Changelog

0.3.1#

Released with mutant 2.6.1. Nothing the extension does has changed.

• The manifest names mutant 2.6.1.
• Development dependencies: brace-expansion 1.1.21 and 5.0.12.

0.3.0#

Released with mutant 2.6.0, which it requires for the task changes below.

• Tasks keep working with mutant 2.6.0, which refuses a bare --password. A task's password field is now passed as --password-insecure, the explicit opt-in to a password on the command line, and a new passwordFile field passes --password-file so the password stays off it. Both now apply to gen and run tasks as well as release. Leave both unset to be prompted in the terminal.

• Names resolve across modules the way the compiler resolves them. The language server now asks the same symbol graph the compiler builds, so stats.mean has go-to-definition, hover, completion, find-references and rename in every file that imports it, under whatever alias each file chose, and call hierarchy finds callers in files nobody has opened. Answers are scoped to the import closure, so go-to-definition no longer jumps into a module nothing imported.

• A new diagnostic: unusedImport, at information severity rather than warning, because an imported module's top-level statements run whether or not its namespace is read. mutant.lint.rules.unusedImport.severity sets it.

• A new diagnostic: classifiedPlaintext. Plaintext read out of a classified record with record_read refuses to go to putln, fs_write, http_post and ten other builtins when the program runs; the editor now says so at the line, before the program runs and asks for a passphrase. record_release(buffer, reason) is the way to comply, and its result is never reported. The rule reads the runtime's own lists, so it cannot warn about a builtin that does not refuse. mutant.lint.rules.classifiedPlaintext.severity sets it.

• A new diagnostic: builtinArgChoice. A string literal that is not one of the words a builtin parameter takes -- db_bfs(h, n, 2, "outbound"), whose direction is out, in or both -- is refused when the program runs; the editor now says so at the argument. The words are the builtin's own list and are compared the way the builtin compares them. mutant.lint.rules.builtinArgChoice.severity sets it.

• classifiedPlaintext follows a search. A hit of record_search, or the context it carries, handed to a builtin that refuses it is flagged at the line. A search's whole result and its hits are not, because a search that found nothing holds no plaintext.

• Four new diagnostics for the case builtins. roleLiteral, filteredLedgerHandle, secretOption and lifecycleState report a call the run time refuses -- a role case_open or ledger_open will not take, the handle ledger_under_view returns passed to a builtin that does not read under a view, an option named passphrase, password, secret or key, and a state case_transition never moves a case to -- at the line and in the run time's own words. Each is a warning by default, set by mutant.lint.rules.<rule>.severity.

• A call inside a return is reported once. Seven rules, most of the security family among them, reported it twice.

• Less garbage on every keystroke. Seven rules built the set of every builtin's name again, about 27 KB each, whenever a document was analysed. They read the one the server builds when it starts.

• Every new builtin is highlighted and hoverable -- 196 of them, across disk image verification, partition-offset opening, deleted-file recovery, journals, slack, the forensic ledger, classified records, disclosure, reading a graph store back, and a case's roles, lifecycle, evidence custody, reviews, retention and erasure -- with no change here, because the grammar and the hover cards are both generated from the registry.

• A new diagnostic: macroSafety. A macro body runs while the compiler runs, where a builtin that reads or writes a file, reaches the network, starts a process, draws entropy, reads the clock or decodes cryptographic material is refused. The editor now says so at the line, in the expander's own words, instead of leaving a clean file and a build that fails with nothing to click. Nothing inside a quote(...) is reported, since that is the source the macro emits; an unquote(...) inside one is. A warning by default, set by mutant.lint.rules.macroSafety.severity.

0.2.0#

New this release:

• Debugging — a mutant debug type. Press F5 on a .mut file with no launch.json and the extension debugs the file in front of you; a saved configuration takes program, stopOnEntry, modulePaths and noDebug. The adapter is the CLI itself (mutant debug), found through the existing mutant.cli.path setting, so the binary that runs a program is the binary that debugs it.
• Every builtin is highlighted. The grammar's list of builtin names was maintained by hand and had fallen to 76 of the language's 497 — the reporting, chain-of-custody and schema-interchange families were all invisible, and so were most of net_* and str_*. The list is now generated from the builtin registry by cmd/gendocs, with a test in the repo that fails if the two ever disagree, so a builtin is highlighted as soon as it exists.
• The bundled language server says what it is. mlsp --version prints the release it was built from and the number of builtins it knows (mlsp 2.5.0 (497 builtins)), so "which version of the language does my editor actually understand?" is a question with an answer. The packaging scripts ask it too: a VSIX can no longer be built around a server older than the language in the tree it was cut from.
• Sigma rules are highlighted and hoverable. The language gained four sigma_* builtins that parse and evaluate Sigma detection rules, and the generated grammar and the server's hover cards picked them up with no change here — which is the point of generating both from the registry.
• A new diagnostic: assignmentTarget. The compiler used to accept counts[host]["n"] = 1 and quietly lose the write; it now emits the write-back chain that makes it work, and refuses by name the two targets it still cannot emit — one with no variable under it, and an index before the last one that would be evaluated more than once. The editor reports both where they are written rather than at the next build, and a test runs the analyzer and the compiler over the same programs, so the two answers cannot drift apart.
• The settings documentation caught up with the settings. This README listed 16 of the 24 lint rules; the eight it omitted were the whole security family — weak crypto, hardcoded secrets, command injection, path traversal, TLS verification, unbounded resources and evidence mutation. Those are the rules most worth knowing about and the least likely to be found by scrolling a settings pane. A test now fails when a rule exists that the README does not mention.

0.1.0#

New this release:

• Snippets — fn, let, letr (multi-value bind), guard, if/ife, for, struct/structlit, enum, macro, and putln/putf.
• Richer editor configuration — folding via // region / // endregion markers, indentation rules, comment-continuation on Enter, and a Mutant-aware word pattern.
• Grammar — macro declarations (let NAME = macro(...)) and capitalized type-usage sites (Point { … }, Color.Green) are now scoped.
• Mutant tasks — a task provider wrapping the real CLI: gen (compile a .mut to .mu), run a compiled .mu, and release (standalone build). The CLI is resolved from the new mutant.cli.path setting (default mutant).
• Language server upgrades (bundled mlsp): member/dot completion (x.field, Enum.Variant), folding ranges, parameter-name inlay hints, semantic-token range + delta requests, richer code-action/rename capabilities, and workspace-wide indexing so go-to-definition, references, and rename work across files you have not opened.
• Type-aware editor smarts — the language server now infers best-effort types (from literals, let bindings, struct/enum, and a curated builtin table) and surfaces them in hover (count : int), completion detail, and inlay type hints on let bindings. Purely informational — no annotations, no type errors, no runtime effect; anything uncertain simply shows nothing.
• Reference codeLens — a clickable "N references" lens above every top-level declaration (opens the references peek).
• Document links — string literals that name an existing file (e.g. fs_read("evidence/$MFT")) become ctrl-clickable links that open the file.
• Status bar indicator — shows Mutant LSP state (running / starting / failed / stopped); click to restart.
• Menus + keybindings — "Mutant: Format Document" in the editor context menu for .mut files, and Ctrl+Alt+R (Cmd+Alt+R on macOS) to restart the LSP.
• Packaging — the extension now bundles with esbuild and publishes per-platform .vsix packages (each carrying only its own mlsp binary), replacing the single ~250 MB all-platforms package.

0.0.6#

New this release:

• Compound assignment and increment/decrement (+= -= *= /= %=, postfix ++ / --) are now lexed, parsed, formatted, and syntax-highlighted. The formatter preserves the compact spelling (i += 1, i++) instead of expanding it, and the bundled language server understands the operators everywhere.

0.0.5#

New this release:

• OS-aware platform-support diagnostic — warns when a program calls a builtin that is not supported on the operating system the language server is running on (for example a Windows/Linux-only builtin used on macOS). Configurable via mutant.lint.rules.platformSupport.severity.
• Unreachable-code diagnostic — flags statements after an unconditional return / break / continue. Configurable via mutant.lint.rules.unreachableCode.severity.
• Hover and completion now show each builtin's capability category (e.g. filesystem, network, graph database) and any platform constraint.
• Semantic highlighting recognizes the &&, ||, <=, and >= operators.
• The discard identifier _ is no longer flagged as a duplicate declaration.
• Bundled mlsp language-server binaries rebuilt for all six targets (Windows / Linux / macOS × amd64 / arm64).

Existing capabilities: canonical strict-semicolon formatting (format-on-save by default), diagnostics (unused / undefined / duplicate declarations, deep nesting, and missing/redundant semicolons with quick fixes), hover, completion, semantic tokens, go-to-definition, references, rename, and document/workspace symbols.