path: root/NEWS

mcpu-cpp 1.0.3
===============

Align the per-user configuration hierarchy with the common MCPU environment
and add an explicit command-line system-root override.

* Move the normal per-user configuration from `$HOME/.mcpu/mcpu-cpp.conf` to
  `$HOME/.mcpu/etc/mcpu-cpp.conf`; the obsolete location is no longer read.
* Add `--sys-root=PATH`.  PATH may be absolute or relative; relative paths are
  resolved from the invocation working directory.
* `--sys-root` makes `PATH/include` the effective system include root and
  suppresses every configuration-file layer, including `--config-file`.
* Keep the relocatable executable-derived root and `--no-config` behavior
  unchanged when `--sys-root` is not present.
* Synchronize command-line help, bilingual Markdown manuals and man pages.

mcpu-cpp 1.0.2
===============

Implemented LibMPU, LibMPUIO congiguring check using libmpu.m4, libmpuio.m4.


mcpu-cpp 1.0.1
===============

Integrate the mcpu-cpp(1) manual pages.

* Add English man/mcpu-cpp.1 and Russian man/ru/mcpu-cpp.1.
* Install the pages through Automake's standard man directory hierarchy.

mcpu-cpp 1.0.0
===============

Remove Russian preprocessing-directive aliases and prepare the public interface
for another verification cycle before 1.0.0.

* Accept only canonical English preprocessing-directive names; Unicode support
  for identifiers and ordinary source text is unchanged.
* Remove Russian-alias regression cases without changing the underlying macro,
  conditional, include, diagnostic, language-stack, or output engines.
* Retitle manual section 10 as "GNU-compatible features" and summarize the
  documented behaviors intentionally aligned with GNU CPP.
* Remove references to the predecessor preprocessor from both normative manuals.

mcpu-cpp 0.0.50
===============

Refresh and bilingualize the normative manual and add a developer bootstrap.

* Replace the single doc/mcpu-cpp.md with synchronized Russian and English
  manuals: doc/mcpu-cpp-ru.md and doc/mcpu-cpp-en.md.
* Bring the manual up to the 0.0.49 feature set, fix stale future-work text and
  conditional-expression subsection numbering, and document the current
  intentionally unsupported legacy features.
* Add a root bootstrap script, modeled after the LibMPU/LibMPUIO bootstrap
  workflow, that regenerates the ZUBR parser and Autotools-generated files.
* Add a root .gitignore describing files that a developer Git tree may omit;
  keep release archives self-contained and the ordinary release build unchanged.

mcpu-cpp 0.0.49
===============

Compact invisible output lines while preserving exact source coordinates.

* Track the next visible logical source position instead of materializing every
  consumed directive/comment/blank source line in .E output.
* Match GNU CPP's 0..7 newline versus 8+ corrective-linemarker boundary.
* Preserve include entry/return markers and avoid synthetic progress markers
  for completely silent included files.
* Keep __LINE__, diagnostics, #line and logical coordinates independent of the
  compacted physical output representation.

mcpu-cpp 0.0.48
===============

Normalize whitespace that belongs to macro replacement lists.

* Canonicalize replacement-list whitespace to one ASCII space after line
  splicing while preserving ordinary source formatting and actual-argument
  whitespace.
* Preserve literal contents and preprocessing-token boundaries around #, ##,
  __VA_ARGS__, __VA_OPT__ and placemarkers.

mcpu-cpp 0.0.47
===============

Add the standard __VA_OPT__(pp-tokens) mechanism.

* Decide variadic emptiness after normal macro substitution.
* Support balanced parentheses, stringification, token concatenation,
  placemarkers and rescan integration.
* Intentionally omit GNU named variadics and the historical comma-swallow
  extension , ## __VA_ARGS__.

mcpu-cpp 0.0.46
===============

Add C99-style variadic function-like macros.

* Accept ... as the final macro parameter and substitute __VA_ARGS__.
* Preserve ordinary prescan, raw stringification and token-paste semantics for
  the complete variadic tail, including empty-tail placemarkers.
* Keep the old GNU named form args... and __VA_OPT__ outside this release.
* Preserve variadic definitions in macro dumps and document the exact contract.
* Add dedicated regression coverage for expansion, #, ##, dumps and errors.

mcpu-cpp 0.0.45
===============

Add GNU-like missing-generated dependency handling.

* Implement -MG for dependency-only -M/-MM modes.
* Record unresolved includes in the existing ordered dependency registry while
  keeping their textual identity separate from physical st_dev/st_ino identity.
* Preserve GNU-like user/system filtering for missing quoted/angle includes and
  system-header context; tolerate missing -include/-imacros operands under -MG.
* Keep -MG invalid with -MD/-MMD or without -M/-MM.
* Document the unresolved-dependency model and add dedicated regression coverage.

mcpu-cpp 0.0.44
===============

Add explicit dependency targets and complete the forced-file documentation.

* Implement GNU-like -MT TARGET and -MQ TARGET, including attached forms.
* Allow repeated -MT/-MQ options and multiple targets in one dependency rule.
* Keep -MT literal while -MQ applies Make quoting; suppress the default target
  whenever an explicit target is present.
* Document -include/-imacros and -MT/-MQ in doc/mcpu-cpp.md.
* Add dedicated dependency-target regression coverage.

mcpu-cpp 0.0.41
===============

Clean the public interface and formalize selected Russian directive aliases.

* Remove obsolete compatibility options from the command-line parser so they
  are diagnosed as unknown options instead of remaining dormant handlers.
* Keep -E as the intentional compiler-driver compatibility no-op.
* Document --object-suffix in --help and require its argument.
* Add Russian aliases for define, line, include, include_next, undef, pragma
  and lang; normalize non-once #управление output to #pragma.
* Rename the product description to "MCPU languages preprocessor" and remove
  predecessor-oriented wording from source comments and documentation.
* Add interface-cleanup and Russian-alias regression coverage.

mcpu-cpp 0.0.40
===============

Add explicit warning-control semantics.

* Implement -Wcomment/-Wcomments and negative forms for nested block-comment
  openers and backslash-newline continuation of // comments.
* Make -Wall enable the optional comment-warning class while preserving the
  priority of an explicit -Wno-comment regardless of option order.
* Implement -Werror/-Wno-error for every warning actually emitted by mcpu-cpp,
  without making -Werror enable optional warning classes.
* Route #warning, zNNN width warnings, macro redefinition and invalid ## paste
  through the common warning policy.
* Add dedicated regression coverage and a normative documentation section.

mcpu-cpp 0.0.39
===============

Split final macro dumping into ordinary and predefined views.

* Make -dM dump only source/include and command-line macros.
* Implement -dMP as predefined macros first, followed by ordinary macros.
* Preserve deterministic name sorting inside each group and final #undef state.
* Keep context-dependent special macros out of the static dump.
* Leave -dD semantics unchanged.
* Add dedicated regression coverage and document the macro-origin contract.

mcpu-cpp 0.0.38
===============

Extend dependency generation with side-effect dependency files.

* Add GNU-like -MD and -MMD without suppressing normal preprocessing output.
* Add -MF FILE, including attached -MFfile spelling and -MF - for stdout.
* Derive the default .d filename from the input basename, or from ordinary -o
  output when present.
* Reuse the existing -M/-MM dependency graph and system-header classification.
* Add dedicated regression coverage and document the dependency-output rules.

mcpu-cpp 0.0.37
===============

Make the MCPU installation tree relocatable and shared by the future toolchain.

* Install under `$libdir/mcpu` rather than `$libdir/mcpu-VERSION`.
* Resolve the real executable at run time and derive `<root>/include` and
  `<root>/etc/mcpu-cpp.conf` from the physical MCPU root.
* Keep `/etc/mcpu/mcpu-cpp.conf` and `$HOME/.mcpu/mcpu-cpp.conf` as higher
  priority overrides, including full replacement of MCPU_CPP_SYSTEM_INCLUDE_PATH.
* Remove the absolute installation default from the packaged config and runtime
  binary contract.
* Add a regression that physically moves a complete MCPU tree and verifies the
  new root, include search, packaged config and symlink invocation.
* Document one relocatable `bin/etc/include/lib` root as a general MCPU
  ecosystem principle for mcpu-cpp, mcpu-as, mcpu-ld and mcpu-run.

mcpu-cpp 0.0.36
===============

Correct --no-config so it suppresses configuration-file reads without
discarding the compiled installation defaults.

* Keep the compiled MCPU_CPP_SYSTEM_INCLUDE_PATH active under --no-config.
* Make -dsearch-dirs, -dconfig and -v expose that built-in default even when
  every configuration file is ignored.
* Preserve explicit -I/-isystem/-idirafter ordering around the compiled system
  tree and keep -nostdinc as the switch that suppresses standard-system paths.
* Add dedicated regression coverage and document the distinction between
  compiled defaults, configuration files and -nostdinc.

mcpu-cpp 0.0.35
===============

Implement the first dependency-generation stage.

* Add -M make-rule output including system headers.
* Add -MM user-dependency output excluding system headers and descendants
  reachable only from system-header contexts.
* Collect dependencies directly from the existing include/#include_next
  pipeline and deduplicate physical files by device/inode identity.
* Keep #line logical filenames out of dependency output and preserve a file in
  -MM when it is also reached through a user include context.
* Add dedicated dependency-generation regression coverage and documentation.

mcpu-cpp 0.0.33
===============

Align the configure-time console presentation with the LibMPU/LibMPUIO/ZUBR
package family and normalize the author's given name spelling.

* Add the shared-style mcpu-cpp configure headline and AC_MSG_CFG_PART section
  headings to make configure output follow the same visual structure as the
  other MCPU/LibMPU projects.
* Keep the existing detailed configuration summary as its own final section.
* Normalize every textual occurrence of the author given name to `Andrey`
  throughout the maintained project tree.

mcpu-cpp 0.0.32
===============

Polish include-path diagnostics after the 0.0.31 installation/search contract.

* Make -v print each effective include configuration variable once, after all
  configuration layers have been resolved.
* Add -dsearch-dirs to print effective include search directories in semantic
  priority order and stop without preprocessing.
* Document the distinction between effective configuration values and the
  expanded global search-directory chain.

mcpu-cpp 0.0.31
===============

Introduce the versioned MCPU installation/configuration contract.

* Install mcpu-cpp under `$libdir/mcpu-0.0.31/bin` and its packaged
  configuration under `$libdir/mcpu-0.0.31/etc`.
* Install `$bindir/mcpu-cpp` as a symbolic link to the versioned executable.
* Do not install or create `/etc/mcpu`; treat `/etc/mcpu/mcpu-cpp.conf` only as
  an optional distributor/administrator override.
* Load configuration in increasing priority: compiled versioned defaults,
  versioned packaged config, optional `/etc` override, then the unversioned
  `$HOME/.mcpu/mcpu-cpp.conf` user override.
* Restore `MCPU_CPP_SYSTEM_INCLUDE_PATH` as the replaceable effective system
  include root, defaulting to `$libdir/mcpu-0.0.31/include`.  Derive
  `<root>/<lang>` internally; no per-language system variables exist.
* Let a higher-priority config replace the system tree completely or disable it
  with an empty value.
* Keep the normative include and #include_next search order established in
  0.0.30, now using the effective `MCPU_CPP_SYSTEM_INCLUDE_PATH`.
* Document the versioned tree, configuration precedence, sandbox workflow and
  wrapper-header rationale as normative interfaces.

mcpu-cpp 0.0.30
===============

Define and enforce the normative include-search contract discovered by manual
wrapper-header testing.

* Give explicit command-line -I, -isystem and -idirafter directories priority
  over persistent configuration classes.
* Search configured user language paths before MCPU_CPP_INCLUDE_PATH.
* Replace MCPU_CPP_SYSTEM_INCLUDE_PATH with the singular
  MCPU_CPP_SYSTEM_INCLUDE_DIR system-tree root.  Derive the fixed <root>/<lang>
  directory internally and then search <root>; no per-language system config
  variables exist.
* Allow an empty MCPU_CPP_SYSTEM_INCLUDE_DIR to disable the configured system
  tree.  -nostdinc suppresses the same configured system classes for one run
  without suppressing explicit -isystem.
* Keep explicit -idirafter before MCPU_CPP_AFTER_INCLUDE_PATH and never derive
  language-specific AFTER directories.
* Make #include_next traverse this exact effective chain by physical provenance.
* Add a dedicated manual-style regression that places the same header in all
  eight search classes and verifies the complete #include_next order.
* Add a separate normative documentation section explaining the search order,
  configurable user paths, fixed system subdirectory names and wrapper-header
  purpose.

mcpu-cpp 0.0.29
===============

Implement GNU-style #include_next wrapper-header search.

* Track the exact configured include-directory entry that supplied each nested
  header and continue #include_next strictly after that entry.
* Preserve the semantic search order across -I, explicit -isystem, configured
  system directories and -idirafter directories.
* Do not re-search the current source-file directory for #include_next; the
  directive intentionally does not distinguish between "file" and <file>.
* Support macro-expanded #include_next operands through the ordinary include
  operand parser and macro-expansion path.
* Keep physical include provenance independent of logical #line filenames.
* Add regression coverage for the wrapper -> configured-system -> idirafter
  chain, quoted-source origin, macro operands, inactive branches and failure
  when no later header exists.

mcpu-cpp 0.0.28
===============

Implement #error and #warning diagnostics.

* Active #error emits a source-location error diagnostic and stops preprocessing.
* Active #warning emits a source-location warning and preprocessing continues.
* Diagnostic directive arguments are not macro-expanded, following the directive contract; whitespace between tokens is folded while quoted text is kept.
* Diagnostic directives are consumed and never copied to normal or -dD output.
* Inactive conditional branches suppress both directives completely.
* Historical Russian aliases #ошибка and #предупреждение are recognized.

mcpu-cpp 0.0.27
===============

Define the MCPU integer-width suffix contract for conditional expressions and
remove the obsolete assertion command-line heritage.

* Accept `zNNN`/`ZNNN` and optional trailing `U`/`u` on integer constants.
  NNN is always decimal, including forms with leading zeroes.
* `z8`, `z16`, `z32` and `z64` normalize the low N bits by sign extension;
  an optional U/u selects zero extension.  All later expression evaluation is
  still exactly 64-bit and does not retain the source width.
* Widths above 64 are errors in conditional directives.  Invalid widths up to
  64 are ignored with a warning; a trailing U/u remains effective.
* Remove C-style L/LL integer suffix acceptance from the #if lexer.
* Diagnose a suffix that runs into an identifier instead of splitting it into
  unrelated tokens.
* Remove the obsolete `-A` assertion option, the assertion action kind and the
  `-dMA` assertion-dump modifier.
* Document the complete normalization and 64-bit evaluation model.
* Add regression coverage for width suffixes, sign/zero extension, truncation,
  64-bit post-normalization operations, errors and removed assertion options.

mcpu-cpp 0.0.26
===============

Use the ZUBR grammar architecture for #if expression parsing.

* Add mcpp-expr.zubr with the grammar and UCS-2 lexical analyzer; generated
  mcpp-expr.c uses the mcpp_zubr_* prefix from ZUBR 4.1.0.
* Keep the established defined() preprocessing and macro-expansion front end.
* Move fixed-width #if arithmetic to mcpp-semantic.c/h; expression integers
  use LibMPU 64-bit types rather than host long/intmax_t.
* Parse binary/octal/decimal/hexadecimal integer lexemes through LibMPU
  iatoui() after validating/copying the ASCII subset from UCS-2.
* Character units are UCS-2 __mpu_uint16_t values; multi-character
  packing is retained up to the 64-bit expression width.
* Preserve the defined precedence, negative-shift direction and short-circuit
  behavior for &&, || and ?:.
* Ship the generated C source; ZUBR is needed only when the .zubr grammar is
  regenerated.  The obsolete code-page -T option is not used.
* Add a dedicated ZUBR expression regression.

mcpu-cpp 0.0.25
===============

Refine preprocessing identifiers and command-line macro declarator parsing.

* Permit `$` after the first preprocessing-identifier character, while keeping
  `$` forbidden as an identifier start.
* Apply the same `$` continuation rule throughout macro names, parameters,
  conditionals, expansion, # stringification and ## rescanning.
* Parse the -D declarator separately from its replacement value so an invalid
  tail before '=' is discarded instead of becoming replacement text.
* Apply the same valid-identifier-prefix rule to -U command-line actions.
* Document shell quoting explicitly: single quotes already protect `$`; an
  unquoted or double-quoted `$` must be escaped when literal text is intended.
* Extend Unicode/command-line macro regressions for `$`, invalid tails, -U and
  forbidden `$` at identifier start.

mcpu-cpp 0.0.24
===============

Correct -dD origin markers for command-line macro definitions.

* Macros installed by -D are tagged as command-line definitions internally.
* -dD emits `# 0 "<command-line>"` before command-line definitions instead of
  incorrectly reporting them as `# 0 "<built-in>"`.
* Predefined/builtin macros retain the `# 0 "<built-in>"` marker.
* Added a regression covering Unicode -D names and both marker classes.

mcpu-cpp 0.0.23
===============

Implement command-line macro definition and undefinition.

* -DNAME defines NAME as 1; -DNAME=VALUE defines VALUE, including an empty
  replacement after an explicit '='.
* Function-like command-line macros use the same parser and macro engine as
  source #define directives, including # stringification and ## concatenation.
* -D and -U payloads alone are decoded from UTF-8 to strict UCS-2; filenames,
  include paths and all other command-line arguments remain byte strings.
* Unicode command-line macro names use the same XID_Start/XID_Continue rules
  as source identifiers.  Invalid UTF-8 and non-UCS-2 characters are rejected.
* -D/-U actions are applied in command-line order after predefined/builtin
  macros are installed, so -U may remove a predefined macro and later -D may
  define it again.
* Added dedicated regression coverage for ASCII and Unicode command-line
  macros, order, function macros, #/##, conditionals and invalid encodings.

mcpu-cpp 0.0.22
===============

Use LibMPUIO 1.0.4 strict UCS-2 Unicode 18.0.0 XID classification for
preprocessing identifiers.

* Identifier start is `_` or Unicode XID_Start.
* Identifier continuation is `_` or Unicode XID_Continue.
* Macro names and parameters may therefore use UCS-2 letters from supported
  scripts, combining marks in continuation positions, and non-ASCII decimal
  digits in continuation positions.
* The same identifier rules are used by #define/#undef, #ifdef/#ifndef,
  defined(), ordinary macro expansion, stringification, token concatenation
  rescanning.
* Classification is locale-independent and inherits LibMPUIO strict UCS-2
  semantics; surrogate code units are never identifiers.
* Configure now verifies that the required LibMPUIO UCS-2 ctype API is present.

mcpu-cpp 0.0.21
===============

Implement GNU-compatible token concatenation (##).

* ## now pastes preprocessing tokens in both object-like and function-like
  macro replacement lists.
* A formal parameter adjacent to ## uses its raw, unexpanded actual argument;
  the completed replacement list is rescanned afterwards.  This preserves the
  standard two-level CAT/XCAT expansion technique.
* Empty pasted arguments use placemarker semantics, and multi-token arguments
  paste only the token adjacent to ##.
* Pasting can form identifiers, preprocessing numbers, multi-character
  punctuators and prefixed string/character tokens.
* An invalid paste is diagnosed and the original two tokens are emitted; ## at
  either end of a replacement list is rejected when the macro is defined.
* Stringification (#) and concatenation (##) may be used in the same macro.
* Added dedicated token-concatenation regression coverage and updated macro
  error tests.

mcpu-cpp 0.0.20
===============

Corrective release of the quoted #lang contract introduced in 0.0.19.

* The #lang implementation contract is unchanged: exactly one quoted language
  name, no whitespace inside the quotes, no escape processing, same-physical-line
  termination, ASCII case-insensitive validation against the internal language
  table only, original spelling preserved, and normalized external whitespace.
* Regression coverage explicitly checks the accepted spellings "diff", "Diff",
  "DIFF" and "dIfF", as well as all other internal languages.
* All package-version regression expectations are synchronized with 0.0.20;
  this fixes the stale-version test failures seen in the first 0.0.19 test run.

mcpu-cpp 0.0.19
===============

Tighten and normalize the #lang directive contract.

* #lang now requires a quoted language name, for example `#lang "diff"`;
  the unquoted form is rejected.
* The quoted text must be one non-empty word without whitespace, must close on
  the same physical source line, and may be followed only by whitespace.
* Escape sequences are not interpreted inside the #lang string.
* Language lookup is ASCII case-insensitive but is restricted to the internal
  language table: diff, dift, alg, as, avm and ACS.
* Output preserves the spelling inside the quotes while normalizing external
  whitespace to exactly `#lang "name"`.
* Updated existing language tests and added a dedicated #lang string-contract
  regression.

mcpu-cpp 0.0.18
===============

Implement conditional-compilation control flow.

* #if, #ifdef, #ifndef, #elif, #else and #endif are consumed by the
  preprocessor and never copied to normal or -dD output.
* Inactive branches are skipped without executing definitions, includes or
  ordinary macro expansion; nested conditionals remain balanced.
* #if expressions follow the expr.zubr precedence model, including
  defined, unknown identifiers as zero, integer operators, ?: and short-circuit
  evaluation.
* Include files keep independent conditional-stack boundaries and diagnose
  unbalanced or unterminated groups.
* Generated source-location information now uses GNU linemarkers (`# N "file"`)
  instead of emitting `#line`; include entry/return markers carry flags 1 and 2.
* Input #line directives are consumed, macro-expanded and update __LINE__ and
  __FILE__; the resulting location is emitted as a GNU linemarker.
* Added regression coverage for include guards, nested groups, all conditional
  directives, defined, macro-expanded expressions, malformed groups and line
  control.

mcpu-cpp 0.0.17
===============

Finish comment whitespace cleanup.

* Removing a // or /* ... */ comment at the end of a non-empty source line no
  longer leaves a trailing blank before the newline.
* The same rule applies when a block comment begins after program text and
  continues onto following physical lines.
* Inline comments between tokens still preserve a separating blank where it is
  required to prevent accidental token concatenation.

mcpu-cpp 0.0.16
===============

Implement -dD and correct comment/UCS-2 phase behavior.

* -dD preserves normal preprocessed output, emits predefined macro definitions
  with <built-in> markers, and keeps encountered #define directives.
* Comment-only lines are left truly empty after // or /* ... */ removal while
  inline comments still preserve token separation.
* Valid UTF-8 scalars outside UCS-2 are accepted inside comments and rejected
  only if they remain in program text after comment removal.
* Added regression coverage for -dD, comment-only whitespace and non-UCS-2
  characters inside versus outside comments.

mcpu-cpp 0.0.13
===============

Implement macro stringification (#).

* Function-like macro replacement lists now support #parameter stringification.
* Stringification uses the raw, unexpanded actual argument, trims leading and
  trailing whitespace, folds internal whitespace outside quoted tokens, and
  escapes quotes/backslashes in the generated string literal.
* Ordinary argument expansion is now lazy, matching the established raw-vs-expanded
  argument semantics and preserving nested calls of the same macro.
* Empty actual arguments are represented correctly; zero-argument macros keep
  their established behavior.
* Invalid # uses are diagnosed at definition time.  ## remains deliberately
  rejected for the next separate token-concatenation port.
* Added regression coverage for raw-vs-expanded stringification, whitespace,
  quoted text, empty arguments, diff apostrophe semantics, diagnostics and -dM.

mcpu-cpp 0.0.12
===============

Correct integer DECIMAL_DIG semantics.

* Replaced direct use of LibMPU _int_digs() for predefined integer precision
  metadata with mcpu-cpp integer-only helpers based on the known type width.
* __INT<bits>_DECIMAL_DIG__ now counts decimal digits of INT<bits>_MAX only;
  it excludes both the sign and any terminating NUL.
* __UINT<bits>_DECIMAL_DIG__ now counts decimal digits of UINT<bits>_MAX only;
  it excludes any terminating NUL.
* Added exact regression values for every integer family width from 8 through
  65536 bits.

mcpu-cpp 0.0.11
===============

Complete predefined precision metadata and MCPU byte/word-order naming.

* Added __INT<bits>_DECIMAL_DIG__ and __UINT<bits>_DECIMAL_DIG__ from LibMPU
  _int_digs() for every configured integer width through NB_I_MAX * 8; MAX
  values remain capped at 256 bits.
* Added __REAL<bits>_DECIMAL_DIG__ and __REAL<bits>_MANT_DIG__ for every Real
  width through MPU_REAL_IO_LIMIT; large numeric MAX/MIN/EPSILON/exponent
  values remain capped at 256 bits.
* Replaced __FLOAT_WORD_ORDER__ with __MCPU_BYTE_ORDER__ and
  __MCPU_WORD_ORDER__; __BYTE_ORDER__ remains as an alias to
  __MCPU_BYTE_ORDER__.
* Added __MCPU_INT_MAX_WIDTH__, __MCPU_REAL_MAX_WIDTH__ and
  __MCPU_COMPLEX_MAX_WIDTH__ capability macros.
* Renamed __SIZEOF_PTRDIFF_T__ to language-native __SIZEOF_PTRDIFF__.
* Extended predefined ABI/range regression coverage for the complete integer
  and Real families.

mcpu-cpp 0.0.10
===============

Predefined ABI naming cleanup and complete Real text-buffer metadata.

* Replaced the C-style public size family with __MCPU_SIZE_TYPE__, WIDTH,
  SIZEOF_SIZE and MAX; renamed __MCPU_SIZEOF_SSIZE_T__ to
  __MCPU_SIZEOF_SSIZE__.
* Renamed __SIZEOF_CHAR8_T__/__SIZEOF_CHAR16_T__ to the language-native
  __SIZEOF_CHAR8__/__SIZEOF_CHAR16__.
* Added __SIZEOF_REAL<bits>_EXP__ from LibMPU _sizeof_exp() for every Real
  width through MPU_REAL_IO_LIMIT.
* Added __REAL<bits>_MAX_STRLEN__ from _real_max_string() for every Real width;
  the value is a character count, not a byte count.
* Extended -dM regression coverage to verify the new names and the largest
  configured Real family.

mcpu-cpp 0.0.9
==============

Complete structural predefined metadata across LibMPU type ranges.

* Corrected MCPU ptrdiff to signed int64 with max 0x7fffffffffffffff.
* Added the MCPU-specific signed-size family: __MCPU_SSIZE_TYPE__, WIDTH,
  SIZEOF and MAX from the configured LibMPU __mpu_ssize_t profile.
* Integer TYPE/WIDTH/SIZEOF families now extend through NB_I_MAX * 8;
  numeric MAX/DECIMAL_DIG values remain capped at 256 bits.
* Real/Complex TYPE/WIDTH/SIZEOF families now extend through
  MPU_REAL_IO_LIMIT; numeric Real characteristics remain capped at 256 bits.
* Corrected __COMPLEX<bits>_WIDTH__ to the language type parameter <bits>;
  __SIZEOF_COMPLEX<bits>__ remains twice the corresponding Real storage.
* Added exhaustive -dM regression coverage across all advertised integer,
  Real and Complex structural families.

mcpu-cpp 0.0.8
==============

Predefined ABI refinement and derived system language include paths.

* Renamed __VERSION__ to __MCPU_CPP_VERSION__ and report the mcpu-cpp version.
* Renamed LibMPU-profile predefines to the __MCPU_* namespace.
* Character type spellings are now char8 and char16.
* MCPU pointer and ptrdiff contracts are fixed at 64 bits independently of host;
  ptrdiff is currently uint64 by ABI decision.
* Integer/Real textual value constants are emitted only through 256 bits.
  Wider families retain only type and width macros.
* Removed computed integer MIN predefines.
* Added integer DECIMAL_DIG values from _int_digs().
* Replaced REAL<bits>_DIG with REAL<bits>_DECIMAL_DIG from _real_digs();
  MANT_DIG comes from _real_mant_digs().
* Configured system include roots automatically add active-language
  subdirectories such as /usr/include/mcpu/diff without extra config variables.
* Added regression coverage for the refined predefined environment and derived
  system language include search.

mcpu-cpp 0.0.7
==============

Predefined ABI/environment layer and inspection tools.

* GNU GCC is now an explicit mandatory build compiler.
* Ported the LibMPU/LibMPUIO GCC type-size, endian and machine-register probes
  into mcpu-cpp acsite.m4 under project-specific macro names.
* Configure reads and verifies the installed LibMPU real-I/O/math limits,
  byte/word order, addressable-unit width, size_t and ptrdiff_t profile.
* Added _ARCH_MCPU and GCC-compatible byte-order, pointer, size_t/ptrdiff_t and
  assembler-prefix predefined macros.
* Added fixed-width int/uint predefined families through 65536 bits.
* Added Real/Complex type families through 65536 bits; Real numeric limits are
  generated with LibMPU and real_to_ascii up to MPU_REAL_IO_LIMIT.
* Large Real numeric predefined values are generated lazily so ordinary
  preprocessing remains fast.
* __VERSION__ is now the future high-level compiler version placeholder 0.0.0.
* Added `-dM` for deterministic macro-table dumps and `-dconfig` for effective
  configuration dumps; neither requires an input file.
* Added ABI, dump-macro and dump-config regression tests.

mcpu-cpp 0.0.6
==============

Implement predefined macros.

* Added the dynamic predefined macros __FILE__, __LINE__, __BASE_FILE__,
  __INCLUDE_LEVEL__, __DATE__, __TIME__ and __VERSION__.
* __FILE__/__LINE__/__INCLUDE_LEVEL__ follow nested #include processing;
  __BASE_FILE__ remains the original translation-unit input file.
* __DATE__ and __TIME__ are fixed once at preprocessing startup, matching the
  translation-unit timestamp model.
* __VERSION__ expands to the mcpu-cpp package version, as preprocessor used its own
  package version for the corresponding macro.
* Predefined expansions are emitted without rescanning, following the established
  special-symbol behavior.
* Built-ins share the normal macro table and can therefore be undefined or
  explicitly redefined.
* K-specific legacy built-ins (__STDK__, __K__, __kxLab__, and related names)
  are not carried into the MCPU language family.

mcpu-cpp 0.0.5
==============

Extend the macro engine.

* Moved project-owned Autoconf macros from m4/ to the root acsite.m4; m4/ is
  reserved for external/vendor macros.
* Documented the future developer-only Bison workflow: make parser, make tests,
  then make dist.  Normal release builds do not require Bison.
* Added function-like #define macros with zero or more formal parameters.
* Added nested argument parsing with defined parenthesis/comma rules.
* Actual arguments are macro-expanded before ordinary substitution, including
  nested calls such as min(min(a,b),c).
* Added diagnostics for duplicate parameters, malformed definitions, incomplete
  calls and wrong argument counts.
* # and ## are explicitly rejected until their semantics are implemented, avoiding silent partial preprocessing.

mcpu-cpp 0.0.4
==============

Implement the first macro-processing layer.

* Added doc/mcpu-cpp.md as a living implementation-order manual as the normative MCPU preprocessing manual.
* Added the preprocessing phase needed before directive and macro processing:
  backslash-newline splicing and comment removal with preserved source lines.
* Added object-like #define and #undef.
* Added recursive/cascaded object-macro expansion with recursion blocking.
* Macro names are not expanded inside quoted strings; the historical diff
  apostrophe rule remains in force.
* Added computed #include through macro expansion.
* Preserved literal #include <...> lexical behavior for comment-like text.
* Began reusing macro regression cases; the self-referential macro
  case is now part of the mcpu-cpp test suite.

mcpu-cpp 0.0.3
==============

Include-root and configuration-layout correction.

* MCPU_CPP_INCLUDE_PATH is now a common user include root visible in every
  language state.
* Active language-specific include directories are additional search paths
  and precede the common configured root.
* Explicit names such as <diff/a.h> can therefore be used through the common
  root regardless of the current #lang state.
* The system configuration moved to /etc/mcpu/mcpu-cpp.conf.
* The per-user configuration moved to $HOME/.mcpu/mcpu-cpp.conf.

mcpu-cpp 0.0.2
==============

Language-contract cleanup before macro-engine development.

* Establish initial language state 0.
* Language 0 is active at startup and cannot be named by #lang.
* Canonical #lang set is: diff, dift, alg, as, avm, ACS.
* Historical vasm was replaced by as for the real MCPU assembler.
* MCPU_CPP_VASM_INCLUDE_PATH became MCPU_CPP_AS_INCLUDE_PATH.
* Removed the temporary MCPU_CPP_C_INCLUDE_PATH.  The root
  MCPU_CPP_INCLUDE_PATH is the main include directory of the unnamed base
  base MCPU language (state 0) and is selected only in that state.
* Added the public `make tests` target.

mcpu-cpp 0.0.1
==============

Initial architectural release.

* New UTF-8 external / UCS-2 internal source model with strict NUL rejection.
* No legacy code-page machinery.
* Hand-written preprocessing scanner; flex is not used.
* Scanner honors quoted strings and the diff apostrophe rule.
* Source/include stack with #include support.
* #lang/#endlang language stack defined for the MCPU language stack.
* Configuration-file and command-line include search paths.
* Initial language set: c, diff, dift, alg, vasm, avm, ACS.