thread_returnToC` to be
  undefined.

* 2003-05-12
  ** Added support for finalization with `MLton.Finalize`.

* 2003-05-09
  ** Fixed a runtime system bug that could cause a segfault.  This bug would
  happen after a GC during heap resizing when copying a heap, if the heap was
  allocated at a very low (<10M) address.  The bug actually showed up on a
  Cygwin system.

* 2003-05-08
  ** Fixed bug in `HashType` that raised `"Vector.forall2"` when the arity of a
  type constructor is changed by `SimplifyTypes`, but a newly constructed type
  has the same hash value.

* 2003-05-02
  ** Switched over to new layered IO implementation, which completes the
  implementation of the `BinIO` and `TextIO` modules.

* 2003-04-28
  ** Fixed bug that caused an assertion failure when generating a jump table for
  a case dispatch on a non-word sized index with non-zero lower bound on the
  range.

* 2003-04-24
  ** Added `-align {4|8}`, which controls alignment of objects.  With `-align
  8`, memory accesses to doubles are guaranteed to be aligned mod 8, and so
  don't need special routines to load or store.

* 2003-04-22
  ** Fixed bug that caused a total failure of time profiling with `-native
  false`.  The bug was introduced with the C codegen improvements that split the
  C into multiple files.  Now, the C codegen declares all profile labels used in
  each file so that they are global symbols.

* 2003-04-18
  ** Added `MLton.Weak`, which supports weak pointers.

* 2003-04-10
  ** Replaced the basis library's `MLton.hostType` with `MLton.Platform.arch`
  and `MLton.Platform.os`.

* 2003-04
  ** Added support for SPARC/SunOS using the C codegen.

* 2003-03-25
  ** Added `MLton.FFI`, which allows callbacks to SML from C.

* 2003-03-21
  ** Fixed `mlprof` so that the default `-graph arg` for data from
  `-profile-stack true` is `(thresh-stack x)`, not `(thresh x)`.


== Version 20030312

Here are the changes from version 20020923 to 20030312.

=== Summary

* Added source-level profiling of both time and allocation.
* Updated basis library to 2002 specification.  To obtain the old
library, compile with `-basis 1997`.
* Added many modules to basis library:
 ** `BinPrimIO`, `GenericSock`, `ImperativeIO`, `INetSock`, `NetHostDB`,
 `NetProtDB`, `NetServDB`, `Socket`, `StreamIO`, `TextPrimIO`, `UnixSock`.
* Completed implementation of `IntInf` and `OS.IO`.

=== Details

* 2003-02-23
  ** Replaced `-profile-combine` wih `-profile-split`.

* 2003-02-11
  ** Regularization of options:
    *** `-l` --> `-link`
    *** `-L` --> `-lib-search`
    *** `-o` --> `-output`
    *** `-v` --> `-verbose`

* 2003-02-10
  ** Added option to `mlton`: `-profile-combine {false|true}`

* 2003-02-09
  ** Added options to `mlprof`: `-graph-title`, `-gray`, `-ignore`, `-mlmon`,
  `-tolerant`.

* 2002-11 - 2003-01
  ** Added source-level allocation and time profiling.  This includes the new
  options to mlton: `-profile` and `-profile-stack`.

* 2002-12-28
  ** Added `NetHostDB`, `NetProtDB`, `NetServDB` structures.
  ** Added `Socket`, `GenericSock`, `INetSock`, `UnixSock` structures.

* 2002-12-19
  ** Fixed bug in signal check insertion that could cause some signals to be
  missed.  The fix was to add a signal check on entry to each function in
  addition to at each loop header.

* 2002-12-10
  ** Fixed bug in runtime that might cause the message `"Unable to set
  cardMapForMutator"`.

* 2002-11-23
  ** Added support for the latest Basis Library specification.
  ** Added option `-basis` to choose Basis Library version.  Currently available
    basis libraries are `2002`, `2002-strict`, `1997`, and `none`.
  ** Added `IntInf.{orb,xorb,andb,notb,<<,~>>}` values.
  ** Added `OS.IO.{poll_desc,poll_info}` types.
  ** Added `OS.IO.{pollDesc,pollToIODesc,infoToPollDesc,Poll}` values.
  ** Added `OS.IO.{pollIn,pollOut,pollPri,poll,isIn,isOut,isPri}` values.
  ** Added `BinPrimIO`, `TextPrimIO` structures.
  ** Added `StreamIO`, `ImperativeIO` functors.

* 2002-11-22
  ** Fixed bug that caused time profiling to fail (with a segfault) when
  resuming a saved world.

* 2002-11-07
  ** Fixed bug in `MLton.eq` that could arise when using `eq` on functions.

* 2002-11-05
  ** Improvements to polymorphic equality.  Equality on IntInfs, vectors, and
  dataypes all do an `eq` test first before a more expensive comparison.

* 2002-11-01
  ** Added allocation profiling.  Now, can compile with either `-profile alloc`
  or `-profile time`.  Renamed `MLton.Profile` as `MLton.ProfileTime`.  Added
  `MLton.ProfileAlloc`.  Cleaned up and changed most `mlprof` option names.

* 2002-10-31
  ** Eliminated `MLton.debug`.
  ** Fixed bug in the optimizer that affected `IntInf.fmt`.  The optimizer had
  been always using base 10, instead of the passed in radix.

* 2002-10-22
  ** Fixed `Real.toManExp` so that the mantissa is in [1, 2), not [0.5, 1).
  ** Added `Real.fromLargeInt`, `Real.toLargeInt`.
  ** Fixed `Real.split`, which would return an incorrect whole part due to the
  underlying primitive, `Real_modf`, being treated as functional instead of
  side-effecting.

* 2002-09-30
  ** Fixed `rpath` problem with packaging.  All executables in packages
  previously made had included a setting for `RPATH`.


== Version 20020923

Here are the changes from version 20020410 to 20020923.

=== Summary

* MLton now runs on FreeBSD.
* Major runtime system improvements.  The runtime now implements mark-compact
and generational collection, in addition to the copying collection that was
there before.  It automatically switches between the the collection strategies
to improve performance and to try to avoid paging.
* Performance when compiling `-exn-history true` has been improved.
* Added `IntInf.log2`, `MLton.GC.pack`, `MLton.GC.unpack`.
* Fixed bug in load world that could cause "sread failed" on Cygwin.
* Fixed optimizer bug that could cause `"no analyze var value property"`
message.

=== Details

* 2002-09
  ** Integrated Sam Rushing's changes to port MLton to FreeBSD.

* 2002-08-25
  ** Changed the implementation of exception history to be completely
  functional.  Now, the extra field in exceptions (when compiling `-exn-history
  true`) is a `string list` instead of a `string list ref`, and `raise` conses a
  new exception with a new element in the list instead of assigning to the list.
  This changes the semantics of exception history (for the better) on some
  programs. See `regression/exnHistory3.sml` for an example.  It also
  significantly improves performance when compiling `-exn-history true`.

* 2002-07 and 2002-08
  ** Added generational GC, and code to the runtime that automatically turns it
  on and off.

* 2002-08-20
  ** Fixed SSA optimizer bug that could cause the following error message: `"x_0
  has no analyze var value property"`

* 2002-07-28
  ** Added `MLton.GC.{pack,unpack}`.  `pack` shrinks the heap so that other
  processes can use the RAM, and its dual, `unpack`, resizes the heap to the
  desired size.

* 2002-06 and 2002-07
  ** Added mark compact GC.
  ** Changed array layout so that arrays have three, not two header words.  The
  new word is a counter word that preceeds the array length and header.
  ** Changed all header words to be indices into an array of object descriptors.

* 2002-06-27
  ** Added patches from Michael Neumann to port runtime to FreeBSD 4.5.

* 2002-06-05
  ** Output file and intermediate file are now saved in the current directory
  instead of in the directory containing the input file.

* 2002-05-31
  ** Fixed bug in overloading of `/` so that the following now type checks:

      fun f (x, y) = x + y / y

* 2002-04-26
  ** Added back `max-heap` runtime option.

* 2002-04-25
  ** Fixed load/save world so that they use binary mode.  This should fix the
  `sread failed` problem that Byron Hale saw on Cygwin that caused `mlton` to
  fail to start.
  ** Added `IntInf.log2`.
  ** Changed call to linker to use `libgmp.a` (if it exists) instead of
  `libgmp.so`.  This is because the linker adds a dependency to a shared library
  even if there are no references to it

* 2002-04-23
  ** Rewrote heap resizing code.  This fixed bug that was triggered with large
  heaps and could cause a spurious out of memory error.
  ** Removed GnuMP from MLton sources (again :-).


== Version 20020410

Here are the changes from version 20011006 to version 20020410.

=== Details

* 2002-03-28
  ** Added BinIO.

* 2002-03-27
  ** Regularization of options
    *** `-g` --> `-degug {false|true}`
    *** `-h n` --> `-fixed-heap n`
    *** `-p` --> `-profile {false|true}`

* 2002-03-22
  ** Set up the stubs so that MLton can be compiled in the standard basis
  library, with no `MLton` structure.  Thus it is now easy to compile MLton with
  an older (or newer) version of itself that has a different `MLton` structure.

* 2002-03-17
  ** Added `MLton.Process.{spawn,spawne,spawnp}`, which use primitives when
  running on Cygwin and fork/exec when running on Linux.

* 2002-02 - 2002-03
  ** Added the ability to cross-compile to Cygwin/Windows.

* 2002-02-24
  ** Added GnuMP back for use with Cygwin.

* 2002-02-10
  ** Reworked object header words so that `Array.maxLen = valOf Int.maxInt`.
  Also fixed a long-standing minor bug in MLton, where `Array.array
  (Array.maxLen, ...)` would raise `Size` instead of attempting to allocate the
  array.  It was an off-by-one error in the meaning of `Array.maxLen`.

* 2002-02-08
  ** Modifications to runtime to behave better in situations where the amount of
  live data is a signifant fraction of the amount of RAM, based on code from
  PolySpace.  MLton executables by default can now use more than the available
  amount of RAM.  Executables will still respect the `max-heap` runtime arg if
  it is set.

* 2002-02-04
  ** Improvements to runtime so that it fails to get space, it attempts to get
  less space instead of failing.  Based on PolySpace's modifications.
  ** Added `MLton.eq`.

* 2002-02-03
  ** Added `MLton.IntInf.gcd`.
  ** Removed GnuMP from MLton sources.  We now link with `/usr/lib/libgmp.a`.
  ** Added `TextIO.getPosOut`.
  ** Renamed type `MLton.Itimer.which` to `MLton.Itimer.t` and
  `MLton.Itimer.whichSignal` to `MLton.Itimer.signal`.
  ** Added `-coalesce` flag, for use with the C backend.

* 2002-01-26
  ** Added `-show-basis-used`, which prints out the parts of the basis library
  that the input program uses.
  ** Changed several other flags (`-print-at-fun-entry`, `-show-basis`,
  `-static`) to follow the `{false|true}` convention.

* 2002-01-22
  ** Improved `MLton.profile` so that multiple profile arrays can exist
  simultaneously and so that the current one being used can be set from the SML
  side.

* 2002-01-18
  ** The Machine IL has been replaced with an RSSA (representation explicit SSA)
  IL and an improved Machine IL.

* 2002-01-16
  ** Added KnownCase SSA optimization

* 2002-01-14
  ** Added rudimentary profiling control from with a MLton compile program via
  the `MLton.Profile` structure.

* 2002-01-09
  ** Fixed bug in match compiler that caused case expressions on datatypes with
  redundant cases to be compiled incorrectly.

* 2002-01-08
  ** Added redundant tuple construction elimination to SSA shrinker.
  ** Improved Flatten SSA optimization.

* 2001-12-06
  ** Changed the interface for `MLton.Signal`.  There is no longer a separate
  `Handler` substructure.  This was done so that programs that just use
  `default` and `ignore` signal handlers don't bring in the entire thread
  mechanism.

* 2001-12-05
  ** Added LocalRef elimination SSA optimization.

* 2001-11-19
  ** The CPS IL has been replaced with an SSA (static-single assignment) IL.
  All of the optimizations have been ported from CPS to SSA.

* 2001-10-24
  ** Fixed bug in `Thread_atomicEnd` -- `limit` was mistakenly set to `base`
  instead of to 0.  This caused assertion failures when for executables compiled
  `-g` because `GC_enter` didn't reset `limit`.
  ** Fixed bug in register allocation of byte registers.

* 2001-10-23
  ** Added `-D` option to `cmcat` for preprocessor defines.  Thanks to Anoq for
  sending the code.
  ** Changed limit check insertion so that limit checks are only coalesced
  within a single basic block -- not across blocks.  This slows many benchmarks
  down, but is needed to fix a bug in the way that limit checks were coalesced
  across blocks.  Hopefully we will figure out a better fix soon.

* 2001-10-18
  ** Fixed type inference of flexrecord so that it now follows the Definition.
  Many programs containing flexrecords were incorrectly rejected.  Added many
  new tests to regression/flexrecord.sml.
  ** Changed the behavior of `-keep dot` combined with `-keep pass` for SSA
  passes.  Dot files are now saved for the program before and after, instead of
  just after.

* 2001-10-11
  ** Fixed a bug in the type inference that caused type variables to be
  mistakenly generalized.  The bug was exposed in Norman Ramsey's `sled.sml`.
  Added a test to `regression/flexrecord.sml` to catch the problem.


== Version 20011006

Here are the changes from version 20010806 to version 20011006.

=== Summary

* Added `MLton.Exn.history`, which is similar to `SMLofNJ.exnHistory`.
* Support for `#line` directives of the form `(*#line line.col "file"*)`.
* Performance improvements in native codegenerator.
* Bug fixes in front-end, optimizer, register allocator,
`Real.{maxFinite,minPos,toManExp}`, and in heap save and restore.

=== Details

* 2001-10-05
  ** Fixed a bug in polymorphic layered patterns, like

      val 'a a as b = []

  These would always fail due to the variable `a` not being handled correctly.
  ** Fixed the syntax of `val rec` so that a pattern is allowed on the left-hand
  side of the `=`.  Thus, we used to reject, but now accept, the following.

      val rec a as b as c = fn _ => ()
      val rec a : unit -> unit : unit -> unit = fn () => ()

  Thanks again to Andreas Rossberg's test files.  This is now tested for in
  `valrec.sml`.
  ** Fixed dynamic semantics of `val rec` so that if `val rec` is used to
  override constructor status, then at run time, the `Bind` exception is raised
  as per rule 126 of the Definition.  So, for example, the following program
  type checks and compiles, but raises `Bind` at run time.

      val rec NONE = fn () => ()
      val _ = NONE ()

  Again, this is checked in `valrec.sml`.
  ** Added `\r\n` to ml.lex so that Windows style newlines are acceptable in
  input files.

* 2001-10-04
  ** Fixed bug in the implementation of `open` declarations, which in the case
  of `open A B` had opened `A` and then looked up `B` in the resulting
  environment.  The correct behaviour (see rule 22 of the Definition) is to
  lookup each _longstrid_ in the current environment, and then open them all in
  sequence.  This is now checked for in the `open.sml` regression test.  Thanks
  to Andreas Rossberg for pointing this bug out.
  ** Fixed bug that caused tyvars of length 1 (i.e. `'`) to be rejected.  This
  is now checked in the `id.sml` regression test.  Again, thanks to Andreas
  Rossberg for the test.

* 2001-10-02
  ** Fixed bugs in `Real.toManExp` (which always returned the wrong result
  because the call to `frexp` was not treated as side-effecting by the
  optimizer) and in `Real.minPos`, which was zero because of a mistake with
  extra precision bits.

* 2001-10-01
  ** Added `MLton.Exn.history`.
  ** Fixed register allocation bug with `fucom` instruction.  Was allowing
  `fucomp` when the first source was not removable.
  ** Changed `Real.isFinite` to use the C `math.h` `finite` function.  This
  fixed the nontermination bug which occurred in any program that used
  `Real.maxFinite`.

* 2001-09-22
  ** Bug fixes found from Ramsey's `lrtl` in `contify.fun` and
  `unused-args.fun`, both of which caused compile-time exceptions to be raised.

* 2001-09-21
  ** Fixed `MLton.World.{load,save}` so that the saved world does not store the
  max heap size.  Instead, the max heap size is computed upon load world in
  exactly the same way as at program startup.  This fixes a long-standing (but
  only recently noticed) problem in which `mlton` (which uses a saved world)
  would attempt to use as much memory as was on the machine used to build
  `world.mlton`.

* 2001-08-29
  ** Overlow checking is now on by default in the C backend.  This is a huge
  performance hit, but who cares, since we never use the C backend except for
  testing anyways.

* 2001-08-22
  ** Added support for #line directives of the form

      (*#line line.col "file"*)

  These directives only affect error messages produced by the parser and
  elaborator.

* 2001-08-17
  ** Fixed bug in RemoveUnused optimzation that caused the following program to
  fail to compile.

      fun f l = case l of [] => f l | _ :: l => f l
      val _ = f [13]

* 2001-08-14
  ** New x86-codegen infrastructure.
    *** support for tracking liveness of stack slots and carrying them in
    registers across basic blocks
    *** more specific `Entry` and `Transfer` datatypes to make calling convention
    distinctions more explicit
    *** new heuristic for carrying values in registers across basic blocks (look
    Ma, no Overflows!)
    *** new "predict" model for generating register allocation hints
    *** additional bug fixes

* 2001-08-07
  ** `MLton.Socket.shutdownWrite` flushes the outstream.


== Version 20010806

Here are the changes from version 20010706 to version 20010806.

=== Summary

* `Word.andb (w, 0xFF)` now works correctly
* `MLton.Rusage.rusage` has a patch to work around a linux kernel bug
* Programs of the form `_exp_ ; _program_` are now accepted
* Added the `MLton.Rlimit` structure
* Added the `-keep dot` flag, which produces call graphs, intraprocedural
control-flow graphs, and dominator trees

=== Details

* 2001-08-06
  ** Added simple CommonBlock elimination CPS optimization.

* 2001-08-02
  ** Took out `-keep il`.

* 2001-07-31
  ** Performance improvements to `TextIO.{input, output, output1}`.

* 2001-07-25
  ** Added RedundantTest elimination CPS optimization.

* 2001-07-21
  ** Added CommonSubexp elimination CPS optimization.

* 2001-07-20
  ** Bug fix to x86 codegen.  The `commuteBinALMD` peephole optimization would
  rewrite `mov 2,Y; add Y,Y` as `mov Y,Y; add 2,Y`.  Now the appropriate
  interference checks are made.
  ** Added intraprocedural unused argument removal.
  ** Added intraprocedural flattener.  This avoids some stupid tuple allocations
  in loops.  Decent speedup on a few benchmarks (`count-graphs`, `psdes-random`,
  `wc-scanStream`) and no noticeable slowdowns.
  ** Added `-keep dot` flag.

* 2001-07-17
  ** Modified grammar to properly handle `val rec`.  There were several problems.
    *** MLton had accepted `val rec 'a ...` instead of `val 'a rec ...`
    *** MLton had not accepted `val x = 13 and rec f = fn () => ()`
    *** MLton had not accepted `val rec rec f = fn () => ()`
    *** MLton had not accepted `val rec f = fn () => () and rec g = fn () => ()`

* 2001-07-16
  ** Workaround for Linux kernel bug that can cause `getrusage` to return a wrong
  system time value (low by one second).  See `fixedGetrusage` in `gc.c`.
  ** Bug fix to x86 codegen.  The register allocator could get confused when
  doing comparisons of floating point numbers and use the wrong operand.  The
  bug seems to have never been detected because it only happens when both of the
  operands are already on the floating point stack, which is rare, since one is
  almost always in memory since we don't carry floating point values in the
  stack across basic blocks.
  ** Added production to the grammar on page 58 of the Definition that had been
  missing from MLton since day one.

      program ::= exp ; <program>

  Also updated docs to reflect change.
  ** Modified grammar to accept the empty program.
  ** Added `-type-check` expert flag to turn on type checking in ILs.

* 2001-07-15
  ** Bug fix to the algebraic simplifier.  It had been rewriting
  `Word32.andb (w, 0wxFF)` to `w` instead of
  `Word32.andb (w, 0wxFFFFFFFF)` to `w`.

* 2001-07-13
  ** Improved CPS shrinker so that `if`-tests where the `then` and `else` branch
  jump to the same label is turned into a direct jump.
  ** Improved CPS shrinker (`Prim.apply`) to handle constructors
    *** `A = A` --> `true`
    *** `A = B` --> `false`
    *** `A x` = `B y` --> `false`
  ** Rewrote a lot of loops in the basis library to use inequalities instead of
  equality for the loop termination test so that the (forthcoming) overflow
  detection elimination will work on the loop index variable.

* 2001-07-11
  ** Fixed minor bugs in `Array2.{array,tabulate}`, `Substring.{slice}` that
  caused the `Overflow` exception to be raised instead of `Size` or `Subscript`
  ** Fixed bug in `Pack32Big.update` that caused the wrong location to be updated.
  ** Fixed several bugs in `Pack32{Big,Little}.{subArr,subVec,update}` that
  caused `Overflow` to be raised instead of `Subscript`.  Also, improved the
  implementation so that bounds checking only occurs once per call (instead of
  four times, which was sometimes happening.
  ** Fixed bugs in `Time.{toMilliseconds,toMicroseconds}` that could cause a
  spurious `Overflow` exception.
  ** Fixed bugs in `Time.{fromMilliseconds,fromMicroseconds}` that could cause a
  spurious `Time` exception.
  ** Improved `Pack32.sub*` by reordering the `orb`-s.
  ** Improved `{Int,IntInf}.mod` to increase chances of constant folding.
  ** Switched many uses of `+`, `-`, `*` in basis library to the non-overflow
  checked versions.  Modules changed were: `Array`, `Array2`, `Byte`, `Char`,
  `Int`, `IntInf`, `List`, `Pack32{Big,Little}`, `Util`, `String`, `StringCvt`,
  `Substring`, `TextIO`, `Time`, `Vector`.
  ** Added regression tests for `Array2`, `Int` (overflow checking), `Pack32`,
  `Substring`, `Time`.
  ** Changed CPS output so that it includes a dot graph for each CPS function.

* 2001-07-09
  ** Change `OS.Process.exit` so that it raises an exception if the exit status
  is not in [0, 256).
  ** Added `MLton.Rlimit` to provide access to `getrlimit` and `setrlimit`.


== Version 20010706

Here are the changes from the 20000906 version to the 20010706 version.

=== Summary

* Native X86 code generator (instead of using `gcc`)
* Significantly improved compile times
* Significantly improved run times for generated executables
* Many bug fixes
* Correct raising of the `Overflow` exception for integer arithmetic
* New modules in the `MLton` structure

=== Details

* 2001-07-06
  ** GC mods from Henry.  Mostly adding `inline` declarations.

* 2001-07-05
  ** Fixed several runtime bugs involving threads, critical sections, and
  signals.

* 2001-06-29
  ** Fixed performance bug in `cps/two-point-lattice.fun` that caused quadratic
  behavior.  This affects the raise-to-jump and useless analayses.  In
  particular, the useless analysis was blowing up when compiling `fxp`.

* 2001-06-27
  ** Henry improved `wordAlign` -- this sped up GC by 27% (during a self
  compile).

* 2001-06-20
  ** Moved `MLton.random` to `MLton.Random.rand` and added other stuff to
  `MLton.Random`
  ** Added `MLton.TextIO.mkstemp`.
  ** Made `Int.{div,quot}` respect the `-detect-overflow` switch.

* 2001-06-20
  ** Added `MLton.Syslog`.

* 2001-06-07
  ** Fixed bug in `MLton.Socket.accept` that was in the runtime implementation
  `Socket_accept`.  It did a `setsockopt SO_REUSEADDR` after the `accept`.  It
  should have been after the call to `socket` in `Socket_listen`.  Thanks to
  Doug Bagley for the fix.

* 2001-05-30
  ** Fixed bug in remove-unused that caused polymorphic equality to return
  `true` sometimes when constructors were never used in a pattern match.  For
  example, the following (in which `A` and `B` are not used as patterns):

      datatype t = A | B
      datatype u = C of t
      val _ = if C A = C B then raise Fail "bug" else ()

* 2001-03-27
  ** Fixed bug that caused all of the following to fail:
  `{LargeWord,Word,SysWord}.{toLargeInt,toLargeIntX,fromLargeInt}` The problem
  was the basis library file `integer/patch.sml` which fixed `Word32` but not
  the other structures that are the same.

* 2001-02-12
  ** Fixed bug in match compiler that caused it to spend a lot of extra time in
  deep patterns.  It still could be exponential however.  Hopefully this will
  get fixed in the release after next.  This bug could cause very slow compile
  times in some cases.  Anyways, this fix cut the `finish infer` time of a self
  compile down from 22 to under 4 seconds.  I.E. most of the time used to be
  spent due to this bug.

* 2001-02-06
  ** Fixed bug in frontend that caused the wrong file and line number to be
  reported with errors in functor bodys.

* 2001-01-03 - 2000-02-05
  ** Changes to CoreML, XML, SXML, and CPS ILs to replace lists by vectors in
  order to decrease space usage.

* 2001-01-16
  ** Fixed a bug in constant propagation where the length of vectors was not
  propagated properly.

* 2000-12-11 - 2001-01-03
  ** Major rewrite of elaborator to use a single hash table for each namespace
  instead of a hash table for every environment.

* 2000-12-20
  ** Fixed some bugs in the SML/NJ compatibility library,
  `src/lib/mlton-subs-in-smlnj`.

* 2000-12-08
  ** More careful removal of tracing code when compiling `MLton_debug=0`.  This
  cut down self compile data size by 100k and compile time by a few seconds.
  ** Added built in character and word cases propagated throughout all ILs.

* 2000-12-06
  ** Added max stack size information to `gc-summary`.

* 2000-12-05
  ** Added `src/benchmark`, which contains an SML program that benchmarks all of
  the SML compilers I have my hands on.  The script has lots of hardwired paths
  for now.

* 2000-12-04
  ** Fixed bug in `Posix.ProcEnv.environ,` which did not work correctly in a
  saved world (the original `environ` was saved).  In fact, it did not work at
  all because the ML primitive expected a constant and the C was a nullary
  function. This caused a segfault with any program using
  `Posix.ProcEnv.environ`.
  ** `Added MLton.ProcEnv.setenv`, since there doesn't seem to be any `setenv`
  in the basis library.

* 2000-11-29
  ** Changed backend so that it should no longer generate machine programs with
  `void` operands.
  ** Added `-detect-overflow` and `-safe` flags.

* 2000-11-27 - 2000-11-28
  ** Changes in many places to use `List.revMap` instead of `List.map` to cut
  down on allocation.

* 2000-11-21
  ** Added `MLton.Word.~` and `MLton.Word8.~` to the `MLton` structure.

* 2000-11-20
  ** Fixed a bug in the CPS shrinker that could cause a compile-time failure.
  It was maintaining occurrence counts incorrectly.

* 2000-11-15
  ** Fixed a (performance) bug in constant propagation that caused the hashing
  to be bad.
  ** Improved translation to XML so that the match compiler isn't called on
  tuple or if expressions.  This should speed up the translation and make the
  output smaller.
  ** Fixed a bug in the match compiler that caused it to not generate integer
  case statements.  This should speed up the mlyacc benchmark and the MLton
  front end.

* 2000-11-09
  ** Added `IntInf_equal` and `IntInf_compare` primitives.
  ** Took out the automatic `-keep c` when compiling `-g`.

* 2000-11-08
  ** Added a whole bunch of algebraic laws to the CPS shrinker, including some
  specifically targeted to `IntInf` primitives.

* 2000-11-03
  ** Improved implementation of properties so that sets don't allocate.
  ** Improved implementation of type homomorphism in type inference.  What was
  there before appears to have been a bug -- it didn't use the property on
  types.

* 2000-11-02
  ** Fixed timers used with `-v` option to use user + sys time.

* 2000-10-27
  ** Split the runtime basis library C files into many separate files so that
  only the needed code would be included by the linker.
  ** Fixed several bugs in the front end grammar and elaborator that caused type
  specifications to be handled incorrectly.  The following three programs used
  to be handled incorrectly, but are now handled correctly.

      signature S = sig type t and u = int end (* reject *)
      signature S = sig type t = int and u = t end (* accept *)
      signature S = sig eqtype t and u = int end (* reject *)

* 2000-10-25
  ** Changes to `main.sml` to run complete compiles with `-native` switch.

* 2000-10-24
  ** Removed defunctorizer.

* 2000-10-20
  ** Fixed bug in `cps-tree.fun` with `PrimExp.maySideEffect`.  This bug could
  cause `"no operand"` failures in the backend.
  ** Fixed bug in the runtime implementation of `MLton.size`.  The size for
  stack objects was using the `used` instead of `reserved`, and so was too low.

* 2000-10-19
  ** Replaced automatically generated dependencies in `src/runtime/Makefile`
  with hand generated ones.  Took out `make depend` from `src/Makefile`.  `make
  depend` was behaving really badly on RHAT 7.0.
  ** Tweaked compiler to shorten width of C output lines to work around bug in
  RHAT 7.0 `cpp` which silently truncates (very) long lines.
  ** Fixed bug in grammar that didn't allow `op` to occur in datatype and
  exception bindings, causing the following to fail

      datatype t = op T
      exception op E = op Fail

  ** Improved error messages in CM processor.  Fixed bug in CM Alias handling.

* 2000-10-18
  ** Fixed two bugs in the gc that did comparisons with `(s->limit -
  s->frontier)`, which of course doesn't work if `frontier` is beyond `limit`,
  since these are unsigned.  This could have caused segfaults, except that the
  mutator checks the `frontier` upon return from the GC.

* 2000-10-17
  ** Fixed bug in backend in the calculation of `maxFrameSize`.  It could be
  wrong (low) in some situations.
  ** Improved CPS inliner's estimate of function sizes.  The size of a function
  now takes into account other inlined functions that the function calls.  This
  also changed the meaning of the size argument to the `-inline` switch.  It now
  corresponds (roughly) to the product of the size of the function and the
  number of calls.  In general, it should be larger than before.

* 2000-10-13
  ** Made some calls to `Array.sub` unsafe in the implementation of `Array2`.
  ** Integrated Matthew's new x86 backend with floating point support.

* 2000-10-09
  ** Fixed CM file processor so that MLton works if it is run from a different
  directory than the main CM file.

* 2000-10-04
  ** Changed LimitCheck so it loops on the `frontier > limit` check.  This fixed
  a potential bug in threads caused when there is enough space available for a
  thread, `t`, before switching to another thread but not enough space when it
  resumes.  This could have caused a segfault.

* 2000-10-03
  ** More rewrites of `TextIO.StreamIO` to improve speed.
  ** Changed `TextIO` so that only `TextIO.stdErr` is unbuffered.
  ** Changed `TextIO` so that FIFOs and sockets are buffered.

* 2000-10-02
  ** Combined remove-unused-constructors, remove-unused-functions, and
  remove-unused-globals into a single pass that runs to fixed-point and produces
  results at least as good as running the previous three in (any) sequence.

* 2000-09-29
  ** Added `GC_FIRST_CHECK`, which does a gc at each limit check the first time
  it reached.
  ** Reimplemented `TextIO.StreamIO` (from 2000-09-12) to use lists of strings
  instead of lists of characters so that the per char space overhead is small.

* 2000-09-21
  ** Fixed bug in profiling labels in C code.  The label was always the basic
  block label instead of the cps function label.
  ** Added `-b` switch to `mlprof` to gather data at the basic block level.
  ** Improved performance of `TextIO.input1` by about 3X.

* 2000-09-15 - 2000-09-19
  ** Added overflow exceptions to CPS and Machine ILs.

* 2000-09-12
  ** Fixed `TextIO.scanStream`.  It was very broken.
  ** Added `TextIO.{getInstream,mkInstream,setInstream}` and
  `TextIO.StreamIO.{canInput,closeIn,endOfStream,input1,input,inputAll,inputLine,inputN}`.

* 2000-09-11
  ** Fixed `Real_qequal` in `mlton-lib.h`.  It was missing a paren that caused
  code using it to not even compile.  It was also semantically incorrect.
  ** Noted that `Real_{equal,lt,le,gt,ge}` may not follow basis library spec,
  since ANSI does not require IEEE compliance, and hence these could return
  wrong results when nans are involved.


== Version 20000906

Here are the changes from the 20000712 version to the 20000906 version.

=== Summary

* Version 20000906 is mostly a bugfix release over 20000712.  The other major
changes are that `mllex` and `mlyacc` are now included and that `mlton` can now
process a limited subset of CM files as input.

=== Details

* 2000-09-06
  ** Fixed `Socket_listen` in `mlton-lib.c` so that it closes the socket if the
  `bind`, `listen`, or `getsockname` fails.  This could have caused a file
  descriptor leak.

* 2000-09-05
  ** Added `-static` commandline switch.
  ** Changed default max heap size to .85 RAM from .95 RAM.
  ** Added `PackRealLittle` structure to basis library.

* 2000-08-25
  ** Added cases on integers to ILs (instead of using sequences of tests) so
  that backend can emit more efficient test (jump table, binary tree, ...).

* 2000-08-24
  ** Fixed bug in `gc.c`.  `dfsInitializeStack` would `smummap` a `NULL` pointer
  whenver `toSpace` was `NULL`.  This could cause `MLton.size` to segfault.
  ** Fixed bug in `Popt` that caused `-k` to fail with no keeps.

* 2000-08-22 - 2000-08-23
  ** Ported `mllex` and `mlyacc` from SML/NJ

* 2000-08-20 - 2000-08-21
  ** Added ability to use a `.cm` file as input to MLton.

* 2000-08-16
  ** Ported `mlprof` to SML.
  ** Fixed bug in `library/basic/assert.sml` that caused asserts to be run even
  when `MLton.debug = false`.

* 2000-08-15
  ** Fixed bug in backend -- computation of `maxFrameSize` was wrong.  It didn't
  count slots in frames that didn't make nontail calls.  This could lead to the
  stack being overwritten because a stack limit check didn't guarantee enough
  space, and lead to a segfault.
  ** Fixed bug in `gc.c` `newThreadOfSize`.  If the thread allocation caused a
  gc, then the stack wasn't forwarded, leading to a segfault.  The solution was
  to ensure enough memory all at once, and then fill in both objects.

* 2000-08-14
  ** Changed limit checks so that checks < 512 bytes are replaced by a check for
  0 bytes.  The runtime also moves the limit down by 512.  This is done so that
  the common case, a small limit check, has less code and is faster.
  ** Fixed bug in `cps/cps-tree.fun`. `Program.hasPrim` returned `true` for any
  program that had *any* primapp, not just programs satisfying the predicate.
  This caused `cps/once.fun` to be overly conservative, since it thought that
  every program used continuations.

* 2000-08-10
  ** Fixed bug in CPS typechecker.  It didn't enforce that handlers should be
  defined before any reference to them -- including implicit references in
  `HandlerPops`.  This caused an evil bug in the liveness analysis where a
  variable that was only live in the handler was missed in a continuation
  because the liveness for the handler wasn't computed yet.
  ** Limited the size for moving up limit checks for arrays whose size is known
  at compile time to avoid huge limit checks getting moved into loops.
  ** added `-indent`, `-kp`, `-show-types` switches.
  ** Put optimization in CPS IL suggested by Neal Glew.  It determines for each
  toplevel function if it can raise an exception to its caller.  Also, it
  removes `HanderPush` and `HandlerPop` for handlers that are not on top of the
  stack for any nontail call.

* 2000-08-08
  ** Changed register allocator so that continuation formals can be allocated in
  pseudo registers -- they aren't necessarily forced to the stack.

* 2000-08-03
  ** Fixed bug in constant folding.  `Word8.>>` had been used to implement
  `Word8.~>>`.
  ** Fixed bug in allocate registers that was not forcing the size argument to
  `Primitive.Array.array` to be a stack slot.  This could cause problems if
  there was a thread switch in the limit check, since upon return the size
  pseudo register would have a bogus value.

* 2000-08-01
  ** Turned back on XML simplification after monomorphisation.

* 2000-07-31
  ** Fixed bug in `MLton.Itimer.set` that caused the time to be doubled.
  ** Fixed bug in `MLton.Thread` that made it look like asynchronous exceptions
  were allowed by `throw`-ing an exception raising thunk to an interrupted
  thread obtained via a signal handler.  Attempting asynchronous exceptions will
  now cause process death, with a helpful error message.

* 2000-07-27
  ** Updated docs to include `structure World: MLTON_WORLD` in `MLton`
  structure.
  ** Added toplevel signatures `MLTON_{CONT, ..., WORLD}` to basis library.
  ** Fixed broken link in docs to CM in `cmcat` section.

* 2000-07-26
  ** Eliminated `GC_switchToThread` and `Thread_switchTo1`, since the inlined
  version `Thread_switchTo` is all that's needed, and Matt's X86 backend now
  handles it.
  ** Added `MLton.Signal.vtalrm`, needed for `Itimer.Set{which =
  Itimer.Virtual, ...}`.

* 2000-07-25
  ** Added `MLton.Socket.shutdownWrite`.

* 2000-07-21
  ** Updated `mlton-lib.c` `MLton_bug` with new email (MLton@sourcelight.com).

* 2000-07-19
  ** Fixed `Posix.Process.kill` to check for errors.

* 2000-07-18
  ** Fixed the following `Posix.ProcEnv` functions to check for errors:
  `setgid`, `setpgid`, `setsid`, `setuid`.
  ** Fixed `doc/examples/callcc.sml`.


== Version 20000712

Here are the changes from the 1999-07-12 to the 20000712 version.

=== Details

* 2000-06-10 - 2000-07-12
  ** Too many changes to count: bug fixes, new basis library modules, optimizer
  improvements.

* 2000-06-30
  ** Fixed bug in monomorphiser that caused programs with non-value carrying
  exception declarations in polymorphic functions to have a compile-time error
  because of a duplicate label.  The problem was that the exception constructor
  wasn't duplicated.

* 2000-05-22 - 2000-06-10
  ** Finished the changes for the new CPS IL.

* 2000-01-01
  ** Fixed some errors in the basis library:
    *** `Real.copySign`
    *** `Posix.FileSys.fpathconf`
    *** `Posix.IO.{lseek, getlk, setlk, setlkw}`
    *** `Posix.ProcEnv.setpgid`
    *** `Posix.TTY.getattr`
    *** `System.FileSys.realPath`

* 1999-12-22
  ** Fixed bug in `src/closure-convert/abstract-value.fun` that caused a
  compiler failure whenever a program had a vector where the element type
  contained an `->`.

* 1999-12-10
  ** Changed dead code elimination in `core-ml/dead-code.fun` so that wildcard
  declarations (`val _ = ...`) in the basis are kept.  Changed places in the
  basis library to take advantage of this.
  ** Added `setTopLevelHander` primitive so that the basis library code can
  define the toplevel handler.
  ** Changed `basis-library/misc/suffix.sml` to call `OS.Process.exit`.  Took
  out `Halt` transfer from CPS, since the program never should reach it.
  ** Cleaned up `basis-library/system/{process.sml, unix.sml}` to use the new
  signal handling stuff.

* 1999-11-28 - 1999-12-20
  ** Added support for threads and cleaned up signal handling.  This involved a
  number of changes:
    *** The stack is now allocated as just another kind of heap object.
    *** Limit checks are inserted at all loop headers, whether or not there is
    any allocation.  This is to ensure that the signal handler always has a
    chance to get called.
    *** The register allocator puts more variables in stack slots.  The new rule
    is that a variable goes in a stack slot if it is ever live across a nontail
    call, in a handler, or (this is the new part) across a limit check.
    *** Arguments are passed on the stack, with the convention determined by
    argument types.
    *** The "locals" array of pointers that was copied to/from for GC is now
    gone, because no registers (in particular no pointer valued registers) can
    be live at a limit check point.

* 1999-11-21
  ** Runtime system
    *** Fixed a bug introduced by the signal code (presumably on 1999-08-09)
    that caused a gc to *not* be performed when doing a save world.  This caused
    the heaps created by save world to be the same size as the heap -- not the
    live data.  This was quite bad.
    *** Cleaned up the `Makefile`.  Add make depend.
    *** Added max gc pause to `gc-summary` info.
    *** Move heap translation variables that had been file statics into the
    `GC_state`.
  ** Made `structure Position` available at toplevel.
  ** Basis Library
    *** Added `MLton.loadWorld`
  ** Added `Primitive.usesCallcc`
  ** Added `Primitive.safe`
  ** Removed special size functions from `cps/save-world` -- they are no longer
  necessary since size doesn't do a gc.
  ** Fixed another (sigh) bug in `cps/simplify-types.fun` that could cause it to
  not terminate.

* 1999-11-16
  ** Cleaned up `backend/machine.fun` a bit so that it spits out macros for
  allocation of objects and bumping of frontier.  Added macros `MLTON_object`
  and `MLTON_incFrontier` to `include/mlton-lib.h`.
  ** Fixed a bug in `backend/limit-check.fun` that caused loops to not be
  detected if they were only reached by a case branch.  This could cause there
  to be loop that allocates with no limit check.  Needless to say, this could
  cause a segfault if the loop ran for long enough.

* 1999-10-18
  ** Added basis library function `Array2.copy`.

* 1999-08-15
  ** Turned off globalization of ref cells (`closure-convert/globalize.fun`)
  because it interacts badly with serialization.

* 1999-08-13
  ** Fixed bug in `mlton-lib.h` in `MLTON_allocArrayNoPointers` that was
  triggered when `bytesPerElt == 0`.  The problem was that it wasn't reserving
  space for the forwarding pointer.  This could cause a segfault.

* 1999-08-08 and 1999-08-09
  ** Added support for signal handling.

* 1999-08-07
  ** Fixed bugs in `Array.tabulate` (and other `tabulate` variants) caused if
  the function argument used `callcc`.

* 1999-08-01
  ** Added serialization, which was mostly code in `src/runtime/gc.c`. +
  `GC_serialize` converts an object to a `Word8Vector.vector`. +
  `GC_deserialize` undoes the conversion. + (de)Serialization should work for
  all objects except for functions, because I haven't yet added the support in
  the flow analysis.

* 1999-07-31
  ** Cleaned up the GC.  Changed headers, by stealing a bit from the number of
  non pointers and making it a mark bit (used in `GC_size`).
  ** Rewrote `GC_size` so that it runs in time proportional to the number of
  pointers in the object.  It does a depth-first-search now, using toSpace to
  hold the stack.

* 1999-07-30
  ** Fixed bug in `SUBSTRING`.  `getc` had the wrong type.  This bug wasn't
  noticed because MLton doesn't do enough type checking.
  ** Fixed bug (segfault) caused when a GC immediately followed a throw.

* 1999-07-29
  ** Fixed bug in `Date.fmt` (`basis-library/system/date.sml`).  It was not
  setting `Tm.buf`, and hence the time was always 0 unless there had been a
  previous call to `setTmBuf`.

* 1999-07-28
  ** Fixed bugs in `Posix.IO.FLock.{getlk,setlk,setlkw}`, which would cause
  compilation to fail because `FLock.toInt` was defined as the C `castInt`,
  which no longer exists.  Instead, expand `FLock.toInt` to
  `MLTON_pointerToInt`, which was added to `include/mlton-lib.h`.
  ** Changed `Posix.Primitive.Flock` to `Posix.Primitive.FLock`.
  ** Added `MLTON_chown`, `MLTON_ftruncate` to `include/mlton-posix.h`.  They
  were missing.  This would cause compilation of any program using
  `Posix.FileSys.{chown,ftruncate}` to fail.  Also made it so all of the
  primitives in `basis-library/posix/primitive.sml` use `MLTON_` versions of
  functions, even if a wrapper is unnecessary.

* 1999-07-25
  ** Added some other missing signature definitions to toplevel.

* 1999-07-24
  ** Added missing `OS_*` signature definitions to
  `basis-library/top-level/top-level.sml`.

* 1999-07-19
  ** Fixed bug in `basis-library/arrays-and-vectors/mono-array.sml`.  Used `:>`
  instead of `:` so that the monomorphic array types are abstract.


== Version 19990712

Here are the changes from the 1999-03-19 version to the 1999-07-12 version.

=== Details

* 1999-07-12
  ** Changed `src/backend/machine.fun` so that the 'pointer locals' array is
  only as large as neccessary in order to copy all pointer-valued locals, not as
  large as the number of pointer-valued locals.

* 1999-07-11
  ** Rewrote `src/backend/allocate-registers.fun` so that it does a better job
  of sharing "registers" (i.e. C local variables) and stack slots.  This should
  cut down on the amount of copying that has to happen before and after a gc.
  It should also cut down on the size of stack slots.

* 1999-07-10
  ** Fixed a bug in `src/backend/parallel-move.fun` that should have been
  triggered on most any parallel move.  I guess parallel moves almost never
  happened due to the old register allocation strategy -- but, with the new one
  (see note for 1999-07-12) parallel moves will be frequent.

* 1999-06-27
  ** Fixed `src/main.sml` so that when compiling `-p`, the `.c` file is compiled
  `-g` and the `.o` is linked `-p`.
  ** In `bakend/machine.fun`, added profiling comments before chunkswitches and
  put in an optimization to avoid printing repeated profiling comments.  Also,
  profiling comments are only output when compiling `-p`.

* 1999-06-17
  ** Changed `-i` to `-inline`, `-f` to `-flatten`, `-np` to `-no-polyvariance`,
  `-u` to `-unsafe`.
  ** Added `-i`, `-I`, `-l`, `-L` flags for includes and libraries.
  ** Updated documentation for these options and for ffi.

* 1999-06-16
  ** Hardwired version number in `src/control/control.sml`.  As it stood, the
  version number was computed when MLton was built after someone downloaded it,
  which was clearly wrong.

* 1999-06-16
  ** Fixed undefined variable `time` in `GC_done` in `src/runtime/gc.c`.

* 19990-06-08
  ** in `include/mlton-lib.h`:
    *** removed `#include <huge_val.h>`
    *** added `#include <math.h>`
    *** and deleted all of the function signatures I had copied from `math.h`
  ** Changed `Real.{minNormalPos, minPos, maxFinite}` so that they are computed
  in `real.sml` instead of appearing as constants in the C.

* 1999-06-07
  `IntInf.pow` added to basis library.

* 1999-06-04
  ** `bin/mlton` changed to use `.arch-n-opsys` if it exists.

* 1999-06-03
  ** `src/Makefile` changed to use `sml-cm` instead of `sml`

* 1999-05-10
  ** Patch to `src/atoms/small-int-inf.fun` to work around a bug in the SML/NJ
  implementation of bignums.  This bug was causing some hex bignum constants to
  be lexed incorrectly.

* 1999-04-15
  ** Comments emitted in C code for profiling.  The comments identify the CPS
  function responsible for each C statement.

* 1999-04-15
  **  `callcc` and `throw` added.

* 1999-04-15
  ** Bug in `src/cps/simplify-types` fixed.  The bug caused nontermination
  whenever there was a circular datatype with a vector on the rhs.
  E.g. `datatype t = T of t vector`


== Version 19990319

Here are the changes from the 1998-08-26 version to the 1999-03-19 version.

=== Summary

* Compile time and code size have decreased.
* Runtime performance of executables has improved.
* Large programs can now be compiled.
* MLton is self hosting.
* The basis library is mostly complete and many bugs have been fixed.
* The monomorphiser (`-m`) is no longer available.
* The heap and stack are automatically resized.
* There are now facilities for heap checkpointing (`MLton.saveWorld`) and object
size computation (`MLton.size`).
* MLton uses the GNU multiprecision (GnuMP) library to provide a fast
implementation of `IntInf`.