porting changes from legacy

  porting changes from legacy branch.  Adding Basis Library extensions 2021-001
  and 2022-001.

port some minor edits from legacy

fix for issue #244 (Unable to interrupt execution by using )

fix for bug (gforge) 270, PPAbsyn.ppDconPat not stripping MARKpat

added two print flags to FLINT_Control relating to prettyprinting FLINT

ppast.sml - fix for bug (gfornge) 270, issue #255?

porting fix for Real64.scan from legacy

  port fixes for bugs #318 and #319 from legacy version

Port bug fixes for #316 and #317 from legacy version

fixed bug #314

update manual

  Fix pretty printing issue (extra newline before the "sig" when printing
  nested structures).

add some comments

  Porting fix for bug #314 (`IEEEReal.setRoundingMode` is a no-op on Linux)

  port fix for bugs #65 and #290 from legacy version

fix for bug 290: FLINT/opt/{lcontract,fcontract}.sml, minor cleanup in elabmod.sml

porting CM changes from legacy branch

fix for bug #312 (CM.make is unable to handle filenames that contain a backslash)

  fix bug #284

update copyright date

  Fix for bug#310 (Error when REPL tries to print value of type Posix.FileSys.ST.stat)

fix shell script

  Fix for bug #306 (Word8VectorSlice: mapping a subslice produces wrong result or
  crashes SML/NJ)

fix for packStr removal, FLINT/kernel changes, minor pp changes

hacking makefiles

makefile fixes for Linux

  Remove conditional inclusion of LLVM backend

fix warning message

  add check for CMake presets support

  disable timing of the code generator 

  Switching the "base" trunk to the LLVM backend.  The previous version of base can be
  found in the branch

    https://smlnj-gforge.cs.uchicago.edu/svn/smlnj/sml/branches/pre-llvm-switch

tweaks to binfile.sml related to new format

remove BUILD64.txt, since it is no longer applicable

remove abcdef.{cm,sml}

remove legacy NOTES directory from base

remove READMES stub

comments

fix bug in printing of char values in the REPL.

remove obsolete comment

  Added support to runtime bootstrap code to work with both old and new-format binfiles.

update header comments

code layout

code layout

code layout

code layout

add support for -C/D argments

code layout and comments

pass -C/D arguments to sml

  remove deprecated Poll structure

  remove deprecated Poll structure

comments

remove last traces of lambda splitting from compiler

support for new binfile header format (not enabled yet)

fixes for bug 281

massive changes for new match compiler, FLINT types reorg, some bug fixes

edited FLINT/opt/fcontract.sml to fix bug 294

  add comment

clean-up of Elaborator/matchcomp

copied Elaborator/matchcomp for merging

fix pretty printing of val decls in REPL

  Changed the *binfile* representation to remove the pickling of
  the *FLINT* intermediate representation.

comments

   fix issue with 32-bit machines

  Changed the way that versions are identified to allow a version suffix
  (e.g., "-rc1" for "release candidate 1").  I removed the bump-release
  mechanism, which might has served a similar purpose, but was restricted
  to integer suffixes and has not been used in recent memory.

  remove Elaborator/matchcomp, which is an old version of the match
  compiler that was never used, but got added to the repo by accident.

  removed old code for the MLRisc stack layout on AMD64, since we now always
  use the LLVM-compatible layout

fix comment

  fix a benign bug where the size of a floating-point spill record was twice
  as large as necessary on 64-bit systems.

porting some improvements from the LLVM branch

  fix bug in toLargeReal

  Updated the help message to mention the "@SMLwordsize" option.

  Split out the Real.toLargeInt implementation into target-word-size
  versions.  For 64-bit targets, the new version uses the bit representation
  of the real number to compute the result.

  add "-cleanall" option to installml script.  This option is like "-clean", but it
  also removes any cached ".cm" files from the various tools and libraries.

undid bad commit from revmc working copy

various changes relating to debugging match compiler

bringing revmc/FLINT/trans up to date with match compiler changes

put comment in correct place

correction to Math64Common.scalb function

  use misc5 and misc6 as temporaries, since they do not have to
  be preserved across the call.

  minor edits to comments/code layout

  Added a hash table to cache inline operators by numeric kind.  This change
  speeds up the compiler by about 2-3% (e.g., compiling the compiler
  went from 58s to 55s on a MacBookPro with a 2.4GHz Intel i9 processor).

editing a comment

adding a comment

  Changed AMD64 frame layout to include a word to hold the
  Overflow exception.  This value is used by the LLVM backend
  to generate the exception for checked arithmetic operations.

  clean up the installation logic

fix typo in comment

  fix conversion of FLINT numeric types to CPS types.

  porting some runtime system tweaks from LLVM version of runtime

  add compare function to IntConst module

  Fix for bug #280 (110.99 config/install.sh -64 fails on MacOS 10.15.7)
  After some investigation, it appears that the problem was inconsistent
  build tools being picked up from the user's path.  To protect against
  this issue, I've made the paths to the "ar" and "ranlib" tools absolute.

  further tweaking of buffer growth

remove obsolete comment

use 256K as the growth limit for buffers

  Fix some correctness and performance issues in the {Char,Word8}Buffer modules.

adding comments

  Switch the layout of the SML stack frame to be compatible with
  where LLVM likes to spill registers.  We also added a new file
  (compiler/CodeGen/amd64/amd64-frame-layout.sml) that collects
  together all of the stack frame information in one place.

  change magic string format for binfiles

change the format of the magic string in binfiles

  fix for bug #274 (Minor pretty printing glitch when printing structure specs)

  fix a couple of regressions

minor pretty-printing cleanup

  Fix for bug #277 (Excess white space when pretty printing a
  module signature)

  Fix bug #276 (missing option to control extra newlines in REPL)

  added helper functions for printing separators

  update pretty printer for new PP device API

  Fix for bug 254 (Real.fromLargeInt produces negative results)

  Fix bug #267 (Returns an incorrect result for a calculation on Position.int
  for 32bit mode).

  Moved the description of the binfile format to `dev-notes/binfile.adoc`

  simplify binfile format

  Simplify binfile format by only supporting a single code
  object per binfile.  Also removed dummy CpsSplitFun functor.

major rewrite of matchcomp for unique lvars

  porting some minor changes from LLVM branch

merge in some minor additions from LLVM branch

remove unused mkRecord.{sig,sml} code

minor edits to eliminate warnings

pretty printer fix

  fix bugs #266, #268, and #271 (pretty printing regressions) by
  reverting the some of the pretty printing code back to 110.97.

reorganize backend code in preparation for LLVM codegen

code formatting

edit comment

header comment

  removed unused code

  Fix for bug #269

minor edits in transutil.sml

resolved minor svn conflict in matchcomp.sml

resolved conflict in translate-new.sml

  Added an additional lowering pass for the `STREQL` primop.  Also changed
  the representation of the primop to include the string being tested
  against, and removed the `STRNEQ` primop.

fix problem with 32-bit linux build

remove trailing whitespace

get the type right for ANDB (should be unsigned)

  remove unnecessary joins for code that does not have any branches in it.

print magic numbers when they don't match

cleanup of the old matchcomp.sml

fixes for lowering pass

working on new matchcomp integration

working on translate-new.sml an transtypes.sml

additions to comments

header comment

fixed spelling in comment

remove trailing whitespace

remove debugging code

comments

remove trailing whitespace

remove trailing whitespace

added top-level match.sml file

near final changes to new matchcomp, calc -> getIndex in DebIndex

minor corrections to get trans/transutil.sml to compile

added translate-new.sml, initially a copy of translate.sml, and transutil.sml containing some utility functions originally in translate

  Adding the `Universal : UNIVERSAL` structure (Basis proposal 2020-001).

  Adding code to do the patching of scripts (for future use)

fix typo

  Potential fix for bug #260

vcode.sml defined pseudo-constructors for mcexp

fixes for module/signature pretty printing problems

  Add new lowering pass.

  Change calc_strictness to calcStrictness

  code layout and comments

fix spelling error

fix for bug 261; moving calc_strictness to TypesUtil

  rename old spill.sml to spill-old.sml

move FLINT/matchcomp (new match compiler) to Elaborator

add FLINT/matchcomp/mctrans.sml (translation from mcexp to Plambda.lexp)

fix bug in CompInfo

add ElabData/pickle/README.md

changed argument field for mkCompInfo

some uncommitted edits of base/compiler/MAP

minor edits/code layout

  Change default word size to 64 bits

Removed the compiler's dependency on html-lib.cm.

  Completed the removal of the trigonometry operators from the primops.

moved vartypes.sml to old

moved old match compiler code to old-versions

reorganizing the matchcomp directory to move old
stuff to the old subdirectory

code generation completed with RHS linkage

remove unused file

add some TODO comments

  Remove use of the hardware trig functions on x86.  We make
  this change for two reasons:
    1) it makes it possible to cross compile from 64 to the
       32-bit x86 (although the same result might be achieved
       using the CMB_CROSS symbol)
    2) it paves the way to removing these operations from the
       compiler's primops.

  Set CMB_CROSS CM variable in first stage of cross compile.
  We use this to avoid compiling things that are problematic,
  but not required.

  Added strength-reduction optimizations to CPS contraction.
  These include optimizing div/mod by powers of two, since
  the plan is to convert these operations to native arithmetic
  (i.e., quot/rem) prior to code generation.

better error message

making changes for typed match compiler

  Change CPS representation to use a boolean to specify re-entrant
  C Calls (which do not seem to currently supported).

add print-clusters control

adding types to new match compiler

comments and removal of unused definitions

  remove fixedArgPassing flag definition, since the default definition
  is correct (and the same).

  add flag and code to enable printing of the clusters of CPS functions
  prior to code generation

print function kind when printing top-level lambda

bug fixes and printing improvements in FLINT/matchcomp

debugging new matchcomp

edit comment

added TODO comment

added parse directory with lexer and parser for patterns

updating FLINT/matchcomp with pretty printing and testing

added lib directory, mcprint.sml; debugging; mctypes.sml compiles

edited 3 files in FLINT/matchcomp

edit mc-code.sml

removed generate.sml

removed files that were moved to old directory

adding files to matchcomp

committing changes to FLINT/matchcomp

editing mctypes.sml, dec-tree.sml, andor.sml

edits to dec-tree.sml

almost final decision tree code in dec-tree.sml

update of FLINT/matchcomp with new match compiler code

adding several new files

added FLINT/matchcomp directory with new (under development) match compiler code

  increase the default printDepth and printLength values

code layout

remove unused module binding

moved files in Elaborator/print into subdirectories of ElabData

update to prettyprinters and pputil.sml

rationalized printing in Basics/print/printutil.s?? and Elaborator/print/pputil.s??

fixed missing ] in ElabData/basics/access.sml

updated MAP to reflect reorganization of Basics

use String.concat

  documentation and code layout

PrettyPrintNew -> PrettyPrint, PPUtilNew -> PPUtil, reorganized Basics

  adding some comments to the ASDL spec and a missing import

  disabling the unused variable warning until we can fix the
  problem with false positives

  add check for unused variables

  make complainer type explicit in ERRORMSG signature

updated compiler/MAP after file and directory changes

moving MiscUtil/print to TopLevel/print, renaming MiscUtil/library to MiscUtil/pickle and moving MiscUtil to Library

  moved feedback.sml from MiscUtil/util to CodeGen/cpscompile.
  Deleted the MiscUtil/util directory, since it is now empty

  moved feedback.sml from MiscUtil/util to CodeGen/cpscompile.
  Deleted the MiscUtil/util directory, since it is now empty

make cmb-cross use the sml command's default wordsize

  Made the LambdaVar.lvar type abstract.

  makeml now uses the sml command's wordsize as the default (instead of 32-bits)

fix bug #256 (Ref.exchange incorrect)

add TODO comment

updated Elaborator/print/pptype.sml for new overload tyvar printing

updated compiler/MAP with changes in overloading files

added missing size option

fix minor bug in makeml

  restore old overload syntax in pervasive.sml so that we can compile
  using the 110.96 compiler.  The type annotations will be ignored
  by the 110.97 compiler.

  Allow type constraint in overload syntax (the type is ignored).

code layout

updated compiler/MAP for changes in Semant

added new files overloadclasses.sml and overloadvar.sml

fixed overloading for bug 220; moved Semant/prim and Semant/pickle and deleted Semant

fix punctuation

  move documentation of Acess types to signature.

  fix makeml bug in handling -boot option.

fix for bug 214 committed

updated bool.sig and list.sig so elabtype can report error on forbidden datatype constructors

changes to elabutil.sml, elabcore.sml, elabtype.sml to fix but 209

fixed bugs 195, 196 and deleted Semant/elaborate and Semant/modules

code layout

  We do not need the '-ldl' flag on macOS; dynamic linking support is
  always available

  Move "-lm" LD flag from makefile into target-specific makefiles,
  since it is not required on macOS.

  code layout

fix for bug #252 (Boyer Benchmark Compile Failure)

better debug infor for DEBUG_LITERALS

fix max_length for 64-bit targets

  Fix for bug #245 (Lazy data types result in Compiler Bug error).

improve internal error message

  Fix bug 244 (Compiler bug: PPObj: ppFields in ppval.sml)

  Better error message

fix for bug #239 (Date.toTime is incorrect (by a factor of 10E9))

  Some preliminary changes to support using a 64-bit compiler to
  generate the bin files.  Unfortunately, this does not work yet
  because of a issue (see bug #238).

code layout

  Use word-sized fields in VProc state vector so that the word-sized
  move operations in the assembly code do not clobber adjacent fields.

  Turn Real.<(a, b) into Real.>(b, a) when generating code, since
  it allows us to avoid having to test for the unordered case
  (that is because Real.> involves testing two condition-code
  bits that both must be 0; they will both be 1 when the arguments
  are unordered).

  Switch to using the negInf/posInf values from Real64Values, instead of
  regenerating them.  Also fixed the signature to refer to LargeReal.real.

fixed typo

  remove assertion checking for AMD64 runtime

  better debug messages when SIGNAL_DEBUG is set

working on AMD64/FreeBSD support

simplify getArchName

  tweaking FreeBSD defines

  switch to ucontext_t representation of machine state for FreeBSD

  working on AMD64/FreeBSD support

move mklinux stuff to Unsupported

change gcc to cc for freebsd

  adding makefile for amd64/freebsd

fix makefile for x86/FreeBSD

  We no longer use/support the "brk" version of the memory library
  so we can remove malloc.c from the repo.

  Merged the TARGET_xxx and HOST_xxx preprocessor symbols into
  a single ARCH_xxx symbol.  This change was also reflected in
  the SMLofNJ.SysInfo module.

  Removed the last vestiges of the old bytecode interpreter from the
  runtime system.

  Fix for bug #231 (New literals-lifting code does not handle pair of reals)

handle non-standard developer path on macOS

  minor runtime edits

  simplifying the build requirements for Cygwin

  some minor edits to support Cygwin64 (but we have some work to do)

  working on 64-bit support for Cygwin

  be consistent about including the -DSIZE_xx definition for all targets

working on cygwin support

fixing overflow detection on x86/linux

  Fix for bug #230 (segmentation fault when compiling MLton sources
  with SML/NJ 64-bit).  The problem was that when a large vector was
  being created, the assembly code did not correctly restore the stack
  state before trying to call the runtime system to do the allocation.

add missing case for DataReplDec

  Fix Real.scan so that 0.0 is always scanned as 0.0, even when there
  is a very large exponent (bug #229).  Still need to check if this
  change fixes the memory fault on Linux.

no longer use MMAP_ADDR symbol

  use MAP_ANONYMOUS on AMD64/LINUX

remove some old DEC Alpha specific code

reapply fix for Windows Math functions

  some small windows fixes

  add amd64-unix to list of targets in allcross script

  minor improvements to the comments

fix bug in MarkRegion that broke the heap in Linux

fix some warnings

  Use `off_t` for file offsets

switch Unsigned32_t to Addr_t

adding some extra assertions to heap import

  Use off_t and size_t types for input functions.

updated BUILD64.txt instructions

adding makefile for linux

  Added support for conditional targets to the installer.

  fix the Overflow bounds for the `floor` function.

minor edit to INSTALL instructions

  Add contraction for TEST and TESTU applied to constants

  Fix for bug #227 (CPS contraction is taking an excessive amount of time on word8 basis test)

bug fix for arithmetic right shift for small sizes

removed tcode_int32

  added contraction optimizations for when conversion operators
  take constants as arguments.

  change the way that tagged integer literals are printed

code generation bug fixes

fix typos

  Modified the Unix installer script to pass a size argument to the configuration script.

update installation instructions

  porting changes from num64 branch

porting bug fix from num64 branch

porting changes from num64 branch

  porting changes from num64 branch

adding target-specific rebindings for 2004 basis

added BUILD64.txt with interrim instructions for cross compiling to 64-bit version

  porting bugfix from num64 branch

  porting installer changes from num64 branch

porting changes from num64 branch

  porting changes from num64 branch

  add -DINDIRECT_CCALL to definitions.  This flag will ensure that C function references
  in the heap look like aligned pointers.

fix for bug 130, changing typesutil.s?? and typecheck.sml

  porting bug fixes for Int64 and IntInf from num64 branch to trunk

  porting minor changes from num64 branch.  a couple of bug fixes, but
  mostly code cleanup.

add missing GC check

improve comment

  new literals runtime support and 64-bit cleanup

code layout

add newLiterals control flag

adding new literals bytecode generation

  adding new literals bytecode generation

code layout

remove unused global machine_id

  bug fix for bug #225 (`Math.ln` giving erroneous answers on Windows).

add rkToString function

add way to pass compiler control flags to cmb-make

fix comment

  Clean up in the Basis Posix library code (both SML and runtime) to
  be consistent about when the `SysWord.word` type is being used to
  communicate information between SML code and the runtime system.

  porting various 64-bit changes from the num64 branch

  add debugging code for literals

  added controls for literals

  Fix for bug #225 (Math.ln giving erroneous answers on Windows).  The problem
  was an inconsistency in the way the Unix and Microsoft assemblers interpreted
  the addressing mode for the FLD instruction.

  expose the Assembly.A.scalb function for testing

align allocptr before allocating double

  remove unused file

  porting runtime changes from num64 branch

  porting runtime 64-bt bug fixes from num64 branch

porting changes from num64 branch

remove 64BIT FIXME comment

  more conversion fixes

  Bug fix: we were dropping on of the arguments to EXTEND_INF

  Generalize the handling of conversions involving tagged integers/words, where
  the size is _not_ the default integer size.  This situation only occurred for
  Word8.word values on 32-bit targets, but also occurs for 32-bit values on
  64-bit targets.

working on 64-bit port

  change Word32 to SysWord

  Add comment clarifying where updtHeapPtr needs to be called.

  Added wrappers for 32-bit int and word operations on 64-bit targets

  add missing files needed for 64-bit targets

  Rewrote the expansion of the `INLLSHIFT`, `INLRSHIFTL`, and `INLRSHIFT`
  primops (`compiler/FLINT/trans/transprim.sml`).  The expansion process
  now correctly handles shift operations on types that are smaller than
  the default tagged-integer size.  This change allows the `Word8` shift
  operations to be inlined.

  One more step in the contraction bug fix

  code cleanup

  Bug fix: constant folding of arithmetic-right-shift operations was not getting
  the sign right for words when the most-significant-bit was set.

  remove trailing whitespace

code formatting fix

  Fix off-by-one error in Real.toManExp function.  This bug caused
  the Real.toLargeInt function to go into an infinite loop in some
  cases.

  correct typo in comment

  Fix bug #173 for Windows.

  Fix for bug #173 (`OS.Process.sleep` only works with whole numbers)

  add missing defn

  Restructured the CPS contraction phase to make the conversion-fusion
  cleaner and more uniform.

slightly better code for Int64.fmt

  Some improvements to mlrisc gen

minor improvements to contract.sml

improve comment

use symbolic names instead of raw numbers

  Add access to IntInf.int internal representation to Unsafe structure.

  Add missing CloseFile

  improve error message

  tweaked the cross-compiling scripts to produce the same directories
  and tarballs as previously

  remove AMD64 cross defs, since the cmb-cross script
  doesn't need it.

  update cmb-cross script to use new cross-compilation scheme

remove cases for unsupported architectures

  Switch to new names for 32-bit to/from IntInf conversions on 32-bit platforms.

switch to new names for IntInf conversions to/from 32-bit ints

fix header comments

  Cleanup of Cluster module; use List combinators where possible.

  added missing copy_word_to_int64 primop that is needed by
  the 64-bit version of core-intinf.sml

Providing target-specific overrides for size and endianess for the purpose of cross-compilation to 64 bit.

  more porting to 64 bits; use InitBibop for global BIBOP.

use InitBibop/FreeBibop for oldBIBOP

add InitBibop/FreeBibop functions

fix type of MarkRegion

porting to 64 bits

porting to 64 bits

  porting runtime to 64 bits

  Use MALLOC/FREE macros

no InitFPE needed for amd64

  assembly code fixes

64-bit fixes

  Fix warning messages

  Fix bug #224 (Word64.fromLargeInt fails).  The problem was an incorrect record
  kind in CPS/opt/infcnv.sml (RK_RECORD instead of RK_RAWBLOCK).

update header comment

  Replaced the "Word32" representation constructor with "Raw" (which covers both
  32 and 64-bit numbers).  We now check the length of the raw object when converting
  to {Int,Word}{32,64}.

fix typo in BIBOP_L2_SZ

improve comment

  Remove runtime types for polymorphic arrays (ported from unrtt branch).

trailing whitespace

  Fix 64-bit issues in limit.sml; remove some bogus allocation counts; code
  layout.

add FIXME for word_to_real64

  removing explicit word32 to real conversion; eventually, we will want to add
  a word_to_real primop.

code formating

  Working on 64-bit support

  Remove unsupported Alpha, HPPA, MIPS, and RS6000 related files.

tweaking inline operators for 64-bit targets

add 64-bit implememtation of CoreIntInf

  code formatting

  add rules for 64-bit targets to init.cmi

  update ASDL to match cps.sig

fix type warning on Window

fix type error in Windows runtime code

fix Windows makefile issue

porting ASM changes to Windows

porting ASM changes to Windows

porting new ASM code to Windows

  Split CHECKLIMIT, ENTRY, and ML_CODE_HDR macros out into x86-macros.h

  porting changes from num64 branch

  working on porting runtime to 64-bits

  Working on porting runtime to 64-bit targets

  working on AMD64 assembly code

add comment about number of live miscregs

added some comments

  Switching to new X86.prim.asm file and working on AMD64.prim.asm.

add _FILE_OFFSET_BITS=64 for Linux

add _FILE_OFFSET_BITS=64 for Linux

  Switch to new version of X86-prim.asm

  Switch to new version of X86.prim.asm

    move old X86 assembler code to Unsupported

  The macros ENTRY and ML_CODE_HDR are defined in x86-syntax.h on
  the x86 and amd64 platforms.

tweak definition of CONCAT

remove unsupported systems

  Working on new X86 syntax

remove #ifdefs for unsupported systems

switch ML_AllocNanoseconds to use Unsigned64_t instead of Int64_t

working on x86-syntax.h

  Add PackWord64 implementations and some clean up of other Basis modules.

  Added 64-bit implementations of target-dependent Basis modules.

fix unsafe word32 unpacking with sign extension

  Added bigEndian flag to Target module

working on AMD64.prim.asm

  Working on new assembler syntax file to replace assyntax.h, assyntax64.h,
  and x86-masm.h.

restore x86-masm.h for now

remove unused x86-masm.h file

fix a couple of runtime type errors

porting changes from num64 branch

adding /nologo flag

porting changes from num64 branch

add 64-bit comment

adding cmb-cross script (again)

porting changes from num64 branch

AMD64 runtime system hacking

  Porting changes from num64 branch in preparation of 110.90
  release.

Fix bug 222.

  porting edits from num64 branch.  These are largely cosmetic.

  merge changes (comments)

  Add `pointerTy` to basic types

  Adding placeholders for exposing the c_pointer type to
  user code (Unsafe.Pointer).

add 64BIT comments

  remove unused/unsupported flags from MACH_SPEC signature
  (unrolled lists and quasi stacks)

  remove unused tcode_int32

  64-bit cleanup: use pty instead of 32 for address arithmetic

  64-bit cleanup

  Fix computation of nanoseconds for file timestamps

  porting changes from num64 branch (switch to 64-bit
  FixedInt, LargeWord, and Position structures)

add 64BIT comment

  Add "c_pointer" basic type to represent runtime-system pointers

add a bit more space in debug printing

add pointerSz to TARGET signature

  Change PT_CFUN to PT_POINTER to allow for adding a raw_pointer basic type.

add 64BIT FIXME comments

code layout

fix comment

remove makefile for unsupported OS (NETBSD2)

  archive assembly code for unsupported architectures

64-bit cleanup

port minor edits from num64 branch

remove junk file

  Runtime-system cleanup.  Removed unsupported systems from include/ml-unixdep.h

  We are now assuming that we have at least C99 support
  (for practical purposes, this assumption is even true
  on Windows).  Therefore, I've switched the implementation
  of small-object allocation from macros to inline functions
  in the runtime (see runtime/include/ml-objects.h).  This
  change allows a graceful handling of 32-bit integers,
  which are heap allocated on 32-bit machines, but tagged
  on 64-bit machines.

  remove makefiles for Alpha and HPPA, since we no longer support
  those architectures

  fix minor issue in f_truncate_64 runtime function

port minor edits from num64 branch to trunk

port minor edits from num64 branch to trunk

  Fixed various bugs in the implementation of the `Word64` operations.

  Some cleanup in the Basis IO code in preparation for switching
  to 64-bit positions

add comment

working on simplified MLRiscGen

  improve compiler error message

working on simplified MLRiscGen

working on simplified MLRiscGen

  remove bogus character

  Working on simplified version of MLRiscGen functor

  header comments

  Code restructuring and reformatting.

  fix error in ASDL spec

  Add contraction for unsigned REM and NEG operations

  code reformatting; add 64BIT comments

  add uN eg to signature

  Move DEVNOTES out of compiler source tree and into dev-notes (renamed old-compiler-notes).

  Fix the way that polyequal handles tag_raw to support 64-bit
  numbers.

  Porting changes from branches/num64.

  merge in benign changes from the num64 branch

added comments

  fix file name to follow convention

  More reorganization of Basis files.  Combined the common exports into
  Exports/bind-common.sml and then added a 32-bit-specific file
  (bind-target-32-bit.sml) for the rest.

add missing file

  Some reorganization of the Basis Library sources to separate 32-bit and
  64-bit target code.

refactor IntInfCnv

add debug print after zeroexpand phase

  Add contraction optimizations for unsigned comparisons with 0

  Removing references to Int31 and Word31 from Basis Library implementation.

  Enable inlining of Char.chr (was supposed to be in 110.87, but missed
  the commit)

fix math64-none.sml and math64-sqrt.sml

  Some of the CPS optimization modules were written as functors over the
  machine spec, when, in fact, the never reference their functor argument,
  so I've switched them back to structures.

  reorder primops to agree with dev-notes/primop-list.md

use InlineT.compose in Fn structure (instead of 'op o')

  Use the InlineT.identity primop for Fn.id.

  Fix regression in printing of character results.

  Renaming the primops in Inline.  This change removes most (but not all) 32-bit
  dependencies in the Basis Library implementation.  See dev-notes/primop-list.md
  for details about the new naming conventions.

  Primop renaming

add FIXME comment

  add missing cases

  Rework the way that primitive values are printed and add clauses
  for the new primtive monomorphic vector/array types.

  Backing out the primop renaming, since I'm having some trouble getting
  to a fixed point.  We build a version from this point (which has all
  of the Primop.primop changes) and then try to get the renaming working
  for the 110.87.

  Major overhaul of primop names.  Also changed InLine structure
  to Inline (to match InlineT).

changed InLine to Inline

get fixedIntSz correct for 32-bit targets

  Replaced the `NUMSUBSCRIPT` and `NUMUPDATE` constructors with
  `NUMSUBSCRIPT`, `NUMSUBSCRIPTV`, `NUMUPDATE`, `INLNUMSUBSCRIPT`,
  `INLNUMSUBSCRIPTV` and `INLNUMUPDATE`

  added some FIXME comments

  Major overhaul of the internal representation of the Primop.primop
  type.  See the HISTORY.txt file for details.

add 64BIT comment

  Removed unused Primop.mayRaise function and added Primop.UADD.

  Renamings to remove 32-bit assumptions.

  Removed unused record kind constructors from CPS representation.

  Code cleanup; use RedBlackTree set for set of functions.

  Renamed DTAG_raw32 to DTAG_raw, since the semantics on 64-bit systems will be
  to require word-size aligned raw data.  Also renamed ML_AllocRaw32 to ML_AllocRaw
  and ML_ShrinkRaw32 to ML_ShrinkRaw for similar reasons.

  Remove more unused flags from Control.

  Removed unused flags from Control.GC

  Split primop contraction out of contract.sml into its own file

  Remove contractions from error messages

  Removing 32-bit assumptions in various places

  Split the translation of primops to PLambda out into its own file
  (`compiler/FLINT/trans/transprim.sml`).  Also cleaned up some
  32-bit dependencies in the code.

  Fixed regression: `Word32.toInt 0wx8002DE32;` would return `187954` instead
  of raising `Overflow`.  The problem was a mistake in the way that the overflow
  trap was being generated in `MLRiscGen`.

  Add fixedIntSz to TARGET signature

  Changed Primop.REAL to Primop.INT_TO_REAL.

  minor tweak to installml script

  Minor tweaks to CoreAccess

  Some minor primop cleanup.  Changed the type of Primop.ROUND and Primop.REAL
  to take bitwidths, instead of numkinds, since the kinds are always the same.
  Renamed ABS to FABS, since it is only used on floating-point numbers.
  Renamed the CPS primitive operator ROUND to REAL_TO_INT and the operator
  REAL to INT_TO_REAL.

improve implementation of Word64 multiplication

  Improvements to the core 64-bit int and word modules in system/smlnj/init.
  Replaced Int64.+, Int64.-, Word64.+ with versions from "Hacker's Delight"
  that use fewer conditional branches.  Also replaced the relational
  operators (<, <=, etc) with more direct implementations.

  Fix for bug #213 (`Int32.div` raises `Div` instead of `Overflow`
  when dividing `minInt` by `~1`).

  Removed the old SPARC assembly code for multiplication and
  division.  The code generator always uses the native hardware
  instructions, so the assembly code is not needed.

  Reimplemented "use" to fix bugs #193, #217, and #219.

  Moved the `F_SGN` operator (which is unary) from the `fcmp` datatype
  in `CPS.P` to the `branch` (and renamed it `FSGN`).

  finishing up the CPS conversion

adding cps-comp.sml

tweaking CPS representation

  Converting CPS data constructors to ASDL version

  minor code cleanup

  Removed default int and real operations from CPSUtil, since they were not being
  used.  Also remove some bogus 31-bit specific code from convert.sml

  Starting to migrate the CPS IR toward the ASDL version.  Changed the
  names of the CPS.P.arith and CPS.P.cmpop constructors to be upper-case
  alpha IDs (many of them were symbolic identifiers).  Also split out the 
  various utility functions into the new CPSUtil module (CPS/cps/cps-util.sml).
  Lastly, moved the literals.sml file from FLINT/main to CPS/main (where
  it belongs).

  code cleanup for cpstrans.sml; fix layout, add comments, and use
  some Basis functions to make the code more concise

  More code reorganization.  Moved FLINT to CPS conversion and CPSTrans
  phase to new CPS/convert directory.  The CPS/cps directory now only
  contains the representation of CPS and some utility code

  Split CPS out from the FLINT directory into its own subdirectory.  This change
  includes replacing the FLINTComp functor with a FLINTOpt structure and a CPSComp
  functor.

add info about Mojave build

tweaking the mojave build process

  Change SwapBytes to SwapBytes32 in cygwin.def

  Use xcode-select to get the path to the developer directory

  Adding support for building the runtime on Mojave.  To do so, you need to
  copy the MacOSX10.13.sdk directory into the Xcode 10.1 SDKs and then use
  the mk.x86-darwin18 makefile.

  mark unused primops

  The getspecial primop can be implemented using the same code as getobjlength.

  moved description of how conversions work to dev-notes.

  Removed several unsupported primitive operators from the compiler.
  See the changelog for details.

  Fix for bug #216 (run-time system fatal error with large top-level value)
  Also changed the way that available space is tracked and handled in the
  literal builder.  Instead of trying to track the amount of available space
  in a local variable, we recompute it for each new instruction.

  A bunch of comment and code layout edits

remove unused primops

  Mark unused primitive operators (these are 8-bit operations)

  remove trailing whitespace

  Remove trailing whitespace

  a bit more information in a compiler bug message

  added header comment

added missing header comments

  remove fast-fp control flag from AMD64, since it is an x86-specific control

  Adding header and 64BIT fixme comments.

  edit comments

  Preparing for a 64-bit basis; rename built-in.sml to built-in32.sml

  Enabled compiling of the AMD64 backend when compiling the compiler (it compiles but
  doesn't yet work).  Fixed issues with the floating point registers on the AMD64.

fixed typo

note that x86-fast-fp may be a bogus control

adding some 64BIT comments

  code layout

remove dead code

  Change CPS operators for wrapping/unwrapping integer and float values to
  be word-size flexible.  We now use a single `wrap` (and `unwrap`) operator
  that is parameterized by a `numkind` value.

  Changed CPS pure operators `wrap` and `unwrap` to `box` and `unbox` to avoid
  colflict with new names for `iwrap`, etc.

minor edit

  More 64-bit cleanup for mlriscGen.sml

more 64-bit cleanup

  Switch GC Types to be ML-value size independent

  More 64-bit cleanup of mlriscGen.sml and fixed a minor bug in tracking
  GC types.

  Working on 64-bit cleanup of code generation

trailing whitespace

code formatting

no need for int31/int63 prim types (use int instead)

update MAP

removed Semant/statenv directory

  Remove unused primopmap.sml file from FLINT/trans.

64BIT fixes to pequal.sml, equal.sml, etc.

adding some comments

  merging in changes from the arith64 branch: mostly removal of the broken
  CPS checking code plus more "64BIT:" comments.

remove unused expandNEW.sml file

  some code formatting and merged in 64-bit cleanup of IntInfCnv (nee InfCnv)  module.

  Replace the implementation of switch with a cleaner version.

  improved comments

clarify comments

fixed (most) 64BIT issues in FLINT/trans/translate.sml

minor updates to FLINT/MAP

64BIT fix for FLINT/trans/pequal.sml

  Added "64BIT" comments

  Last minute bug fix for 110.84; the subslice function had a typo in it.

fix error in windows installer

  Reimplementation of the array/vector-slice modules to use a (base, start, length)
  representation (as does Substring in system/smlnj/init/substring.sml).  Also
  fixed a bug in the findi functions, where the index being passed to the predicate
  function was not adjusted to be slice-relative.

adding SMLNJ-ASDL-LIB to pathconfig

  add proxy for ASDL library

  Fix header comment

  Fix incorrect signature name

  better error messages

  signatures for SML Basis Library proposal 2018-002 (Additional slice operations)

  update copyright

  Improved implementation of CharVectorSlice.{map,mapi} and implementation of SML Basis
  Library propoosal 2018-002

  addd header comment

latex fix for btcomp

  Various tweaks to the CM documentation.

  Switch most verbatim blocks to lstlisting blocks.

update document history

  merging in changes from cm-tool-hacking branch

  Fixed bug in implementation of monomorphic buffers (CharBuffer and Word8Buffer).

  Working on code cleanup for 64-bit support.

trailing whitespace

fix bug with error propagation

code layout/comments

  change representation of integer types in FLINT to be target-word-size
  independent

  remove OBJECT constructor from CPS.value datatype, since it is not used

remove condmove primop from CPS

add FIXME comment

edit comment

comments etc.

fix for bug 206

correct typo in comment

  Fixed minor bug in Date.toString (missing leading "0" for day of month).

cleanup, documentation, and debugging support for match compiler (FLINT/trans/matchcomp.sml, etc.)

  Fixed parser to allow parentheses around `val rec` patterns.  This issue was bug
  number 1487 in the old bug list.

  Fixed lexer to produce the correct error message for bad escape sequences.
  Previously, it produced an "unclosed string" message.  This bug was bug number
  1551 in the old bug list.

  Fixed open bug number 1383: Char.toCString #"\000" returned "\\0", instead
  of #"\\000", which caused String.toCString to produce invalid results.

adding some signature comments

  Added an implementation of SML Basis Library proposal 2018-001
  (Addition of monomorphic buffers).

  Fix a bug where "0w" was being accepted as a prefix for a hexidecimal
  word value in Word.fromString/scan (ignoring case, only "0x" and "0wx"
  are valid prefixes).

  Fix parsing issue with 'op' (bug1370)

  Tweaking error messages to make them more consistent

  bug fix: lexer gave unmatched close comment error on "*)", when it should have
  scanned it as "*" ")".

use ListPair.foldl instead of List.foldl + ListPair.zip

code layout

code layout

  code layout

  The sameSign function returned incorrect results in the Int31 and Int32 modules.

  Fixed various minor parsing and scanning issues:
    - correct syntax for type variables
    - signature/structure/functor IDs should always be alpha IDs
    - the equality ID ('=') cannot appear in a binding context

added FIXME comment

adding comments

removed FLINT/const-arith directory

removed all files in FLINT/const-arith

  Cleanup switch code a bit.

  Switch CPS to use IntInf.int to represent integer/word values

better classification of real literals

code layout

  Fix bug in handling of zero with large exponents

code formatting

  Fixed bug with converting strings to real literals and added
  support for Successor ML's use of "_" as a numeric separator.

  improvements to the Basics/reals code

const-arith moved to MiscUtil

  Switch FLINT IR to use IntInf.int for integer and word constants.

resolved tree conflict on FLINT/const-arith

  working on numeric constants

  Added contraction rules for unsigned *, +, and -

  moved const-arith into MiscUtil so that it is available to the
  CodeGen phase (in addition to cpsopt).

  code formatting

  Changed representation of integer/word constants in PLambda to
  use IntInf.int.

  Switch to using IntInf.int to represent object descriptors in
  the code generator.  Part of the transition toward 64 bits.

  documentation edits

deleted PACKexp from Absyn and other files referring to it

  Working on integer/word literals stuff

  Use RealConst.t representation in FLINT and CPS IRs.

  Propogate the source string for numeric literals into the
  Absyn so that it can be used in error messages.

  moved literal-to-num.sml to the FLINT/trans directory, since that is
  the only place that it is used.

  Changed Absyn representation of real constants to prepare
  for eventual 32-bit float support.

  code formatting and comments

  fix representation of 0.0 to be the empty list of digits

  Fix for bug #191 (Compiler crash when handling large reals).  We now issue a warning
  for real literals that will round to zero and an error for real literals that are
  too large to represent.  There still needs to be done some work to support sub-normal
  literal values (these are currently rounded to 0).

  Adding library to support constant integer/word arithmetic at
  arbitrary precisions.

Select -std=c99 since gcc does not understand the inline keyword in -ansi mode.

  Switched the representation of real literals from strings to the
  RealLit.t type.  This change is the first step toward fixing
  bug #191 and supporting 32-bit reals.

  Removed real patterns from Absyn and FLINT, since they are not allowed by SML'93.

old-literals.c as bridge

  Fixed bug 194 (Real.fromString overflows or hangs)

  Preparing the way for a new implementation of the build-literals mechanism.
  Move the version-1 interpreter into old-literals.c.  Also added Int64_t and
  Unsigned64_t typedefs.
  See dev-notes/new-literals.md for a description of the new-literals bytecode.

  adding some documentation

  Some documentation and simplified the interface to the Literals
  module in FLINT.

  bug fix for literal range checking.  We need to do *all* overload resolution
  before checking ranges, since we might have literals whose type variable has
  been instantiated to another overloaded type varible that has not yet be resolved.

prune type in numInfo and better tyToString

  fix typo

  Pickling support for IntInf.int values; for now we just pickle them as
  strings.

  bug fix for windows: use SayDebug for _ml_win32_debug function

  Add checking of numeric literal ranges to overload resolution.

  Add some utility functions for dealing with int/word literals

  No longer compiler the file FLINT/opt/abcopt.sml, since its optimization
  is never enabled.

added comment to cmb-make

adding comment

  Remove "open" and add some comments to overload.sml

  Working on change in integer/word literal representation.
  Stage 1 -- changed representation in Absyn.

  Adding common representation of integer and word literals.

  fix bug (typo)

  Some restructuring/cleanup for literal overloading.  Got rid of Types.litKind
  datatype; changed Type.ovldSource datatype to have different constructors for
  each kind of literal that can be overloaded.

  change EOL comment to regular comment to avoid pre-processor warning
  on linux.

added/revised some comments regarding FE primops; made primop_bind type abstract

update header

  Fixed a bug in IntInf.div and IntInf.mod, where the Div exception
  was not raised when both arguments are 0.

minor code cleanup

add FIXME comment

  Changed representation of CPS.cty to track sizes of untagged integer
  and float types.

  add unpickler bug fix that got clobbered by accident

changed cty in FLINT/cps/cps.{sig,sml}

add mlValueSz and comments to target modules

trailing whitespace

fix type errors

  Bug fix.  Also add headers and remove trailing whitespace.

minor fix to tyToString in TypesUtil

added a crude tyToString function to TypesUtil

  fix bug in printing

removed old debugging print statements in elabOVERLOADdec in elabcore.sml

bug fixes for new primops

  Working towards 64-bit support.  Added Target module, which specifies the
  properties of the target (e.g., size in bits of default int type).  Reworked
  generation of InlineT structure to be target specific.  Some cleanup for
  primop bindings and FLINT's primtycs.

minor edits of primtyc.sig and pickmod.sml

getting rid of ptnum stuff

  Merge CoreBasicTypes into BasicTypes

  working on primops

minor edit

  Working on new primops.  We've moved a bunch of code from Semant to ElabData
  and we've changed the representation of primop IDs in structures.

working on new primops

updated FLINT/trans/translate for new primop-bindings

updated ElabData/basics/primop.{sig/sml} with new primop_bind

moved primop.{sig/sml} from FLINT/kernel to Semant/prim

working on new primops

working on new primops

working on new primops

redeleted Semant/types

comment editiing in FLINT/kernel/primop.sml

new primop representation

  remove BOXEDUPDATE primop, since it is the same as UPDATE.

  Removed unused primops (GETRUNVEC, DEFLVAR, USELVAR)

fix for bug 123, updated documentation for primops

  Fix bug #183 (status code returned by sml REPL).  This restores the
  110.79 behavior of having "sml foo.sml" exit with a non-zero status
  when there is a type-checking error in foo.sml.  It also cleans up
  the error messages associated with "use" when there is a syntax
  error.

  add some comments about bug #173

minor 64-bit prep

fix for gen-posix-names.sh

  Added missing options (@SMLsuffix and @SMLversion) to help message
  (bug #185).  Also reordered the list of options to match the Manual
  page (some reordering of the manual page was also done).

code formatting

  For now, we remove AMD64 from the visible compiler since
  it is incomplete.

adding header comment