Release 0.2.1

Added

• Memory inspector: command PIC18 Assembly: Open Memory Inspector opens a read-only
  memory view as a tab in the editor area. Memory is fetched from the active debug session
  with the standard DAP readMemory request, so it works with the MPLAB debug adapter
  (mplab-core-da) without depending on how that adapter is implemented.

  • Memory space selector (ram / program , configurable) that builds
    a <space>:<address> memory reference, plus a raw mode for typing a reference directly.
    Switching space adopts that space's MAU size, grouping and length, so program lands on
    2-byte (instruction word) groups over 4096 bytes.
  • Configurable start address (hex) and length, row width (Auto or a fixed byte count),
    MAU size, MAU per group, byte order, and display format (hex / dec / oct / bin).
  • MPLAB X style layout: rows aligned to the row width with a hex address column (0x0000,
    0x0010, …) and column headers giving the byte offset within the row.
  • Optional ASCII column, and optional highlighting of values that changed since the
    previous read.
  • Re-reads automatically when the target stops; format changes re-render without touching
    the target again.

• SFR view: command PIC18 Assembly: Open SFR View opens a read-only table of the
  device's Special Function Registers (address, name, value) as a tab in the editor area.

  • The first read evaluates each register name once and keeps the memoryReference the
    adapter answers with (e.g. sfr:0xFD8), so the addresses come from the adapter rather
    than from this extension. Every refresh afterwards is a single block readMemory, with
    each register's value rebuilt from the response's own base address — fast, and portable
    to other devices.
  • Registers with named bits (e.g. STATUS → N, OV, Z, DC, C) expand into sub-rows; a bit
    row shows its bit number in the address column and bit rows always keep bit order, no
    matter how the parent registers are sorted.
  • Click the Address or Register column header to sort by it; click again to reverse.
  • Every value is shown in hex, decimal and binary at once, in their own columns.
  • Search box matching a register name, a bit name, or an address (F80 or 0xF80); an
    "Only changed" filter that keeps just the registers whose value moved since the previous
    read; highlighting of changed values.

• SFR table generated by scripts/generateSfrMap.js. Bit positions are derived from the
  BANKMASK(<register>), <bit> form of the xc8 defines and cross-checked against the
  datasheet bit layout already shipped in register.json; the generator reports any
  disagreement rather than silently picking one source.

• Setting pic18-assembly.memoryInspector.memorySpaces for the memory space list,
  pic18-assembly.memoryInspector.spaceDefaults for the per-space presets, and
  pic18-assembly.memoryInspector.debugTypes for the debug session types to read from.

• Setting pic18-assembly.projectScope (folder / workspace) to choose whether each
  workspace folder is its own project or all folders share a single index.

• Output channel "PIC18 Assembly" (command PIC18 Assembly: Show Log) logging every Debug
  Adapter Protocol exchange the memory and SFR views make — the request arguments, the
  resolved address, how many bytes were readable, and the adapter's error text — so a failing
  read can be diagnosed and exported.

Fixed

• The memory view mis-drew the unreadable region: per the DAP specification data holds the
  bytes that were read and unreadableBytes counts further bytes after them, but the view
  treated the unreadable run as overlapping the returned data. The grid is now built from the
  address the adapter reports for its data and stops at the last byte it returned;
  unreadable addresses are left out entirely and the status line says where they start.
• Switching the memory view to another memory space no longer marks every cell as changed.
  Addresses only mean the same thing within one space, so a read is now compared against the
  previous one only when both came from the same space.
• The memory view's column headers no longer widen with the display format (they showed three
  digits in decimal mode). They are sized from the largest byte offset in a row instead, which
  never needs more than two hex digits.
• Decimal values in the memory view are no longer zero padded, matching the SFR view.
• Both views now scroll inside themselves instead of scrolling the whole tab, and their
  column header row stays pinned to the top while the rows scroll under it.
• The memory view treated the MAU size as a grouping width rather than as the size of an
  addressable unit. Setting the MAU to 2 bytes now means one address covers two bytes, so
  eight bytes of data span the addresses 0x00–0x03 instead of 0x00–0x07. The row length is
  counted in addressable units accordingly (mausPerRow in
  pic18-assembly.memoryInspector.spaceDefaults, renamed from bytesPerRow), which keeps
  the address column stepping by 0x10 per row at any MAU size.
• A group in the memory view is now one column holding one uninterrupted value, the way a
  multi-byte addressable unit already was. On a little-endian target its highest address is
  the most significant part, so F6 at 0x00 and 6A at 0x01 read as 6AF6 with two units
  per group; big-endian keeps plain address order. Each column header carries the lowest
  address of its group (00 02 04 … at two units per group). A group is only drawn when
  every unit of it was read, and the ASCII column stays in plain address order, being a
  rendering of the raw bytes.
• The SFR view no longer assumes one memory reference syntax. It reads the block from the
  space origin the adapter itself reported, falling back to the lowest register address and
  then a bare address, remembers whichever form was accepted, and lists every form it tried
  when they all fail.
• Multi-root workspaces are no longer indexed as a single project. Each workspace folder now
  has its own symbol index, so identically named labels, variables or macros in unrelated
  folders no longer report as Duplicate label / Duplicate macro name, and completions,
  go-to-definition, find-references, hovers and semantic highlighting stay inside the folder
  the file belongs to.
• Include paths that fall back to a basename match no longer resolve to a file in a different
  workspace folder.
• A folder's .gitignore no longer excludes files of a sibling workspace folder from indexing.
• Adding or removing a workspace folder, or changing pic18-assembly.exclude, now reindexes
  the workspace instead of leaving stale entries behind.

Release 0.2.0

Added

• Assembler dialect setting pic18-assembly.assembler (mpasm / xc8): directive completions,
  SFR definition values and hover documentation now follow the selected assembler.
• Compatibility diagnostics: warn when a directive of the other assembler dialect is used
  (can be disabled with pic18-assembly.diagnostics.assemblerCompatibility).
• Directive completions with descriptions at the opcode position, and hover documentation for directives.
• Preprocessor directive completions after typing # (dialect aware).
• Hover documentation for SFR / standard include definitions (including register bit layouts).
• Support for EXTERN (MPASM spelling), SET, CBLOCK/ENDC members, name RES n / name DS n
  and value-less #define when indexing symbols.
• Comma-separated name lists in EXTRN / GLOBAL are now fully indexed (previously only the first name).
• LOCAL symbols inside macros are recognized and scoped to the macro.
• Labels defined in included files are now valid branch targets (textual inclusion).
• Setting pic18-assembly.exclude for additional indexing exclusions, and
  pic18-assembly.documentation.includeTraditionalChinese to toggle the zh-TW instruction docs.
• When accepting a not-yet-imported label completion with the xc8 dialect selected, a
  GLOBAL declaration is also added to the file that defines the label (pic-as requires the
  defining module to export the symbol), in addition to the EXTRN inserted locally.
• xc8-only diagnostic: warn when an EXTRN symbol is defined in another file without a
  matching GLOBAL export, with a Quick Fix that inserts the missing GLOBAL declaration.
• "Find All References" (Shift+F12) for labels, variables (EQU/#define/CBLOCK), macros,
  macro parameters and EXTRN-only C symbols — works from the definition site and from any
  usage, including symbols passed as macro invocation arguments.
• Ctrl+Click on a definition site (label, variable, macro) now shows the reference list,
  matching the behavior of other language extensions; Ctrl+Click on a C-symbol usage jumps
  to its EXTRN declaration.
• Numeric literal diagnostics: digit-leading tokens in instruction operands that match no
  supported number format (binary 0101B, octal 17Q, decimal 10, hex 22h/0FFh/0x1F)
  are reported as errors — also inside the expansion of #define aliases and built-in
  defines, and in EQU/SET values. The TextMate grammar's number rules were aligned
  (lowercase binary suffix b, hex suffix without a leading zero like 22h).
• ISA operand-count diagnostics: report too few / too many operands per the PIC18 instruction
  signatures (MOVFF fs, fd, BSF f, b {,a}, ...). Aliases are expanded before counting:
  user #define substitutions that stand for several operands (e.g. #define LED LATA, 0, a)
  and dialect-dependent built-ins (pic-as LATA0 = BANKMASK(LATA), 0, a) are handled, and
  instructions whose operands are macro parameters are skipped (caller-dependent).
• Underscore-prefixed EXTRN names (C symbols such as _main from linked C object files) are
  treated as valid references: no error, an informational (blue underline) marker instead,
  with label coloring, an explanatory hover, and auto-completion that behaves like regular
  labels (including automatic EXTRN insertion).

Fixed

• Language features now update live while typing; previously most data refreshed only on save.
• Removed catastrophic-backtracking-prone regular expressions; parsing is now a linear scan
  (significantly better performance and stability on large files).
• Label completion cache no longer serves stale results after edits.
• Hover / Go to Definition / Find References no longer misidentify parts of numeric or
  character literals as symbols (e.g. the binary suffix in 01010101B being confused with
  the assembler designator b, hex 0FFh, radix literals B'0101', char literals 'A').
• .gitignore entries are now converted to proper glob patterns instead of being used verbatim.
• Fixed macro argument count validation messages and unclosed-macro detection.

Changed

• Complete architectural rewrite: single-pass parser + workspace index; all providers
  (completion, hover, definition, semantic tokens, diagnostics) read from one shared model.
• Macro definition snippet now includes the macro name placeholder (name MACRO ... ENDM).

Removed

• Runtime dependency on lru-cache.

Release 0.1.10

Added

• AI-translated instruction details in Traditional Chinese.

Fixed

• N/A

Changed

• N/A

Removed

• N/A

Release 0.1.9

Added

• N/A

Fixed

• Can use labels as variable names.

Changed

• N/A

Removed

• N/A

Release 0.1.8

Added

• Support use macro parameters as branch labels.

Fixed

• N/A

Changed

• N/A

Removed

• N/A

Release 0.1.7

Added

• Auto exclude files in .gitignore from language features.

Fixed

• Hightlighting issue when use EQU with lowercase label.
• Hightlighting issue when use macro with prefix space before macro name.

Changed

• Add .history to the excluded files in the extension.

Removed

• N/A

Release 0.1.6

Added

• Hover documentation for instruction.
• Hover documentation for label.
• Hover documentation for variable.
• Hover documentation for macro.

Fixed

• Fix variable still exists after removing all variable definitions in a file.

Changed

• N/A

Removed

• N/A

Release 0.1.5

Added

• N/A

Fixed

• Fix variable still exists after removing all variable definitions in a file.

Changed

• N/A

Removed

• N/A