ed/pikuma_ps1
1
0
Fork 0
mirror of https://github.com/Ed94/pikuma_ps1.git synced 2026-08-25 02:20:33 +00:00
Code Issues Packages Projects Releases Wiki Activity

Review Pass: Type annotations.

Browse Source
This commit is contained in:
ed
2026-08-19 23:18:32 -04:00
parent c226e8a7d3
commit 449216967b
19 changed files with 5602 additions and 429 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/duffle.lua
+53
View File
@@ -1,10 +1,52 @@
--- duffle.lua — facade over duffle_scan / duffle_isa / duffle_emit.
--- @class DuffleExport
--- bag: open module-export keys from duffle_scan / duffle_isa / duffle_emit
--- @type DuffleExport
local scan = require("duffle_scan")
--- @type DuffleExport
local isa = require("duffle_isa")
--- @type DuffleExport
local emit = require("duffle_emit")
--- @type DuffleExport
local M = {}
--- @alias Path string
--- @alias LineNum integer
--- @alias ByteOff integer
--- @alias MacroName string
--- @alias AtomName string
--- @alias Severity string
--- @class SourceFile
--- @field path Path
--- @field text string
--- @field dir string
--- @field basename string
--- @field scan SourceScan|nil
--- @class CorpusView
--- @field register_alias_registry table<string, AliasEntry>
--- @field type_name_registry table<string, TypeNameEntry>
--- @field atom_views table<AtomName, AtomViewEntry>
--- @field atom_ctxs table<AtomName, AtomCtxEntry>
--- @field atom_phases table<string, AtomPhaseGroup>
--- @field binds_by_name table<string, BindsEntry>
--- @field atoms_by_name table<AtomName, AtomEntry>
--- @field atom_infos AtomInfoEntry[]
--- @field components table<string, ComponentDef>
--- @field component_atom_infos AtomInfoEntry[]|nil
--- @field component_body_index table<string, ComponentBodyEntry>
--- @field tape_chains table<string, TapeChain>|nil
--- @field source_order SourceFile[]
--- @field collisions CorpusCollision[]
--- @param src DuffleExport
--- @param label string
--- @return nil
local function merge(src, label)
--- @type string, any
for k, v in pairs(src) do
if M[k] ~= nil and M[k] ~= v then
error("duffle facade name collision on " .. tostring(k) .. " from " .. label, 0)
@@ -17,7 +59,10 @@ merge(scan, "duffle_scan")
merge(isa, "duffle_isa")
merge(emit, "duffle_emit")
--- @param ctx PassCtx
--- @return CorpusView
function M.corpus_view(ctx)
--- @type Corpus
local corpus = ctx and ctx.shared and ctx.shared.corpus
if not corpus then error("requires ctx.shared.corpus", 0) end
return {
@@ -38,8 +83,16 @@ function M.corpus_view(ctx)
}
end
--- @param rules CheckRule[]
--- @param phase string
--- @param item AtomEntry|SourceFile
--- @param pipe_ctx PipeCtx
--- @param findings CheckFinding[]
--- @return nil
function M.run_check_rules(rules, phase, item, pipe_ctx, findings)
--- @type integer, CheckRule
for _, rule in ipairs(rules) do
--- @type (fun(item: AtomEntry|SourceFile, pipe_ctx: PipeCtx, findings: CheckFinding[]): nil)|nil
local fn = rule[phase]
if fn then fn(item, pipe_ctx, findings) end
end
scripts/duffle_emit.lua
+250 -14
View File
@@ -1,8 +1,13 @@
--- duffle_emit.lua — project_emission + decl finders.
--- @type DuffleScan
local scan = require("duffle_scan")
--- @type DuffleIsa
local isa = require("duffle_isa")
--- @type DuffleEmit
local M = {}
--- @type string, any
for k, v in pairs(scan) do M[k] = v end
--- @type string, any
for k, v in pairs(isa) do M[k] = v end
-- Section 8: Cross-source component-body index + word-event expansion
@@ -12,25 +17,55 @@ for k, v in pairs(isa) do M[k] = v end
-- built once from the pre-tokenized bodies.
--- @class ComponentBodyEntry
--- @field body_tokens table -- pre-tokenized {{tok=string, rel=integer}, ...}
--- @field body_tokens BodyToken[]
--- @field body_off integer -- byte offset of body[1] in `source`
--- @field line_of fun(pos:integer):integer -- byte-offset → 1-based line number in `source`
--- @field source string -- absolute path of the source containing the declaration
--- @field declaration integer -- 1-based line number of the MipsAtomComp_(ac_X) declaration
--- @field kind string -- "comp_bare" | "comp_proc"
--- @field arg_names string[]|nil
--- @field sub_map table<string, string>|nil -- bag: formal -> substituted operand
--- @class EmissionWalkCtx
--- @field component_index table<string, ComponentBodyEntry>
--- @field word_counts WordCounts
--- @field components table<string, ComponentDef>
--- @field reg_use_schema RegUseSchema|nil
--- @field reg_use_param string|nil
--- @field atom_name AtomName|nil
--- @field schema_name string|nil
--- @field visiting table<string, boolean>|nil -- bag: component name on DFS stack
--- @field root_call_path string|nil
--- @field root_call_line integer|nil
--- @class RegUseCtx
--- @field reg_use_schema RegUseSchema|nil
--- @field reg_use_param string|nil
--- @field atom_name AtomName|nil
--- @field schema_name string|nil
--- @class DuffleEmit
-- The cross-source component-body index is owned by the corpus (`corpus.component_body_index`, populated by `passes/components.lua`).
-- Consumers (`passes/static_analysis.lua`, `passes/emission_model.lua`) read it directly; per-pass memoization helpers stay out of scope.
-- ASCII byte constants used by split_call_args (kept local to keep Section 8 self-contained).
--- @type integer
local E_BYTE_OPEN_PAREN = 0x28
--- @type integer
local E_BYTE_OPEN_BRACE = 0x7B
--- @type integer
local E_BYTE_OPEN_BRACK = 0x5B
--- @type integer
local E_BYTE_DQUOTE = 0x22
--- @type integer
local E_BYTE_SQUOTE = 0x27
--- @type integer
local E_BYTE_COMMA = 0x2C
-- Map an open-delimiter byte to its matching close string for read_balanced.
--- @type table<integer, string> -- bag: open-delimiter byte -> close string
local E_OPEN_CLOSE = {
[E_BYTE_OPEN_PAREN] = ")",
[E_BYTE_OPEN_BRACE] = "}",
@@ -44,15 +79,22 @@ local E_OPEN_CLOSE = {
--- @param inner string
--- @return string[]
local function split_call_args(inner)
--- @type string[]
local args = {}
if not inner or inner == "" then return args end
--- @type integer
local pos = 1
--- @type integer
local len = #inner
--- @type integer
local start = 1
while pos <= len do
--- @type integer
local c = inner:byte(pos)
--- @type string|nil
local close = E_OPEN_CLOSE[c]
if close then
--- @type integer, integer
local _, after = M.read_balanced(inner, string.char(c), close, pos)
pos = after
elseif c == E_BYTE_DQUOTE or c == E_BYTE_SQUOTE then
@@ -74,17 +116,22 @@ end
--- @param tok string
--- @return string, string[]
local function token_ident_and_args(tok)
--- @type string|nil, integer
local ident, after = M.read_ident(tok, 1)
if not ident then return "?", {} end
--- @type integer
local paren_pos = M.skip_ws_and_cmt(tok, after)
if tok:sub(paren_pos, paren_pos) ~= "(" then return ident, {} end
--- @type string|nil
local inner = M.read_parens(tok, paren_pos)
if not inner then return ident, {} end
return ident, split_call_args(inner)
end
-- The macro-name prefix that marks a `mac_X(...)` component invocation.
--- @type string
local E_MAC_PREFIX = "mac_"
--- @type integer
local E_MAC_PREFIX_LEN = 4
--- Expand a body entry into the flat sequence of emitted machine-word events.
@@ -104,10 +151,10 @@ local E_MAC_PREFIX_LEN = 4
---
--- Pure: reads `body_entry` / `component_index` / `word_counts`. Memoization is the caller's responsibility.
--- Callers wanting `word_events` / `word_event_errors` precomputed for many atoms should memoize them per atom.
--- @param body_entry table -- `{body_tokens, body_off, line_of, source, declaration}` (declaration = root atom's atom.line)
--- @param component_index table -- the bare-name → ComponentBodyEntry map from M.get_component_body_index
--- @param word_counts table -- macro name → emitted-word count (from `ctx.shared.word_counts`)
--- @return WordEvent[], WordEventError[]
--- @param body_entry ComponentBodyEntry
--- @param component_index table<string, ComponentBodyEntry>
--- @param word_counts WordCounts
--- @return WordEvent[], EmitError[]
-- ════════════════════════════════════════════════════════════════════════════
-- Section 11: project_emission (per-atom emission projection)
@@ -122,12 +169,12 @@ local E_MAC_PREFIX_LEN = 4
-- word_counts table is authored-metadata + current-component count table.
--- @class EmissionProjection
--- @field items table[] -- Ordered stream of word|label|offset|invoke_begin|invoke_end
--- @field word_events table[] -- Dense view of items where kind == "word"
--- @field markers table[] -- Dense view of items where kind == "label"|"offset"
--- @field items EmissionItem[]
--- @field word_events WordEvent[]
--- @field markers EmissionMarker[]
--- @field invocations InvocationRecord[] -- dense view of items where kind == "invoke_begin"|"invoke_end"
--- @field errors table[] -- Token-resolution failures surfaced without fail-loud
--- @field warnings table[] -- Opaque warnings (e.g. unknown uncounted macro)
--- @field errors EmitError[]
--- @field warnings EmitWarning[]
--- @class InvocationRecord
--- Lives at `atom.paths.invocations[*]`. Constructed once at the single invocation-construction site
@@ -148,7 +195,7 @@ local E_MAC_PREFIX_LEN = 4
--- @field end_word integer -- 1-based items index of the `invoke_end` item (set by `emit_invoke_end`)
--- @field word_count integer -- Number of `word` items emitted between `start_word` and `end_word` (inclusive)
--- @field debug_skip boolean -- `debug_skip` stamp; true iff `corpus.components[name].debug_skip` is true at construction. Always boolean (never `nil`).
--- @field errors table[] -- Per-invocation construction errors (cycle / count_mismatch); does not include pass-level errors
--- @field errors EmitError[]
-- Internal recursive walker. The items stream holds every emitted event in order; `word_events`, `markers`,
-- `invocations`, `errors`, `warnings` are dense views / side outputs appended alongside.
@@ -166,35 +213,58 @@ local E_MAC_PREFIX_LEN = 4
-- then breaks out without recursing (the cycle entry still receives an invocation ID + paired `invoke_begin` / `invoke_end` items, so the boundary invariant is preserved).
-- * Component declared-count mismatch (declared vs. measured) is a construction error (kind = "count_mismatch"); recorded on the invocation record and pass-level errors list.
-- * Final boundary check: if any invocation is still open at end of walk, surface a "unbalanced" construction error.
--- @param root_body_entry ComponentBodyEntry
--- @param ctx_table EmissionWalkCtx
--- @return EmissionProjection
local function _project_emission_inner(root_body_entry, ctx_table)
--- @type EmissionItem[]
local items = {}
--- @type WordEvent[]
local word_events = {}
--- @type EmissionMarker[]
local markers = {}
--- @type InvocationRecord[]
local invocations = {}
--- @type EmitError[]
local errors = {}
--- @type EmitWarning[]
local warnings = {}
--- @type integer
local word_idx = 0
--- @type InvocationRecord[]
local invocation_stack = {} -- stack of currently-open invocation records
--- @type integer
local next_inv_id = 0
--- @type RegUseSchema|nil
local reg_use_schema = ctx_table.reg_use_schema
--- @type string|nil
local reg_use_param = ctx_table.reg_use_param
--- @type AtomName|nil
local atom_name = ctx_table.atom_name
--- @type table<string, boolean> -- bag: slot name -> readonly
local slot_readonly = {}
if reg_use_schema then
--- @type integer, RegUseSlot
for _, slot in ipairs(reg_use_schema.slots or {}) do
slot_readonly[slot.name] = slot.readonly == true
end
end
--- @param sub_map table<string, string>|nil
--- @param operand string|any
--- @return any
local function apply_sub(sub_map, operand)
if not (sub_map and type(operand) == "string") then return operand end
if sub_map[operand] then return sub_map[operand] end
--- @type integer|nil
local dot = operand:find(".", 1, true)
if dot then
--- @type string
local head = operand:sub(1, dot - 1)
--- @type string|nil
local mapped = sub_map[head]
if type(mapped) == "string" then
return mapped .. operand:sub(dot)
@@ -203,39 +273,65 @@ local function _project_emission_inner(root_body_entry, ctx_table)
return operand
end
--- @param operand string|any
--- @return string|nil, string|nil, string|nil
local function resolve_gpr_key(operand)
if type(operand) ~= "string" then return nil end
if operand:sub(1, 2) == "R_" then return operand end
if not (reg_use_schema and reg_use_param) then return nil end
--- @type string
local prefix = reg_use_param .. "."
if operand:sub(1, #prefix) ~= prefix then return nil end
--- @type string
local member_path = operand:sub(#prefix + 1)
--- @type string|nil
local slot = reg_use_schema.alias_to_slot[member_path]
if not slot then return nil, member_path end
return "reguse:" .. atom_name .. ":" .. slot, nil, slot
end
--- @return integer[]
local function open_invocation_ids_snapshot()
--- @type integer[]
local ids = {}
--- @type integer, InvocationRecord
for _, inv in ipairs(invocation_stack) do
ids[#ids + 1] = inv.id
end
return ids
end
--- @param encoder string
--- @param args string[]|nil
--- @param line integer
--- @param word_call_text string|nil
--- @param def_source_now string|nil
--- @param def_line_now integer|nil
--- @param immediate_call_text string|nil
--- @param root_call_text_w string|nil
--- @param sub_map table<string, string>|nil
--- @return nil
local function emit_word(encoder, args, line, word_call_text, def_source_now, def_line_now, immediate_call_text, root_call_text_w, sub_map)
--- @type integer[]
local inv_ids = open_invocation_ids_snapshot()
--- @type integer
local outermost = inv_ids[1] or 0
-- For words emitted at the root atom body, `immediate_call_text` is nil and the walker's `word_call_text` (the word's own token, e.g. "nop") becomes the effective call_text.
-- For words emitted inside a component expansion, `immediate_call_text` is the immediate outer `mac_X(...)` token text;
-- The call that triggered the body expansion we're currently walking.
--- @type string|nil
local eff_call_text = immediate_call_text or word_call_text
--- @type string|nil
local eff_root_call_text = root_call_text_w
--- @type string[]|nil
local gpr_keys = nil
if reg_use_schema or sub_map then
gpr_keys = {}
--- @type integer, string
for pos, arg in ipairs(args or {}) do
--- @type any
local effective = apply_sub(sub_map, arg)
--- @type string|nil, string|nil, string|nil
local key, unresolved, slot = resolve_gpr_key(effective)
gpr_keys[pos] = key
if unresolved then
@@ -247,8 +343,10 @@ local function _project_emission_inner(root_body_entry, ctx_table)
}
end
if key and slot and slot_readonly[slot] then
--- @type InstructionRow|nil
local row = M.instr(encoder)
if row and row.writes then
--- @type integer, integer
for _, wpos in ipairs(row.writes) do
if wpos == pos then
errors[#errors + 1] = {
@@ -266,16 +364,25 @@ local function _project_emission_inner(root_body_entry, ctx_table)
if not reg_use_schema then
gpr_keys = nil
end
--- @type InstructionRow|nil
local isa = M.instr(encoder)
--- @type string
local isa_kind = isa and isa.kind or "unknown"
--- @type integer
local nop_words = (encoder == "nop" and 1) or (encoder == "nop2" and 2) or 0
--- @type boolean
local is_yield = (encoder == "mac_yield" or encoder == "mac_yield_tail")
--- @type string|nil
local gp0_shape = type(encoder) == "string"
and encoder:match("^mac_format_([%w_]+)_color$")
or nil
--- @type boolean
local is_load = (isa_kind == "load")
--- @type boolean
local is_branch = (isa_kind == "branch")
--- @type boolean
local is_unconditional_jump = (encoder == "jump" or encoder == "call_addr")
--- @type boolean
local is_terminal_jump = (encoder == "jump_reg" or encoder == "call_reg" or encoder == "jump_link")
items[#items + 1] = {
kind = "word",
@@ -324,10 +431,21 @@ local function _project_emission_inner(root_body_entry, ctx_table)
word_idx = word_idx + 1
end
--- @param kind string
--- @param name string
--- @param target string|nil
--- @param line integer
--- @param immediate_call_text string|nil
--- @param root_call_text_w string|nil
--- @param consuming_encoder string|nil
--- @param consuming_arg_pos integer|nil
--- @return nil
local function emit_marker(kind, name, target, line,
immediate_call_text, root_call_text_w,
consuming_encoder, consuming_arg_pos)
--- @type integer[]
local inv_ids = open_invocation_ids_snapshot()
--- @type integer
local outermost = inv_ids[1] or 0
-- Markers carry the open invocation stack snapshot. `call_text` / `root_call_text` belong to words, not markers — markers are zero-width and skip per-word call-site attribution.
-- `consuming_encoder` + `consuming_arg_pos` carry the surrounding control-transfer instruction context
@@ -338,6 +456,7 @@ local function _project_emission_inner(root_body_entry, ctx_table)
if kind == "offset" and (consuming_encoder == nil or consuming_encoder == "") then
return
end
--- @type EmissionItem
local it = {
kind = kind,
name = name,
@@ -364,21 +483,32 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- Count top-level commas in `tok` between position `from_pos` (inclusive) and `to_pos` (exclusive).
-- Tracks paren depth so commas inside nested () don't count. Skips string literals + comments.
-- Used by `emit_embedded_markers` to compute `consuming_arg_pos` for each embedded marker.
--- @param tok string
--- @param from_pos integer
--- @param to_pos integer
--- @return integer
local function count_top_level_commas(tok, from_pos, to_pos)
--- @type integer
local depth = 0
--- @type integer
local count = 0
--- @type integer
local i = from_pos
while i < to_pos do
--- @type string
local c = tok:sub(i, i)
if c == "'" or c == '"' then
--- @type integer
local next_pos = M.skip_str_or_cmt(tok, i)
i = (next_pos > i) and next_pos or (i + 1)
elseif c == "/" and tok:sub(i + 1, i + 1) == "/" then
-- line comment: skip to end of line
--- @type integer|nil
local nl = tok:find("\n", i, true)
i = (nl and nl + 1) or (#tok + 1)
elseif c == "/" and tok:sub(i + 1, i + 1) == "*" then
-- block comment: skip to matching */
--- @type integer|nil
local close = tok:find("*/", i + 2, true)
i = (close and close + 2) or (#tok + 1)
elseif c == "(" then
@@ -399,9 +529,13 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- Find the position of the consuming instruction's open paren (the `(` that starts the consuming instruction's argument list).
-- Returns nil if the token's leading text isn't an ident followed by `(` (e.g. the ident is at the start of a non-instruction token).
--- @param tok string
--- @return integer|nil
local function find_consuming_paren(tok)
--- @type integer
local i = 1
while i <= #tok do
--- @type string
local c = tok:sub(i, i)
if c == "(" then return i end
if not c:match("[%w_]") and c ~= " " then return nil end
@@ -410,24 +544,33 @@ local function _project_emission_inner(root_body_entry, ctx_table)
return nil
end
--- @param tok string
--- @param tok_line integer
--- @param consuming_encoder string|nil
--- @return nil
local function emit_embedded_markers(tok, tok_line, consuming_encoder)
-- When called with a non-nil `consuming_encoder`, the marker is nested inside that instruction's argument list.
-- We compute each marker's arg position by counting top-level commas between the consuming instruction's `(` and the marker's start.
--- @type integer|nil
local consuming_paren = nil
if consuming_encoder then consuming_paren = find_consuming_paren(tok) end
--- @type integer
local pos = 1
while pos <= #tok do
-- Trim leading whitespace and comments before each scan.
pos = M.skip_ws_and_cmt(tok, pos)
if pos > #tok then break end
--- @type string|nil, integer
local ident, after = M.read_ident(tok, pos)
if not ident then
-- Not an ident: token is a string or comment; skip or one-step.
--- @type integer
local next_pos = M.skip_str_or_cmt(tok, pos)
pos = (next_pos > pos) and next_pos or (pos + 1)
goto continue_loop
end
if M.DELAY_MARKERS[ident] then
--- @type integer|nil
local arg_pos = nil
if consuming_encoder and consuming_paren then
arg_pos = count_top_level_commas(tok, consuming_paren + 1, pos) + 1
@@ -442,7 +585,9 @@ local function _project_emission_inner(root_body_entry, ctx_table)
goto continue_loop
end
-- Marker ident: parse the (...) arguments.
--- @type integer
local open = M.skip_ws_and_cmt(tok, after)
--- @type string|nil, integer
local inner, after_paren = M.read_parens(tok, open)
if not inner then
-- (...) Unreadable: fall back to non-marker behavior.
@@ -453,10 +598,12 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- For embedded markers, propagate the consuming_encoder + the marker's arg position
-- (1-based) so `passes/offsets.lua` can dispatch per-consuming-instruction offset encoding.
-- Offset markers are emitted only when a consuming encoder is present.
--- @type integer|nil
local arg_pos = nil
if consuming_encoder and consuming_paren then
arg_pos = count_top_level_commas(tok, consuming_paren + 1, pos) + 1
end
--- @type string[]
local args = split_call_args(inner)
if ident == "atom_label" then emit_marker("label", args[1] or "", nil, tok_line, nil, nil, consuming_encoder, arg_pos)
elseif consuming_encoder then emit_marker("offset", args[1] or "", args[2] or "", tok_line, nil, nil, consuming_encoder, arg_pos)
@@ -466,6 +613,13 @@ local function _project_emission_inner(root_body_entry, ctx_table)
end
end
--- @param inv_kind string
--- @param component_name string
--- @param call_text string
--- @param root_call_text string|nil
--- @param call_path string
--- @param call_line integer
--- @return InvocationRecord
local function emit_invoke_begin(inv_kind, component_name, call_text,
root_call_text, call_path, call_line)
next_inv_id = next_inv_id + 1
@@ -476,7 +630,9 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- The walker has already found the component body in `ctx_table.component_index[component_name]`, so the matching entry MUST exist in `ctx_table.components[component_name]`
-- (both registries are populated from the same source by the components pass).
-- A missing entry is a corpus-plumbing bug; we fail loudly here rather than silently stamp `false` and mask the regression.
--- @type table<string, ComponentDef>|nil
local components = ctx_table.components
--- @type ComponentDef|nil
local component_def = components and components[component_name] or nil
if not component_def then
error("duffle.emit_invoke_begin: component " .. string.format("%q", component_name)
@@ -486,7 +642,9 @@ local function _project_emission_inner(root_body_entry, ctx_table)
, 0
)
end
--- @type boolean
local debug_skip_stamp = component_def.debug_skip == true
--- @type InvocationRecord
local inv = {
id = next_inv_id,
parent_id = 0, -- patched below by caller
@@ -521,6 +679,8 @@ local function _project_emission_inner(root_body_entry, ctx_table)
return inv
end
--- @param inv InvocationRecord
--- @return nil
local function emit_invoke_end(inv)
-- 0-based emitted-word position of the LAST word inside this invocation.
-- After the last body word was emitted, `word_idx` was incremented past it, so `word_idx - 1` is the 0-based position of the last word.
@@ -532,6 +692,7 @@ local function _project_emission_inner(root_body_entry, ctx_table)
word_index = word_idx,
invocation_ids = open_invocation_ids_snapshot(),
}
--- @type integer
for i = #invocation_stack, 1, -1 do
if invocation_stack[i] == inv then
table.remove(invocation_stack, i)
@@ -542,9 +703,14 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- Resolve the per-token word count. If unresolved, surface ONE warning
-- and fall back to 1 opaque word so the cycle budget still accounts for the slot.
--- @param ident string
--- @param tok_line integer
--- @return integer
local function resolve_count(ident, tok_line)
--- @type WordCounts|nil
local wc = ctx_table.word_counts
if wc and wc[ident] then return wc[ident] end
--- @type string
local canon = M.gte_canon(ident)
if canon ~= ident and wc and wc[canon] then return wc[canon] end
warnings[#warnings + 1] = {
@@ -561,24 +727,40 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- walk_root_call_text: Outermost `mac_X(...)` token text (preserved across recursion).
-- walk_immediate_call_text: IMMEDIATE outer `mac_X(...)` token text for words emitted in this body — nil for the root atom body.
-- Two trackers are propagated as separate parameters so words deep inside nested expansions correctly identify both their immediate call site and the outermost call site.
--- @param body_entry ComponentBodyEntry
--- @param walk_parent_inv_id integer
--- @param walk_root_call_text string|nil
--- @param walk_immediate_call_text string|nil
--- @return nil
local function walk_body_entry(body_entry, walk_parent_inv_id,
walk_root_call_text, walk_immediate_call_text)
--- @type BodyToken[]
local tokens = body_entry.body_tokens or {}
--- @type integer
local body_off = body_entry.body_off or 0
--- @type LineIndexFn
local line_of = body_entry.line_of or M.LineIndex("")
--- @type string
local def_source = body_entry.source or ""
--- @type integer
local def_line = body_entry.declaration or 0
--- @type table<string, string>|nil
local sub_map = body_entry.sub_map
-- Per-token dispatch: each matched branch returns; only the fall-through
-- "opaque word" emit handles direct encoders + mac_X-without-component.
--- @param bt BodyToken
--- @return nil
local function process_token(bt)
--- @type string
local tok = M.trim(bt.tok or "")
-- Substituted MipsCode args can carry // comments from the call site.
while tok ~= "" do
if tok:sub(1, 2) == "//" then
--- @type integer|nil
local nl = tok:find("\n")
tok = M.trim(nl and tok:sub(nl + 1) or "")
elseif tok:sub(1, 2) == "/*" then
--- @type integer|nil
local close = tok:find("*/", 3, true)
tok = M.trim(close and tok:sub(close + 2) or "")
else
@@ -586,19 +768,25 @@ local function _project_emission_inner(root_body_entry, ctx_table)
end
end
if tok == "" then return end
--- @type string|nil, integer
local ident, after = M.read_ident(tok, 1)
if not ident then ident = "?" end
--- @type string, string[]
local _, args = token_ident_and_args(tok)
--- @type integer
local tok_line = line_of(body_off + bt.rel) or 0
if M.DELAY_MARKERS[ident] then
emit_marker("delay", ident, nil, tok_line)
--- @type string
local rest = tok:sub(after or (#tok + 1))
while true do
rest = M.trim(rest)
if rest:sub(1, 2) == "//" then
--- @type integer|nil
local nl = rest:find("\n")
rest = nl and rest:sub(nl + 1) or ""
elseif rest:sub(1, 2) == "/*" then
--- @type integer|nil
local close = rest:find("*/", 3, true)
if not close then rest = ""; break end
rest = rest:sub(close + 2)
@@ -615,6 +803,7 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- Pass `ident` as the consuming instruction so `emit_embedded_markers` can compute each marker's arg position + record the consuming_encoder for the offsets pass.
-- Canonicalize `jump_rel` to `branch_equal` (its preprocessor-expanded form) so the `consuming_encoder` metadata in marker records is canonical.
-- `jump_rel`: unconditional jump alias from `code/duffle/mips.h`.
--- @type string
local consuming_encoder_for_markers = (ident == "jump_rel") and "branch_equal" or ident
if ident ~= "atom_label" and ident ~= "atom_offset" then
emit_embedded_markers(tok, tok_line, consuming_encoder_for_markers)
@@ -628,6 +817,7 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- MipsCode formals (nop_slot1, …): the ident is a sub_map key.
-- Re-process the replacement token so load_word(...) becomes a real encoder.
if sub_map and type(sub_map[ident]) == "string" and sub_map[ident] ~= ident then
--- @type string
local repl = M.trim(sub_map[ident])
if repl ~= "" then
process_token({ tok = repl, rel = bt.rel })
@@ -635,15 +825,20 @@ local function _project_emission_inner(root_body_entry, ctx_table)
end
end
if ident:sub(1, 4) == "mac_" then
--- @type string
local bare = ident:sub(5)
--- @type ComponentBodyEntry|nil
local comp = ctx_table.component_index[bare]
if comp then
--- @type string
local invocation_root_call_text = walk_root_call_text or tok
if ctx_table.visiting[bare] then
-- Cycle: still allocate inv_id, emit zero-width begin/end, record the cycle error; do NOT recurse.
--- @type InvocationRecord
local inv = emit_invoke_begin(comp.kind or "comp_bare", bare, tok, invocation_root_call_text, def_source, tok_line)
inv.parent_id = walk_parent_inv_id
inv.call_text = tok
--- @type EmitError
local err = {
kind = "cycle",
msg = string.format("project_emission: component cycle detected: %q", bare),
@@ -657,6 +852,7 @@ local function _project_emission_inner(root_body_entry, ctx_table)
end
-- First visit: descend + count + count_mismatch-check below.
ctx_table.visiting[bare] = true
--- @type InvocationRecord
local inv = emit_invoke_begin(comp.kind or "comp_bare", bare, tok, invocation_root_call_text, def_source, tok_line)
inv.parent_id = walk_parent_inv_id
inv.call_text = tok
@@ -665,11 +861,14 @@ local function _project_emission_inner(root_body_entry, ctx_table)
-- Propagate trackers into the recursive walk:
-- immediate_call_text = this call's tok (the IMMEDIATE outer call for words emitted in this body)
-- root_call_text = the OUTERMOST call (immutable across the recursion)
--- @type string[]|nil
local formal_names = ctx_table.component_index[bare]
and ctx_table.component_index[bare].arg_names
--- @type table<string, string>|nil
local child_map = nil
if formal_names then
child_map = {}
--- @type integer, string
for i, fname in ipairs(formal_names) do
child_map[fname] = apply_sub(sub_map, args[i])
end
@@ -688,8 +887,11 @@ local function _project_emission_inner(root_body_entry, ctx_table)
ctx_table.visiting[bare] = nil
emit_invoke_end(inv)
-- Count `word` items inside [start_word, end_word].
--- @type integer
local wc_inside = 0
--- @type integer
for i = inv.start_word, inv.end_word do
--- @type EmissionItem|nil
local it = items[i]
if it and it.kind == "word" then
wc_inside = wc_inside + 1
@@ -698,8 +900,10 @@ local function _project_emission_inner(root_body_entry, ctx_table)
inv.word_count = wc_inside
-- count_mismatch is a construction error: word_counts["mac_X"] is the declared count populated by the components pass;
-- We compare against the measured word count.
--- @type integer|nil
local declared = ctx_table.word_counts["mac_" .. bare]
if declared and wc_inside ~= declared then
--- @type EmitError
local err = {
kind = "count_mismatch",
msg = string.format("project_emission: mac_%s declared=%d measured=%d", bare, declared, wc_inside),
@@ -715,13 +919,17 @@ local function _project_emission_inner(root_body_entry, ctx_table)
end
-- Direct encoder, or mac_X-without-component: resolve count + emit n words.
-- Resolve_count may emit a warning if the count is unresolved.
--- @type integer
local n = resolve_count(ident, tok_line)
--- @type string
local out_ident = (ident == "nop2") and "nop" or ident
--- @type integer
for _ = 1, n do
emit_word(out_ident, args, tok_line, tok, def_source, def_line, walk_immediate_call_text, walk_root_call_text, sub_map)
end
end
--- @type integer, BodyToken
for _, bt in ipairs(tokens) do
process_token(bt)
end
@@ -780,11 +988,12 @@ end
--- for the open invocation stack at that word.
---
--- @param body_text string -- the raw atom body string
--- @param component_index table -- bare-name → component record (corpus.component_body_index)
--- @param word_counts table -- macro name → emitted word count
--- @param components table -- bare-name → component definition (corpus.components); REQUIRED — consumed at the invocation-construction site to stamp
--- @param component_index table<string, ComponentBodyEntry>
--- @param word_counts WordCounts
--- @param components table<string, ComponentDef>
--- `invocation.debug_skip`. A missing or non-table `components` raises a fail-loud error rather than silently falling back.
--- @return EmissionProjection
--- @param reg_use_ctx RegUseCtx|nil
function M.project_emission(body_text, component_index, word_counts, components, reg_use_ctx)
-- The recursive walk delegates to `_project_emission_inner` so component bodies (which arrive as
-- `{body_tokens, body_off, line_of, source, declaration}` records from `corpus.component_body_index`)
@@ -816,6 +1025,7 @@ function M.project_emission(body_text, component_index, word_counts, components,
}
end
--- @type BodyToken[]
local tokens = M.tokenize_body(body_text)
return _project_emission_inner({
body_tokens = tokens,
@@ -849,10 +1059,17 @@ end
-- (FI_, atom_dbg_skip, comments) until it finds an ident followed by "(".
-- That ident is the function name; the parens contents are the args.
-------------------------------------------------------------------------------
--- @param source string
--- @param before_pos integer
--- @param slice_mips_code_len integer
--- @return string|nil, string|nil
function M.find_function_decl_for(source, before_pos, slice_mips_code_len)
--- @type integer
local search_pos = 1
--- @type integer|nil
local last_match = nil
while true do
--- @type integer|nil
local found = source:find("Slice_MipsCode", search_pos, true)
if not found or found >= before_pos then break end
last_match = found
@@ -860,10 +1077,12 @@ function M.find_function_decl_for(source, before_pos, slice_mips_code_len)
end
if not last_match then return nil, nil end
--- @type integer
local pos = last_match + slice_mips_code_len
while pos < before_pos do
-- skip whitespace
while pos <= #source do
--- @type string
local c = source:sub(pos, pos)
if c == " " or c == "\t" or c == "\n" or c == "\r" then
pos = pos + 1
@@ -880,17 +1099,21 @@ function M.find_function_decl_for(source, before_pos, slice_mips_code_len)
end
-- skip block comments
if source:sub(pos, pos + 1) == "/*" then
--- @type integer|nil
local close = source:find("*/", pos + 2, true)
if not close then break end
pos = close + 2
goto continue
end
-- try to read an ident
--- @type string|nil, integer
local ident, ident_end = M.read_ident(source, pos)
if not ident then break end
-- check if the next non-ws char after ident is "("
--- @type integer
local next_pos = M.skip_ws_and_cmt(source, ident_end)
if source:sub(next_pos, next_pos) == "(" then
--- @type string|nil
local inner = M.read_parens(source, next_pos)
if inner then
return ident, inner
@@ -917,11 +1140,18 @@ end
-- The walk finds the LAST "MipsAtom*" before before_pos, then skips whitespace + qualifiers (internal, I_, FI_, comments)
-- until it finds an ident followed by "(".
-------------------------------------------------------------------------------
--- @param source string
--- @param before_pos integer
--- @param mips_atom_ptr_len integer
--- @return string|nil, string|nil, string|nil, integer|nil
function M.find_atom_proc_decl_for(source, before_pos, mips_atom_ptr_len)
--- @type integer
local search_pos = 1
--- @type integer|nil
local last_match = nil
while true do
-- plain=true: "*" is literal, no escaping needed
--- @type integer|nil
local found = source:find("MipsAtom*", search_pos, true)
if not found or found >= before_pos then break end
last_match = found
@@ -929,10 +1159,12 @@ function M.find_atom_proc_decl_for(source, before_pos, mips_atom_ptr_len)
end
if not last_match then return nil, nil end
--- @type integer
local pos = last_match + mips_atom_ptr_len
while pos < before_pos do
-- skip whitespace
while pos <= #source do
--- @type string
local c = source:sub(pos, pos)
if c == " " or c == "\t" or c == "\n" or c == "\r" then
pos = pos + 1
@@ -949,17 +1181,21 @@ function M.find_atom_proc_decl_for(source, before_pos, mips_atom_ptr_len)
end
-- skip block comments
if source:sub(pos, pos + 1) == "/*" then
--- @type integer|nil
local close = source:find("*/", pos + 2, true)
if not close then break end
pos = close + 2
goto continue
end
-- try to read an ident
--- @type string|nil, integer
local ident, ident_end = M.read_ident(source, pos)
if not ident then break end
-- check if the next non-ws char after ident is "("
--- @type integer
local next_pos = M.skip_ws_and_cmt(source, ident_end)
if source:sub(next_pos, next_pos) == "(" then
--- @type string|nil, integer
local inner, after_paren = M.read_parens(source, next_pos)
if inner then
return ident, inner, ident, after_paren
scripts/duffle_isa.lua
+130
View File
@@ -1,16 +1,126 @@
--- duffle_isa.lua — encoder / GTE / hardware tables.
--- @class InstructionImm
--- @field arg integer
--- @field signed boolean|nil
--- @field width integer
--- @class InstructionValue
--- @field dest integer
--- @field op string
--- @field sources integer[]|nil
--- @field immediate integer|nil
--- @field source integer|nil
--- @class InstructionRow
--- @field cycles integer
--- @field kind string
--- @field reads integer[]|nil
--- @field writes integer[]|nil
--- @field imm InstructionImm[]|nil
--- @field value InstructionValue|nil
--- @field delay_slot boolean|nil
--- @field suppress_arg1 table<string, string>|nil -- bag: GPR ident -> reason
--- @class TapeAtomMacroRow
--- @field kind string
--- @field binds boolean
--- @class GteCommandPort
--- @field register string
--- @field role string
--- @class GteCommandLatch
--- @field register string
--- @field required integer
--- @class GteCommandRow
--- @field aliases string[]
--- @field cycles integer
--- @field inputs string[]
--- @field outputs GteCommandPort[]
--- @field latch GteCommandLatch[]
--- @class GteCrAliasGroup
--- @field [1] integer -- C2 control-register slot
--- @field [2] string[] -- aliases that share that slot
--- @class GtePackedSlotRelation
--- @field slot integer
--- @field first string
--- @field second string
--- @class HardwareRelationPort
--- @field domain string
--- @field arg integer
--- @class HardwareRelationVisibility
--- @field kind string
--- @field required integer
--- @class HardwareRelationEvidence
--- @field confidence string
--- @field source string
--- @class HardwareRelationRow
--- @field id string
--- @field semantic string
--- @field consumer string
--- @field token string
--- @field direction string
--- @field reads HardwareRelationPort
--- @field writes HardwareRelationPort
--- @field visibility HardwareRelationVisibility|nil
--- @field evidence HardwareRelationEvidence
--- @field violation_kind string
--- @field destination_match string|nil
--- @field fanout_to string[]|nil
--- @field required integer|nil
--- @field clear_on_consumer boolean|nil
--- @field stage boolean|nil
--- @field cu2_transition boolean|nil
--- @field status_register integer|nil
--- @class Cu2TransitionPolicy
--- @field status_register integer
--- @field enable_bit integer
--- @field required integer
--- @field visibility_kind string
--- @field evidence HardwareRelationEvidence
--- @class DuffleIsa
--- @field TAPE_ATOM_MACROS table<string, TapeAtomMacroRow>
--- @field DELAY_MARKERS table<string, boolean>
--- @field INSTRUCTION table<string, InstructionRow>
--- @field GTE_COMMAND table<string, GteCommandRow>
--- @field ALIAS_TO_CANONICAL table<string, string>
--- @field instr fun(ident: string): InstructionRow|nil
--- @field gte_canon fun(ident: string): string
--- @field gte fun(ident: string): GteCommandRow|nil
--- @field GTE_CR_ALIAS_GROUPS GteCrAliasGroup[]
--- @field GTE_PACKED_SLOT_RELATIONS GtePackedSlotRelation[]
--- @field OPERAND_READ_POSITIONS table<string, integer[]>
--- @field GP0_CMD_SIZE table<integer, integer>
--- @field GP0_CMD_BY_SHAPE table<string, integer>
--- @field UNKNOWN_INSTRUCTION_CYCLES integer
--- @field HARDWARE_RELATIONS HardwareRelationRow[]
--- @field CU2_TRANSITION_POLICY Cu2TransitionPolicy
--- @type DuffleIsa
local M = {}
-- Section 7: domain tables
-- ════════════════════════════════════════════════════════════════════════════
-- atom_info sub-calls: atom_bind, atom_reads, atom_writes, atom_view, atom_reg_types, atom_ctx, atom_phase.
--- @type table<string, TapeAtomMacroRow>
M.TAPE_ATOM_MACROS = {
["atom_info"] = { kind = "info", binds = false },
}
-- Empty C macros that prefix the next encoder. Zero words.
-- BdSlot_ nop is one nop word. The marker is not the BD instruction.
--- @type table<string, boolean> -- bag: marker prefix -> true
M.DELAY_MARKERS = {
["GteDelay_"] = true,
["LdSlot_"] = true,
@@ -19,6 +129,7 @@ M.DELAY_MARKERS = {
}
-- One row per encoder. Read through duffle.instr.
--- @type table<string, InstructionRow>
M.INSTRUCTION = {
["BdSlot_"] = { cycles = 0, kind = "marker", },
["LdSlot_"] = { cycles = 0, kind = "marker", },
@@ -115,6 +226,7 @@ M.INSTRUCTION = {
}
-- One row per GTE command. Alias cycle numbers live here, not on INSTRUCTION.
--- @type table<string, GteCommandRow>
M.GTE_COMMAND = {
["gte_cmdw_avsz3"] = {
aliases = { "gte_avg_sort_z3", "gte_avsz3", "gte_cmdw_avg_sort_z3" },
@@ -296,14 +408,24 @@ M.GTE_COMMAND = {
},
}
--- @param ident string
--- @return InstructionRow|nil
function M.instr (ident) return M.INSTRUCTION [ident] end
--- @param ident string
--- @return string
function M.gte_canon(ident) return M.ALIAS_TO_CANONICAL [ident] or ident end
--- @param ident string
--- @return GteCommandRow|nil
function M.gte (ident) return M.GTE_COMMAND[M.gte_canon(ident)] end
--- @return nil
local function build_alias_map()
--- @type table<string, string> -- bag: alias or canon -> canon
M.ALIAS_TO_CANONICAL = {}
--- @type string, GteCommandRow
for canon, row in pairs(M.GTE_COMMAND) do
M.ALIAS_TO_CANONICAL[canon] = canon
--- @type integer, string
for _, alias in ipairs(row.aliases or {}) do
M.ALIAS_TO_CANONICAL[alias] = canon
end
@@ -319,6 +441,7 @@ build_alias_map()
--- Cross-alias writes inside one atom body, or across the wave-context boundary, silently clobber each other.
--- The `check_gte_cr_alias_writes` check warns about each pair per source. See `docs/gte_reference.md` §"Control-register alias table"
--- for the HW rationale and the libgte outer-product convention.
--- @type GteCrAliasGroup[]
M.GTE_CR_ALIAS_GROUPS = {
{ 24, { "gte_cr_RBK", "gte_cr_OFX" } }, -- background R vs screen offset X
{ 25, { "gte_cr_GBK", "gte_cr_OFY" } }, -- background G vs screen offset Y
@@ -326,6 +449,7 @@ M.GTE_CR_ALIAS_GROUPS = {
}
-- Packed RT slots named by the gte.h packed-slot comment. First must be written before second.
--- @type GtePackedSlotRelation[]
M.GTE_PACKED_SLOT_RELATIONS = {
{ slot = 2, first = "gte_cr_RT13", second = "gte_cr_RT22" },
}
@@ -340,6 +464,7 @@ M.GTE_PACKED_SLOT_RELATIONS = {
-- * The check tracks one entry per destination GPR per MFC2 / CFC2 event.
-- A subsequent event counts as a "use" iff any of its read operand positions reference that destination GPR's ident (e.g. `R_T0`).
-- * Branch delay slots are out of scope (MIPS control-flow; tracked separately).
--- @type table<string, integer[]> -- bag: encoder ident -> GPR operand positions
M.OPERAND_READ_POSITIONS = {
-- CPU ALU with one or two GPR operands. Reads every GPR operand.
["add_ui"] = {1, 2},
@@ -439,6 +564,7 @@ M.OPERAND_READ_POSITIONS = {
-- set_poly_gt3(p) -> set_len(p, 9) -> 10 total GP0 0x34
-- set_poly_g4(p) -> set_len(p, 8) -> 9 total GP0 0x38
-- set_poly_gt4(p) -> set_len(p, 12) -> 13 total GP0 0x3C
--- @type table<integer, integer> -- bag: GP0 cmd byte -> word count
M.GP0_CMD_SIZE = {
[0x20] = 5, -- Poly_F3
[0x24] = 8, -- Poly_FT3
@@ -452,6 +578,7 @@ M.GP0_CMD_SIZE = {
-- Shape suffix (after `ac_format_` / `mac_format_` prefix) -> GP0 cmd byte.
-- Lets the static-analysis check derive the cmd byte from a macro name like `mac_format_g4_color` -> `g4` -> 0x38 -> 9 expected words.
--- @type table<string, integer> -- bag: shape suffix -> GP0 cmd byte
M.GP0_CMD_BY_SHAPE = {
["f3"] = 0x20, ["ft3"] = 0x24,
["f4"] = 0x28, ["ft4"] = 0x2C,
@@ -459,6 +586,7 @@ M.GP0_CMD_BY_SHAPE = {
["g4"] = 0x38, ["gt4"] = 0x3C,
}
--- @type integer
M.UNKNOWN_INSTRUCTION_CYCLES = 1
-- Hardware-relation policy table.
@@ -491,6 +619,7 @@ M.UNKNOWN_INSTRUCTION_CYCLES = 1
-- * passes/static_analysis.lua::analyze_hardware_relations (forward walker).
-- * passes/static_analysis.lua::transfer_hazards CHECK_RULES reader (renders hazards onto `findings`).
-- This table is consumed by the hardware-relation analyzer and hazard renderer.
--- @type HardwareRelationRow[]
M.HARDWARE_RELATIONS = {
-- CPU → COP2 data register (MTC2). The ordinary default is 2 cached words between producer and consumer (cpuspecifications.md:407-419).
{
@@ -681,6 +810,7 @@ M.HARDWARE_RELATIONS = {
-- Bounded Status/SR.CU2 transition policy.
-- The value lattice and the transition consumer both read this immutable row; no second value pass is permitted.
-- The source says the enable/disable transition takes "2 clock cycles or so", so the boundary is conservative rather than exact.
--- @type Cu2TransitionPolicy
M.CU2_TRANSITION_POLICY = {
status_register = 12,
enable_bit = 0x40000000,
scripts/duffle_paths.lua
+14
View File
@@ -16,9 +16,14 @@
--- Net effect: the caller gets the duffle module in one statement; no separate `dofile(...)` + `require("duffle")` dance.
---
--- @class DufflePaths
--- @field setup fun(): nil
--- @type DufflePaths
local M = {}
-- Cache key for the repo root. Stored in `package.loaded` (process-global) so all 8 entry scripts + passes scripts share one resolution.
--- @type string
local CACHE_KEY = "__duffle_repo_root__"
--- Resolve the repo root from this script's own path. Zero shell spawn.
@@ -30,13 +35,16 @@ local CACHE_KEY = "__duffle_repo_root__"
local function find_repo_root()
if package.loaded[CACHE_KEY] then return package.loaded[CACHE_KEY] end
--- @type string
local source = debug.getinfo(1, "S").source
-- Strip the leading `@` (Lua's dofile marker) and the trailing `/duffle_paths.lua` filename.
-- What remains is the directory containing this script, i.e. `<repo>/scripts/`.
--- @type string|nil
local scripts_dir = source and source:match("^@?(.*)[/\\]duffle_paths%.lua$")
if not scripts_dir then return nil end
-- The repo root is the parent of `scripts/`. Strip the trailing `scripts/` (with or without trailing slash).
--- @type string
local root = scripts_dir:gsub("scripts[\\/]?$", "")
root = root:gsub("\\", "/")
if root == "" then root = "./" end
@@ -50,7 +58,9 @@ end
--- This script does NOT touch the OS environment: no `os.setenv`, no `os.putenv`, no `$PATH` mods.
--- It just sets `package.path` and `package.cpath` (the standard Lua way to register module search dirs).
--- lpeg is built by `update_deps.ps1` to `toolchain/lpeg/`, which we wire into `package.cpath` here (so `require("lpeg")` from `duffle.lua` resolves without any global state).
--- @return nil
function M.setup()
--- @type string|nil
local repo_root = find_repo_root()
if not repo_root then
-- Unreachable in practice: find_repo_root() derives the repo root from this script's own source path via debug.getinfo(1, "S").source (no subprocess, no git CLI, <1ms).
@@ -59,7 +69,9 @@ function M.setup()
os.exit(2)
end
--- @type string
local scripts_dir = repo_root .. "scripts/"
--- @type string
local passes_dir = repo_root .. "scripts/passes/"
package.path = scripts_dir .. "?.lua;"
.. scripts_dir .. "?/init.lua;"
@@ -70,7 +82,9 @@ function M.setup()
-- lpeg: built by `update_deps.ps1` to `toolchain/lpeg/lpeg.dll`.
-- lfs: compiled from pcsx-redux's vendored luafilesystem source to `toolchain/lfs/lfs.dll`.
-- Wire both directories into cpath so `require("lpeg")` and `require("lfs")` resolve.
--- @type string
local lpeg_dir = repo_root .. "toolchain/lpeg/"
--- @type string
local lfs_dir = repo_root .. "toolchain/lfs/"
package.cpath = lpeg_dir .. "?.dll;"
.. lfs_dir .. "?.dll;"
scripts/duffle_scan.lua
+349 -25
View File
File diff suppressed because it is too large Load Diff
scripts/elf32.lua
+159 -18
View File
@@ -22,6 +22,94 @@
-- spec: System V ABI gABI v1.2 §"ELF Header" (Table 1) + §"Section Header Table"
-- spec: System V ABI gABI v1.2 §"Symbol Table" (Elf32_Sym layout)
--- @class Elf32Adapter
--- @field read_u8_at fun(off: integer): integer|nil
--- @field read_u16_at fun(off: integer): integer|nil
--- @field read_u32_at fun(off: integer): integer|nil
--- @field read_size fun(): integer
--- @class Elf32Header
--- @field e_entry integer
--- @field e_shoff integer
--- @field e_shentsize integer
--- @field e_shnum integer
--- @field e_shstrndx integer
--- @field error string|nil
--- @class Elf32Section
--- @field sh_name integer
--- @field sh_type integer
--- @field sh_flags integer
--- @field sh_addr integer
--- @field sh_offset integer
--- @field sh_size integer
--- @field sh_link integer
--- @field name string
--- @class Elf32Sym
--- @field value integer
--- @field size integer
--- @field info integer
--- @field shndx integer
--- @class Elf32HeaderLayout
--- @field magic_offset integer
--- @field magic string
--- @field class_offset integer
--- @field endian_offset integer
--- @field header_bytes integer
--- @field e_entry_offset integer
--- @field e_shoff_offset integer
--- @field e_shentsize_offset integer
--- @field e_shnum_offset integer
--- @field e_shstrndx_offset integer
--- @class Elf32SectionLayout
--- @field sh_name_offset integer
--- @field sh_type_offset integer
--- @field sh_flags_offset integer
--- @field sh_addr_offset integer
--- @field sh_offset_offset integer
--- @field sh_size_offset integer
--- @field sh_link_offset integer
--- @field sh_entsize_bytes integer
--- @class Elf32SymLayout
--- @field st_name integer
--- @field st_value integer
--- @field st_size integer
--- @field st_info integer
--- @field sym_entry_bytes integer
--- @class Elf32Mod
--- @field ELFCLASS32 integer
--- @field ELFDATA2LSB integer
--- @field EM_MIPS integer
--- @field SHT_SYMTAB integer
--- @field SHT_STRTAB integer
--- @field SHT_NOBITS integer
--- @field SHF_WRITE integer
--- @field SHF_ALLOC integer
--- @field SHF_EXECINSTR integer
--- @field ELF32_HEADER Elf32HeaderLayout
--- @field ELF32_SECTION Elf32SectionLayout
--- @field ELF32_SYM Elf32SymLayout
--- @field dw_dwarf32_terminator integer
--- @field read_u32 fun(adapter: Elf32Adapter, off: integer): integer|nil
--- @field read_u16 fun(adapter: Elf32Adapter, off: integer): integer|nil
--- @field read_u8 fun(adapter: Elf32Adapter, off: integer): integer|nil
--- @field size fun(adapter: Elf32Adapter): integer
--- @field read_u32_le fun(buf: string, off: integer): integer
--- @field read_u16_le fun(buf: string, off: integer): integer
--- @field validate_adapter fun(adapter: any): boolean, string|nil
--- @field get_str fun(strtab: string, off: integer): string|nil
--- @field parse_elf32_headers fun(adapter: Elf32Adapter): Elf32Header|nil, string|nil
--- @field walk_sections fun(adapter: Elf32Adapter, hdr: Elf32Header): Elf32Section[]|nil, string|nil
--- @field read_section_bytes fun(adapter: Elf32Adapter, section: Elf32Section): string|nil
--- @field read_named_section fun(adapter: Elf32Adapter, sections: Elf32Section[], name: string): string|nil, string|nil
--- @field collect_symbols fun(adapter: Elf32Adapter, sections: Elf32Section[]): table<string, Elf32Sym>|nil, string|nil
--- @type Elf32Mod
local M = {}
-- ════════════════════════════════════════════════════════════════════════════
@@ -39,7 +127,7 @@ local M = {}
--- **Call form:** explicit-pass. The reader receives `adapter` as the first positional argument and the offset as the second; no `self` is passed.
--- Test fixtures declare `function(offset) ... end` and the parsers call them via dot syntax `adapter.read_u8_at(off)`.
--- The colon form `adapter:read_u8_at(off)` would prepend the adapter table as `offset` and break the contract.
--- @param adapter table
--- @param adapter Elf32Adapter
--- @param off integer -- zero-based wire offset
--- @return integer|nil
function M.read_u32(adapter, off)
@@ -50,7 +138,7 @@ function M.read_u32(adapter, off)
end
--- Read a 2-byte little-endian unsigned integer from `adapter` at zero-based wire offset `off`.
--- @param adapter table
--- @param adapter Elf32Adapter
--- @param off integer -- zero-based wire offset
--- @return integer|nil
function M.read_u16(adapter, off)
@@ -59,7 +147,7 @@ function M.read_u16(adapter, off)
end
--- Read a 1-byte unsigned integer from `adapter` at zero-based wire offset `off`.
--- @param adapter table
--- @param adapter Elf32Adapter
--- @param off integer -- zero-based wire offset
--- @return integer|nil
function M.read_u8(adapter, off)
@@ -67,7 +155,7 @@ function M.read_u8(adapter, off)
end
--- Total adapter byte length.
--- @param adapter table
--- @param adapter Elf32Adapter
--- @return integer
function M.size(adapter)
return adapter.read_size()
@@ -76,7 +164,11 @@ end
--- Forwarders kept for backward compat with scripts/elf_dwarf.lua.
--- The metaprogram side keeps `read_u32_le` / `read_u16_le`;
--- both layers now use the same byte-level helpers under the hood.
--- @param buf string
--- @param off integer
--- @return integer
function M.read_u32_le(buf, off)
--- @type integer
local byte_off = off + 1
return buf:byte(byte_off)
+ buf:byte(byte_off + 0x01) * 0x00000100
@@ -89,6 +181,7 @@ end
--- @param off integer -- zero-based wire offset
--- @return integer
function M.read_u16_le(buf, off)
--- @type integer
local byte_off = off + 1
return buf:byte(byte_off) + buf:byte(byte_off + 0x01) * 0x00000100
end
@@ -116,6 +209,7 @@ M.SHF_EXECINSTR = 0x4 -- spec: gABI v1.2 §"Section Attributes" — executable
-- ELF32 header layout (System V ABI gABI v1.2 §"ELF Header" Table 1)
-- ---------------------------------------------------------------------------
-- All offsets are zero-based wire offsets. The header is 52 bytes total (header_bytes = 0x34 = 52).
--- @type Elf32HeaderLayout
M.ELF32_HEADER = {
magic_offset = 0x00, -- 4 bytes; expected "\127ELF"
magic = "\127ELF",
@@ -134,6 +228,7 @@ M.ELF32_HEADER = {
-- ---------------------------------------------------------------------------
-- Each entry is 40 bytes (sh_entsize_bytes = 0x28 = 40);
-- zero-based, field offsets relative to the start of the entry.
--- @type Elf32SectionLayout
M.ELF32_SECTION = {
sh_name_offset = 0x00, -- 4-byte LE; offset into .shstrtab
sh_type_offset = 0x04, -- 4-byte LE; section type (SHT_*)
@@ -150,6 +245,7 @@ M.ELF32_SECTION = {
-- ---------------------------------------------------------------------------
-- Each entry is 16 bytes (sym_entry_bytes = 0x10 = 16);
-- zero-based, field offsets relative to the start of the entry.
--- @type Elf32SymLayout
M.ELF32_SYM = {
st_name = 0x00, -- 4-byte LE; offset into the linked string table
st_value = 0x04, -- 4-byte LE; symbol value (address / absolute)
@@ -190,6 +286,7 @@ end
--- @return string|nil
function M.get_str(strtab, off)
if off < 0 or off >= #strtab then return nil end
--- @type integer|nil
local end_pos = strtab:find("\0", off + 1, true)
if not end_pos then return nil end
return strtab:sub(off + 1, end_pos - 1)
@@ -205,38 +302,50 @@ end
--- On failure returns nil + a stable error code:
--- bad_magic, unsupported_elf_class, unsupported_elf_data, truncated_header
--- The header's machine field is NOT validated here — callers (e.g. the helper's prime path) decide whether to require EM_MIPS before symbol reads.
--- @param adapter table
--- @return table|nil, string|nil
--- @param adapter Elf32Adapter
--- @return Elf32Header|nil, string|nil
function M.parse_elf32_headers(adapter)
--- @type boolean, string|nil
local ok, err = M.validate_adapter(adapter)
if not ok then return nil, err end
-- 4-byte magic: 0x7F 'E' 'L' 'F'.
-- The byte readers take the adapter explicitly.
-- The production `Support.File` adapter is wrapped by the caller to drop its implicit `self` so the parser shape is flat pass-style.
--- @type integer|nil
local b1 = M.read_u8(adapter, 0)
--- @type integer|nil
local b2 = M.read_u8(adapter, 1)
--- @type integer|nil
local b3 = M.read_u8(adapter, 2)
--- @type integer|nil
local b4 = M.read_u8(adapter, 3)
if not (b1 and b2 and b3 and b4)
or not (b1 == 0x7f and b2 == 0x45 and b3 == 0x4c and b4 == 0x46) then
return nil, "bad_magic"
end
--- @type integer|nil
local class = M.read_u8(adapter, M.ELF32_HEADER.class_offset)
if class ~= M.ELFCLASS32 then
return nil, "unsupported_elf_class"
end
--- @type integer|nil
local data = M.read_u8(adapter, M.ELF32_HEADER.endian_offset)
if data ~= M.ELFDATA2LSB then
return nil, "unsupported_elf_data"
end
--- @type integer|nil
local e_entry = M.read_u32(adapter, M.ELF32_HEADER.e_entry_offset)
--- @type integer|nil
local e_shoff = M.read_u32(adapter, M.ELF32_HEADER.e_shoff_offset)
--- @type integer|nil
local e_shentsize = M.read_u16(adapter, M.ELF32_HEADER.e_shentsize_offset)
--- @type integer|nil
local e_shnum = M.read_u16(adapter, M.ELF32_HEADER.e_shnum_offset)
--- @type integer|nil
local e_shstrndx = M.read_u16(adapter, M.ELF32_HEADER.e_shstrndx_offset)
if not (e_entry and e_shoff and e_shentsize and e_shnum and e_shstrndx) then
return nil, "truncated_header"
@@ -254,10 +363,11 @@ end
--- Read one section-header entry from `adapter` at `sh_off`.
--- Returns a table with the wire fields plus a (yet-unresolved) `name` field.
--- @param adapter table
--- @param adapter Elf32Adapter
--- @param sh_off integer
--- @return table|nil, string|nil -- entry, error
--- @return Elf32Section|nil, string|nil
local function read_section_entry(adapter, sh_off)
--- @type Elf32Section
local entry = {
sh_name = M.read_u32(adapter, sh_off + M.ELF32_SECTION.sh_name_offset),
sh_type = M.read_u32(adapter, sh_off + M.ELF32_SECTION.sh_type_offset),
@@ -279,21 +389,26 @@ end
--- (the section at logical index 0 is at array position 1, etc.).
--- Each entry has the wire fields plus a resolved `name` derived from `.shstrtab`.
--- Returns nil + a stable error code on failure: truncated_section_headers, missing_shstrtab, truncated_strtab
--- @param adapter table
--- @param hdr table -- the table returned by parse_elf32_headers
--- @return table|nil, string|nil
--- @param adapter Elf32Adapter
--- @param hdr Elf32Header
--- @return Elf32Section[]|nil, string|nil
function M.walk_sections(adapter, hdr)
if not hdr or hdr.error then return nil, hdr and hdr.error or "truncated_section_headers" end
--- @type integer
local file_size = M.size(adapter)
if hdr.e_shoff + hdr.e_shnum * hdr.e_shentsize > file_size then
return nil, "truncated_section_headers"
end
-- Read every section header first; we need .shstrtab to resolve names.
--- @type Elf32Section[]
local sections = {}
--- @type integer
for i = 0, hdr.e_shnum - 1 do
--- @type integer
local sh_off = hdr.e_shoff + i * hdr.e_shentsize
--- @type Elf32Section|nil, string|nil
local entry, err = read_section_entry(adapter, sh_off)
if not entry then return nil, err end
sections[i + 1] = entry
@@ -303,6 +418,7 @@ function M.walk_sections(adapter, hdr)
return nil, "missing_shstrtab"
end
--- @type Elf32Section|nil
local shstrtab = sections[hdr.e_shstrndx + 1]
if not shstrtab or shstrtab.sh_type ~= M.SHT_STRTAB then
return nil, "missing_shstrtab"
@@ -310,9 +426,11 @@ function M.walk_sections(adapter, hdr)
if shstrtab.sh_offset + shstrtab.sh_size > file_size then
return nil, "truncated_section_headers"
end
--- @type string|nil
local shstrtab_bytes = M.read_section_bytes(adapter, shstrtab)
if not shstrtab_bytes then return nil, "truncated_section_headers" end
--- @type integer, Elf32Section
for _, s in ipairs(sections) do
s.name = M.get_str(shstrtab_bytes, s.sh_name) or ""
end
@@ -322,14 +440,18 @@ end
--- Read the bytes of one section. Returns a string, or nil if the adapter returns nil for any byte (out-of-bounds).
--- The caller is responsible fors sizing the buffer (the section's sh_offset + sh_size must fit in adapter.size).
--- @param adapter table
--- @param section table -- one entry from walk_sections
--- @param adapter Elf32Adapter
--- @param section Elf32Section
--- @return string|nil
function M.read_section_bytes(adapter, section)
--- @type integer
local size = section.sh_size
if size == 0 then return "" end
--- @type string[]
local out = {}
--- @type integer
for i = 0, size - 1 do
--- @type integer|nil
local b = M.read_u8(adapter, section.sh_offset + i)
if b == nil then return nil end
out[#out + 1] = string.char(b)
@@ -339,14 +461,16 @@ end
--- Convenience: walk sections, then look up the named section, then read its bytes.
--- Returns nil + a stable error code if the section is absent or out-of-bounds.
--- @param adapter table
--- @param sections table -- 1-based array from walk_sections
--- @param adapter Elf32Adapter
--- @param sections Elf32Section[]
--- @param name string
--- @return string|nil, string|nil
function M.read_named_section(adapter, sections, name)
if not sections then return nil, "missing_section" end
--- @type integer, Elf32Section
for _, s in ipairs(sections) do
if s.name == name then
--- @type string|nil
local bytes = M.read_section_bytes(adapter, s)
if not bytes then return nil, "truncated_section_data" end
return bytes, nil
@@ -359,15 +483,19 @@ end
--- Each stored entry is `{ value = st_value, size = st_size, info = st_info, shndx = st_shndx }`.
--- Both STB_LOCAL and STB_GLOBAL symbols are included; the live ELF stores `smem` as a local symbol.
--- Returns nil + a stable error code on failure: missing_symtab_strtab, truncated_section_headers
--- @param adapter table
--- @param sections table
--- @return table|nil, string|nil
--- @param adapter Elf32Adapter
--- @param sections Elf32Section[]
--- @return table<string, Elf32Sym>|nil, string|nil
function M.collect_symbols(adapter, sections)
if not sections then return nil, "missing_sections" end
--- @type table<string, Elf32Sym> -- bag: symbol name -> Elf32Sym
local symbols = {}
--- @type integer
local file_size = M.size(adapter)
--- @type integer, Elf32Section
for _, s in ipairs(sections) do
if s.sh_type == M.SHT_SYMTAB then
--- @type Elf32Section|nil
local strtab = sections[s.sh_link + 1]
if not strtab or strtab.sh_type ~= M.SHT_STRTAB then
return nil, "missing_symtab_strtab"
@@ -375,30 +503,43 @@ function M.collect_symbols(adapter, sections)
if strtab.sh_offset + strtab.sh_size > file_size then
return nil, "truncated_section_headers"
end
--- @type string|nil
local strtab_bytes = M.read_section_bytes(adapter, strtab)
if not strtab_bytes then return nil, "truncated_section_headers" end
if s.sh_offset + s.sh_size > file_size then
return nil, "truncated_section_headers"
end
--- @type string|nil
local symtab_bytes = M.read_section_bytes(adapter, s)
if not symtab_bytes then return nil, "truncated_section_headers" end
--- @type number
local n = #symtab_bytes / M.ELF32_SYM.sym_entry_bytes
--- @type integer
for j = 0, n - 1 do
--- @type integer
local e = s.sh_offset + j * M.ELF32_SYM.sym_entry_bytes
--- @type integer|nil
local st_name = M.read_u32(adapter, e + M.ELF32_SYM.st_name)
if st_name then
--- @type integer|nil
local st_value = M.read_u32(adapter, e + M.ELF32_SYM.st_value)
--- @type integer|nil
local st_size = M.read_u32(adapter, e + M.ELF32_SYM.st_size)
--- @type integer|nil
local st_info = M.read_u8(adapter, e + M.ELF32_SYM.st_info)
-- st_shndx is at offset 14 (2 bytes) — derived from the layout
-- the metaprogram reads too. Inline the read to keep the
-- adapter as the only I/O surface.
--- @type integer|nil
local b1 = M.read_u8(adapter, e + 14)
--- @type integer|nil
local b2 = M.read_u8(adapter, e + 15)
if not (b1 and b2) then
return nil, "truncated_section_headers"
end
--- @type integer
local st_shndx = b1 + b2 * 0x100
--- @type string
local name = M.get_str(strtab_bytes, st_name) or ""
if name ~= "" then
symbols[name] = {
scripts/elf_dwarf.lua
+388 -8
View File
@@ -9,13 +9,118 @@
-- ════════════════════════════════════════════════════════════════════════════
-- lfs is wired into package.cpath by `duffle_paths.lua` (vendored under `toolchain/lfs/lfs.dll`).
-- Elf32Adapter / Elf32Header / Elf32Section / Elf32Sym / Elf32Mod: see elf32.lua.
-- Path: see duffle.lua. NmAddr: see passes/atoms_source_map.lua.
--- @class LfsMod
--- @field attributes fun(path: string, request: string): string|nil
--- @class Dwarf4Aranges
--- @field unit_length_offset integer
--- @field version_offset integer
--- @field cu_offset_offset integer
--- @field addr_size_offset integer
--- @field seg_size_offset integer
--- @field entry_size integer
--- @field terminator_size integer
--- @field version_expected integer
--- @field addr_size_expected integer
--- @field seg_size_expected integer
--- @class Dwarf5Rnglists
--- @field unit_length_offset integer
--- @field version_offset integer
--- @field addr_size_offset integer
--- @field seg_size_offset integer
--- @field offset_count_offset integer
--- @field first_entry_offset integer
--- @field end_of_list integer
--- @field start_length integer
--- @field version_expected integer
--- @field addr_size_expected integer
--- @field seg_size_expected integer
--- @field offset_count_expected integer
--- @class DwarfLineOps
--- @field DW_LNS_extended integer
--- @field DW_LNS_copy integer
--- @field DW_LNS_advance_pc integer
--- @field DW_LNS_advance_line integer
--- @field DW_LNS_set_file integer
--- @field DW_LNS_negate_stmt integer
--- @field DW_LNE_end_sequence integer
--- @field DW_LNE_set_address integer
--- @field opcode_base integer
--- @field line_base integer
--- @field line_range integer
--- @field end_sequence_payload_size integer
--- @field set_address_payload_size integer
--- @class Dwarf5DebugLine
--- @field version_offset_post_il integer
--- @field addr_size_offset integer
--- @field seg_size_offset integer
--- @field header_length_offset integer
--- @field program_header_start integer
--- @field form_addr_bytes integer
--- @field form_strp_bytes integer
--- @field form_data16_bytes integer
--- @field form_line_strp integer
--- @field form_string integer
--- @field form_udata integer
--- @field form_data16 integer
--- @field lnct_path integer
--- @field lnct_directory_index integer
--- @field lnct_md5 integer
--- @class AbbrevAttr
--- @field name integer
--- @field form integer
--- @class AbbrevDecl
--- @field code integer
--- @field tag integer
--- @field has_children integer
--- @field attrs AbbrevAttr[]
--- @class ElfDwarf
--- @field DW_TAG table<string, integer> -- bag: tag name -> encoding
--- @field DW_AT table<string, integer> -- bag: attr name -> encoding
--- @field DW_FORM table<string, integer> -- bag: form name -> encoding
--- @field DW_ATE table<string, integer> -- bag: base-type encoding name -> code
--- @field MIPS_BYTES_PER_WORD integer
--- @field dw_dwarf32_terminator integer
--- @field DWARF4_ARANGES Dwarf4Aranges
--- @field DWARF5_RNGLISTS Dwarf5Rnglists
--- @field DWARF_LINE_OPS DwarfLineOps
--- @field DWARF5_DEBUG_LINE Dwarf5DebugLine
--- @field read_u32_le fun(buf: string, off: integer): integer
--- @field read_u16_le fun(buf: string, off: integer): integer
--- @field read_uleb128_at fun(buf: string, pos: integer): integer|nil, integer
--- @field read_sleb128_at fun(buf: string, pos: integer): integer|nil, integer
--- @field find_abbrev_table_end fun(table_bytes: string, table_start: integer): integer|nil
--- @field read_ref_sig8 fun(buf: string, pos: integer): integer, integer, integer
--- @field find_type_unit_by_signature fun(info: string, target_sig_lo: integer, target_sig_hi: integer): integer|nil, integer|nil
--- @field write_u32_le fun(value: integer): string
--- @field write_u16_le fun(value: integer): string
--- @field read_elf_sections fun(elf_path: Path, section_names: string[]): table<string, string>
--- @field read_nm fun(elf_path: Path): table<string, NmAddr>
--- @field uleb128 fun(n: integer): string
--- @field sleb128 fun(n: integer): string
--- @field uleb128_size fun(n: integer): integer
--- @field sleb128_size fun(n: integer): integer
--- @field read_line_unit_file_table fun(elf_path: string): table<string, integer>|nil, table<integer, string>|nil, table<integer, string>|nil
--- @type LfsMod
local lfs = require("lfs")
-- scripts/elf32.lua contains format-constant tables + the byte-level walker.
-- The this file re-exports `read_u32_le` / `read_u16_le` (and the DWARF32 terminator).
-- read_u32_le is this module's reader; implementation in elf32.lua.
--- @type Elf32Mod
local E = require("elf32")
--- @type ElfDwarf
local M = {}
-- ════════════════════════════════════════════════════════════════════════════
@@ -23,6 +128,8 @@ local M = {}
-- ════════════════════════════════════════════
-- (DWARF5 §7.5.5 "Tag Encodings" + Table 7.1; gcc emits these exact values for the DWARF3-extension and DWARF5 line units.)
--- bag: DWARF tag name -> encoding
--- @type table<string, integer>
M.DW_TAG = {
compile_unit = 0x11,
subprogram = 0x2E,
@@ -38,6 +145,8 @@ M.DW_TAG = {
-- We index the canonical gcc-emitted tags. Anything else falls through.
}
--- bag: DWARF attr name -> encoding
--- @type table<string, integer>
M.DW_AT = {
name = 0x03,
low_pc = 0x11,
@@ -59,6 +168,8 @@ M.DW_AT = {
decl_line = 0x3B,
}
--- bag: DWARF form name -> encoding
--- @type table<string, integer>
M.DW_FORM = {
addr = 0x01,
data1 = 0x0B,
@@ -75,6 +186,8 @@ M.DW_FORM = {
sec_offset = 0x17,
}
--- bag: DWARF base-type encoding name -> code
--- @type table<string, integer>
M.DW_ATE = {
address = 0x01,
boolean = 0x02,
@@ -87,6 +200,7 @@ M.DW_ATE = {
}
-- DWARF5 §7.5.6 DW_FORM_implicit_const
--- @type integer
local DW_FORM_implicit_const = 0x21
-- ════════════════════════════════════════════════════════════════════════════
@@ -120,6 +234,7 @@ M.dw_dwarf32_terminator = E.dw_dwarf32_terminator
-- All offsets are zero-based wire offsets.
--- spec: DWARF5 spec §7.4 (Address Range Table) — 32-bit DWARF form
--- @type Dwarf4Aranges
M.DWARF4_ARANGES = {
unit_length_offset = 0x00, -- 4-byte LE; length of unit body (excludes these 4 bytes)
version_offset = 0x04, -- 2-byte LE; expected = 2
@@ -139,6 +254,7 @@ M.DWARF4_ARANGES = {
-- All offsets are zero-based wire offsets.
--- spec: DWARF5 spec §2.17 + §7.21 (Range List Table) — 32-bit DWARF form
--- @type Dwarf5Rnglists
M.DWARF5_RNGLISTS = {
unit_length_offset = 0x00, -- 4-byte LE
version_offset = 0x04, -- 2-byte LE; expected = 5
@@ -161,6 +277,7 @@ M.DWARF5_RNGLISTS = {
-- Compare to the *_offset fields above which are hex.
--- spec: DWARF5 spec §6.2.5 (Line Number Program Opcodes)
--- @type DwarfLineOps
M.DWARF_LINE_OPS = {
-- Standard opcodes (§6.2.5.2)
DW_LNS_extended = 0, -- spec: §6.2.5.2 — extended opcode marker byte
@@ -197,6 +314,7 @@ M.DWARF_LINE_OPS = {
-- These are documented inline at each parse site in read_line_unit_file_table below.
--- spec: DWARF5 spec §6.2.4 (Line Number Program Header — version >= 5)
--- @type Dwarf5DebugLine
M.DWARF5_DEBUG_LINE = {
-- Header fields (zero-based, AFTER unit_length has been read).
version_offset_post_il = 0x00, -- 2-byte LE; expected = 5
@@ -255,10 +373,17 @@ end
-- Offsets are 0-based; returns (value, next_pos).
-- Promoted from `local function` to M.* exports so passes/dwarf_injection.lua can import them as file-scope locals per the 2nd-caller lift precedent
-- (the uleb128 + sleb128 encoders were promoted the same way).
--- @param buf string
--- @param pos integer
--- @return integer|nil
--- @return integer
function M.read_uleb128_at(buf, pos)
--- @type integer, integer
local value, shift = 0, 0
--- @type integer
local len = #buf
while pos < len do
--- @type integer
local b = buf:byte(pos + 1)
value = value + (b % 0x80) * (2 ^ shift)
shift = shift + 7
@@ -268,10 +393,17 @@ function M.read_uleb128_at(buf, pos)
return nil, pos
end
--- @param buf string
--- @param pos integer
--- @return integer|nil
--- @return integer
function M.read_sleb128_at(buf, pos)
--- @type integer, integer
local value, shift = 0, 0
--- @type integer
local len = #buf
while pos < len do
--- @type integer
local b = buf:byte(pos + 1)
value = value + (b % 0x80) * (2 ^ shift)
shift = shift + 7
@@ -287,27 +419,36 @@ end
-- Find the 0-based offset of the table-terminator byte (a single 0) for the abbrev table starting at `table_start`.
-- Returns nil on truncated input. Walks declaration headers
-- (code, tag, has_children, attr/form pairs, DW_FORM_implicit_const constant) until it finds a 0 byte that follows a complete declaration.
--- @param table_bytes string
--- @param table_start integer
--- @return integer|nil
function M.find_abbrev_table_end(table_bytes, table_start)
--- @type integer, integer
local pos, len = table_start, #table_bytes
if pos >= len or table_bytes:byte(pos + 1) == 0 then return pos end
while pos < len do
--- @type integer|nil, integer
local _code, code_end = M.read_uleb128_at(table_bytes, pos)
if not _code then return nil end
pos = code_end
--- @type integer|nil, integer
local _tag, tag_end = M.read_uleb128_at(table_bytes, pos)
if not _tag then return nil end
pos = tag_end
if pos >= len then return nil end
pos = pos + 1 -- has_children byte
while pos < len do
--- @type integer|nil, integer
local attr, attr_end = M.read_uleb128_at(table_bytes, pos)
if not attr then return nil end
pos = attr_end
--- @type integer|nil, integer
local form, form_end = M.read_uleb128_at(table_bytes, pos)
if not form then return nil end
pos = form_end
if attr == 0 and form == 0 then break end
if form == DW_FORM_implicit_const then
--- @type integer|nil, integer
local _c, ce = M.read_sleb128_at(table_bytes, pos)
if not _c then return nil end
pos = ce
@@ -321,8 +462,13 @@ end
-- Read the null-terminated C string at 0-based offset `off` in `buf`.
-- Stops at the first 0 byte or end of buffer.
--- @param buf string
--- @param off integer
--- @return string
local function read_c_string_at(buf, off)
--- @type integer
local len = #buf
--- @type integer
local start = off
while off < len and buf:byte(off + 1) ~= 0 do off = off + 1 end
return buf:sub(start + 1, off)
@@ -331,31 +477,45 @@ end
-- Walk the .debug_abbrev table starting at 0-based offset `table_start` and return a list of declarations:
-- {code, tag, has_children, attrs={ {name, form}, ... }}.
-- Stops at the table terminator.
--- @param table_bytes string
--- @param table_start integer
--- @return AbbrevDecl[]|nil
--- @return string|nil
local function parse_abbrev_table(table_bytes, table_start)
--- @type integer|nil
local table_end = M.find_abbrev_table_end(table_bytes, table_start)
if not table_end then return nil, "no terminator" end
--- @type AbbrevDecl[]
local decls = {}
--- @type integer
local pos = table_start
while pos < table_end do
--- @type integer|nil, integer
local code, code_end = M.read_uleb128_at(table_bytes, pos)
if not code then return nil, "truncated code" end
pos = code_end
--- @type integer|nil, integer
local tag, tag_end = M.read_uleb128_at(table_bytes, pos)
if not tag then return nil, "truncated tag" end
pos = tag_end
--- @type integer
local has_children = table_bytes:byte(pos + 1)
pos = pos + 1
--- @type AbbrevAttr[]
local attrs = {}
while true do
--- @type integer|nil, integer
local attr, attr_end = M.read_uleb128_at(table_bytes, pos)
if not attr then return nil, "truncated attr" end
pos = attr_end
--- @type integer|nil, integer
local form, form_end = M.read_uleb128_at(table_bytes, pos)
if not form then return nil, "truncated form" end
pos = form_end
if attr == 0 and form == 0 then break end
attrs[#attrs + 1] = { name = attr, form = form }
if form == DW_FORM_implicit_const then
--- @type integer|nil, integer
local _c, ce = M.read_sleb128_at(table_bytes, pos)
if not _c then return nil, "truncated const" end
pos = ce
@@ -371,52 +531,121 @@ end
-- For DW_FORM_strp we return the inline string resolved from `str_buf`.
-- For DW_FORM_ref4 we return the absolute CU-relative offset.
-- The caller decides whether to interpret that as a section offset.
--- @type table<integer, fun(buf: string, str_buf: string, pos: integer): (string|integer|nil, integer)>
local FORM_READERS = {
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.addr] = function(buf, _, pos)
return M.read_u32_le(buf, pos), pos + 4
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return string
--- @return integer
[M.DW_FORM.string] = function(buf, _, pos)
--- @type string
local s = read_c_string_at(buf, pos)
return s, pos + #s + 1
end,
--- @param buf string
--- @param str_buf string
--- @param pos integer
--- @return string
--- @return integer
[M.DW_FORM.strp] = function(buf, str_buf, pos)
-- DW_FORM_strp: 4-byte offset into .debug_str.
--- @type integer
local strp_off = M.read_u32_le(buf, pos)
return read_c_string_at(str_buf, strp_off), pos + 4
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer|nil
--- @return integer
[M.DW_FORM.udata] = function(buf, _, pos)
return M.read_uleb128_at(buf, pos)
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.data1] = function(buf, _, pos)
return buf:byte(pos + 1), pos + 1
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.data2] = function(buf, _, pos)
return M.read_u16_le(buf, pos), pos + 2
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.data4] = function(buf, _, pos)
return M.read_u32_le(buf, pos), pos + 4
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.ref4] = function(buf, _, pos)
return M.read_u32_le(buf, pos), pos + 4
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.sec_offset] = function(buf, _, pos)
-- DW_FORM_sec_offset: 4-byte offset (size depends on DWARF version;
-- on DWARF5 32-bit it's always 4 bytes).
return M.read_u32_le(buf, pos), pos + 4
end,
--- @param _ string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.flag_present] = function(_, _, pos)
return 1, pos
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return nil
--- @return integer
[M.DW_FORM.exprloc] = function(buf, _, pos)
-- DW_FORM_exprloc: ULEB byte count + that many bytes of DW_OP_*.
--- @type integer|nil, integer
local len, ne = M.read_uleb128_at(buf, pos)
if not len then return nil, pos end
return nil, ne + len
end,
--- @param _ string
--- @param _ string
--- @param pos integer
--- @return nil
--- @return integer
[DW_FORM_implicit_const] = function(_, _, pos)
-- The constant is declared in the abbrev; no value bytes in the DIE.
return nil, pos
end,
--- @param buf string
--- @param _ string
--- @param pos integer
--- @return integer
--- @return integer
[M.DW_FORM.ref_sig8] = function(buf, _, pos)
-- DW_FORM_ref_sig8 (DWARF5 §7.4.2): An 8-byte value identifying a type by signature.
-- The low 4 bytes (LE) are the type signature (content hash);
@@ -425,11 +654,19 @@ local FORM_READERS = {
-- then the high 4 to resolve the specific type within it.
-- Return the low 4 as the primary value to preserve the (value, next_pos) shape;
-- the high 4 is exposed via M.read_ref_sig8 (which returns both halves).
--- @type integer, integer, integer
local _, _, next_pos = M.read_ref_sig8(buf, pos)
return M.read_u32_le(buf, pos), next_pos
end,
}
--- @param buf string
--- @param str_buf string
--- @param pos integer
--- @param form integer
--- @return string|integer|nil
--- @return integer
local function read_form_value(buf, str_buf, pos, form)
--- @type (fun(buf: string, str_buf: string, pos: integer): (string|integer|nil, integer))|nil
local r = FORM_READERS[form]
if not r then
return nil, pos
@@ -464,15 +701,20 @@ function M.read_ref_sig8(buf, pos) return M.read_u32_le(buf, pos), M.read_u32_le
--- @param target_sig_hi integer -- high 4 bytes (LE) of the desired signature
--- @return integer|nil, integer|nil -- unit offset, type_offset within the unit
function M.find_type_unit_by_signature(info, target_sig_lo, target_sig_hi)
--- @type integer
local pos = 0
--- @type integer
local section_len = #info
while pos + 4 < section_len do
--- @type integer
local unit_length = M.read_u32_le(info, pos)
if unit_length == 0xFFFFFFFF then
return nil, nil -- DWARF64 not supported
end
-- unit_length is the body size, NOT including the 4-byte unit_length field itself.
--- @type integer
local body_start = pos + 4
--- @type integer
local body_end = body_start + unit_length
if body_end > section_len then
return nil, nil -- malformed
@@ -495,10 +737,13 @@ function M.find_type_unit_by_signature(info, target_sig_lo, target_sig_hi)
-- byte 4-7: debug_abbrev_offset (4)
-- byte 8-15: type_signature (8)
-- byte 16-19: type_offset (4)
--- @type integer
local unit_type = info:byte(body_start + 2 + 1) -- 0-based +2 = unit_type in 1-indexed
if unit_type == 0x02 then -- DW_UT_type
--- @type integer, integer, integer
local sig_lo, sig_hi, _ = M.read_ref_sig8(info, body_start + 8) -- 0-based +8 = type_signature in 1-indexed
if sig_lo == target_sig_lo and sig_hi == target_sig_hi then
--- @type integer
local type_offset = M.read_u32_le(info, body_start + 16) -- 0-based +16 = type_offset in 1-indexed
return pos, type_offset
end
@@ -551,11 +796,15 @@ end
function M.read_elf_sections(elf_path, section_names)
-- Initialize result with all requested names set to "" so callers can do `sections[X]
-- or ""` for missing sections without nil-checks.
--- @type table<string, string>
local result = {}
--- @type integer, string
for _, name in ipairs(section_names) do result[name] = "" end
-- O(1) lookup set.
--- @type table<string, boolean> -- bag: requested section name -> true
local wanted = {}
--- @type integer, string
for _, name in ipairs(section_names) do wanted[name] = true end
-- Existence check (lfs.attributes avoids an io.open-vs-fail race).
@@ -564,45 +813,63 @@ function M.read_elf_sections(elf_path, section_names)
return result
end
--- @type file*|nil
local f = io.open(elf_path, "rb")
if not f then
io.stderr:write(string.format("[elf_dwarf.read_elf_sections] io.open failed: %s\n", elf_path))
return result
end
--- @type integer
local file_size
do
f:seek("end", 0)
file_size = f:seek("cur", 0)
end
--- @type Elf32Adapter
local adapter = {
--- @param offset integer
--- @return integer|nil
read_u8_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b = f:read(1)
if not b then return nil end
return b:byte()
end,
--- @param offset integer
--- @return integer|nil
read_u16_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b1 = f:read(1)
--- @type string|nil
local b2 = f:read(1)
if not b1 or not b2 then return nil end
return b1:byte() + b2:byte() * 0x100
end,
--- @param offset integer
--- @return integer|nil
read_u32_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b1 = f:read(1)
--- @type string|nil
local b2 = f:read(1)
--- @type string|nil
local b3 = f:read(1)
--- @type string|nil
local b4 = f:read(1)
if not b1 or not b2 or not b3 or not b4 then return nil end
return b1:byte() + b2:byte() * 0x100
+ b3:byte() * 0x10000 + b4:byte() * 0x1000000
end,
--- @return integer
read_size = function() return file_size end,
}
-- Delegate the header parse + section walk to E.*.
--- @type Elf32Header|nil, string|nil
local hdr, hdr_err = E.parse_elf32_headers(adapter)
if not hdr then
io.stderr:write(string.format("[elf_dwarf.read_elf_sections] header parse failed: %s\n", tostring(hdr_err)))
@@ -610,6 +877,7 @@ function M.read_elf_sections(elf_path, section_names)
return result
end
--- @type Elf32Section[]|nil, string|nil
local sections, walk_err = E.walk_sections(adapter, hdr)
if not sections then
io.stderr:write(string.format("[elf_dwarf.read_elf_sections] section walk failed: %s\n", tostring(walk_err)))
@@ -618,8 +886,10 @@ function M.read_elf_sections(elf_path, section_names)
end
-- Resolve the requested sections.
--- @type integer, Elf32Section
for _, s in ipairs(sections) do
if wanted[s.name] then
--- @type string|nil
local bytes = E.read_section_bytes(adapter, s)
if bytes then result[s.name] = bytes end
end
@@ -640,50 +910,69 @@ end
--- - `st_size > 0` filter excludes undefined/imported symbols.
---
--- @param elf_path Path
--- @return table<string, {integer, integer}>
--- @return table<string, NmAddr>
function M.read_nm(elf_path)
--- @type table<string, NmAddr>
local addrs = {}
-- Existence check first; an empty or missing ELF returns an empty map.
if lfs.attributes(elf_path, "mode") ~= "file" then return addrs end
--- @type file*|nil
local f = io.open(elf_path, "rb")
if not f then return addrs end
-- Build the file adapter for E.*.
--- @type integer
local file_size
do
f:seek("end", 0)
file_size = f:seek("cur", 0)
end
--- @type Elf32Adapter
local adapter = {
--- @param offset integer
--- @return integer|nil
read_u8_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b = f:read(1)
if not b then return nil end
return b:byte()
end,
--- @param offset integer
--- @return integer|nil
read_u16_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b1 = f:read(1)
--- @type string|nil
local b2 = f:read(1)
if not b1 or not b2 then return nil end
return b1:byte() + b2:byte() * 0x100
end,
--- @param offset integer
--- @return integer|nil
read_u32_at = function(offset)
f:seek("set", offset)
--- @type string|nil
local b1 = f:read(1)
--- @type string|nil
local b2 = f:read(1)
--- @type string|nil
local b3 = f:read(1)
--- @type string|nil
local b4 = f:read(1)
if not b1 or not b2 or not b3 or not b4 then return nil end
return b1:byte() + b2:byte() * 0x100
+ b3:byte() * 0x10000 + b4:byte() * 0x1000000
end,
--- @return integer
read_size = function() return file_size end,
}
-- Delegate the header + section walk to E.*.
--- @type Elf32Header|nil, string|nil
local hdr, hdr_err = E.parse_elf32_headers(adapter)
if not hdr then
io.stderr:write(string.format("[elf_dwarf.read_nm] header parse failed: %s\n", tostring(hdr_err)))
@@ -691,6 +980,7 @@ function M.read_nm(elf_path)
return addrs
end
--- @type Elf32Section[]|nil, string|nil
local sections, walk_err = E.walk_sections(adapter, hdr)
if not sections then
io.stderr:write(string.format("[elf_dwarf.read_nm] section walk failed: %s\n", tostring(walk_err)))
@@ -700,6 +990,7 @@ function M.read_nm(elf_path)
-- E.collect_symbols returns every defined symbol (no binding filter).
-- The metaprogram then applies its STB_LOCAL / STB_GLOBAL + size>0 filter, matching `nm`'s default (external symbols only).
--- @type table<string, Elf32Sym>|nil, string|nil
local symbols, sym_err = E.collect_symbols(adapter, sections)
if not symbols then
io.stderr:write(string.format("[elf_dwarf.read_nm] symbol collection failed: %s\n", tostring(sym_err)))
@@ -709,9 +1000,11 @@ function M.read_nm(elf_path)
f:close()
--- @type string, Elf32Sym
for name, entry in pairs(symbols) do
-- High nibble of st_info = binding (STB_LOCAL=0, STB_GLOBAL=1, STB_WEAK=2).
-- math.floor(/16) is portable across LuaJIT 2.0/2.1 and plain Lua 5.x.
--- @type integer
local binding = math.floor(entry.info / 16)
if (binding == 0 or binding == 1) and entry.size > 0 then
addrs[name] = { entry.value, entry.size }
@@ -744,13 +1037,16 @@ end
-- Spec: DWARF5 §7.6 "Variable-Length Data" / Appendix C.
-- Top bit of each LEB128 byte. Set if more bytes follow in the encoding.
--- @type integer
local LEB_CONT_BIT = 0x80
-- Low 7 bits of each LEB128 byte. The actual data payload.
--- @type integer
local LEB_DATA_MASK = 0x7F
-- Bit 6 of the 7-bit data (i.e. 0x40). For SLEB128: the sign-bit position used by the decoder for sign extension.
-- Encoders MUST stop when the next byte would be redundant AND the sign bit in the last byte matches the value's sign.
--- @type integer
local SLEB_SIGN_BIT = 0x40
--- ULEB128 (Unsigned Little-Endian Base 128) encoder. Returns the byte string for the non-negative integer `n`.
@@ -768,8 +1064,10 @@ function M.uleb128(n)
error("uleb128 requires non-negative number")
end
assert(n >= 0, "uleb128 requires non-negative input")
--- @type string[]
local bytes = {}
repeat
--- @type integer
local b = n % (LEB_DATA_MASK + 1) -- extract low 7 bits
n = (n - b) / (LEB_DATA_MASK + 1) -- shift right by 7 bits
if n > 0 then b = b + LEB_CONT_BIT end -- set continuation bit if more bytes follow
@@ -789,9 +1087,12 @@ end
--- @param n integer -- any integer (negative allowed)
--- @return string
function M.sleb128(n)
--- @type string[]
local bytes = {}
--- @type boolean
local more = true
while more do
--- @type integer
local b = n % (LEB_DATA_MASK + 1) -- extract low 7 bits
n = (n - b) / (LEB_DATA_MASK + 1) -- arithmetic shift right by 7
-- Termination: remaining value bits fit in the sign bit of the last byte.
@@ -811,6 +1112,7 @@ end
function M.uleb128_size(n)
assert(n >= 0, "uleb128_size requires non-negative input")
if n == 0 then return 1 end
--- @type integer
local bytes = 1
while n >= 0x80 do
n = (n - (n % (LEB_DATA_MASK + 1))) / (LEB_DATA_MASK + 1) -- arithmetic shift right by 7
@@ -827,10 +1129,14 @@ end
--- @param n integer -- any integer (negative allowed)
--- @return integer
function M.sleb128_size(n)
--- @type boolean
local more = true
--- @type integer
local bytes = 0
--- @type integer
local v = n
while more do
--- @type integer
local b = v % (LEB_DATA_MASK + 1) -- extract low 7 bits
v = (v - b) / (LEB_DATA_MASK + 1) -- arithmetic shift right by 7
if v == 0 and b < SLEB_SIGN_BIT then more = false end -- positive terminator
@@ -869,21 +1175,26 @@ end
--- downstream `resolve_provenance_file_index(path)` consumers consult the map directly.
---
--- @param elf_path string -- absolute path to the post-link ELF (typically the gcc-emitted `.elf` BEFORE dwarf_injector's splice; both shapes work since the splice preserves `.debug_line`)
--- @return table|nil, table|nil, table|nil
--- basename_to_index: { [basename] = 1-based-per-unit-file-index, ... }
--- basenames: { [1-based-per-unit-file-index] = basename, ... }
--- paths: { [1-based-per-unit-file-index] = full path (mixed slashes), ... }
--- @return table<string, integer>|nil
--- @return table<integer, string>|nil
--- @return table<integer, string>|nil
function M.read_line_unit_file_table(elf_path)
--- @type table<string, string>
local sections = M.read_elf_sections(elf_path, { ".debug_line", ".debug_line_str" })
--- @type string
local line = sections[".debug_line"]
--- @type string
local lstr = sections[".debug_line_str"] or ""
if not line or line == "" then
io.stderr:write("[elf_dwarf.read_line_unit_file_table] no .debug_line section in: " .. tostring(elf_path) .. "\n")
return nil
end
--- @type table<integer, string> -- bag: 1-based file index -> basename
local basenames = {}
--- @type table<string, integer> -- bag: basename -> 1-based file index
local basename_to_index = {}
--- @type table<integer, string> -- bag: 1-based file index -> full path
local paths = {}
--- Read one form-code's bytes from `buf` at position `p` according to `form`.
@@ -891,15 +1202,25 @@ function M.read_line_unit_file_table(elf_path)
--- * the resolved string (DW_FORM_line_strp / DW_FORM_string)
--- * the ULEB128 number (DW_FORM_udata)
--- * nil + skip-bytes (DW_FORM_data16; we don't surface the MD5)
--- @param buf string
--- @param lstr_buf string
--- @param p integer
--- @param form integer
--- @return string|integer|nil
--- @return integer
local function read_form(buf, lstr_buf, p, form)
if form == M.DWARF5_DEBUG_LINE.form_line_strp then
--- @type integer
local strp = M.read_u32_le(buf, p)
--- @type integer
local end_pos = lstr_buf:find("\0", strp + 1, true) or (#lstr_buf + 1)
return lstr_buf:sub(strp + 1, end_pos - 1), p + M.DWARF5_DEBUG_LINE.form_strp_bytes
elseif form == M.DWARF5_DEBUG_LINE.form_string then
--- @type integer
local nul = buf:find("\0", p + 1, true) or (#buf + 1)
return buf:sub(p + 1, nul - 1), nul
elseif form == M.DWARF5_DEBUG_LINE.form_udata then
--- @type integer|nil, integer
local v, after = M.read_uleb128_at(buf, p)
return v, after
elseif form == M.DWARF5_DEBUG_LINE.form_data16 then
@@ -916,34 +1237,51 @@ function M.read_line_unit_file_table(elf_path)
--- Parse one DWARF-version-3-style unit (DWARF3/4 line program; gcc default in the PS1 toolchain still emits DWARF3 for line programs in `-g` mode).
--- Layout: null-terminated directory list, then path(null) + dir_idx(ULEB) + time(ULEB) + size(ULEB) file entries terminated by an empty null.
--- `content_start` = zero-based wire offset of the first byte of program-header content (after version + header_length fields).
--- @return unit_basenames { [idx_in_unit_1_based] = basename }
--- @return unit_paths { [idx_in_unit_1_based] = full path }
--- @param buf string
--- @param content_start integer
--- @param body_end integer
--- @return table<integer, string>
--- @return table<integer, string>
local function parse_dwarf3_unit(buf, content_start, body_end)
--- @type integer
local up = content_start
-- 5 fixed bytes: min_insn, default_is, line_base (signed), line_range, opcode_base
up = up + 5
--- @type integer
local opcode_base = buf:byte(content_start + 5)
up = up + (opcode_base - 1) -- std_opcode_lengths
--- @type string[]
local dirs = {}
while up < body_end do
--- @type integer
local nul = buf:find("\0", up + 1, true) or (body_end + 1)
if nul > body_end then break end
--- @type integer
local len = nul - up - 1
if len == 0 then up = nul break end
dirs[#dirs + 1] = buf:sub(up + 1, nul - 1)
up = nul
end
--- @type table<integer, string> -- bag: 1-based unit file index -> basename
local unit_basenames = {}
--- @type table<integer, string> -- bag: 1-based unit file index -> full path
local unit_paths = {}
while up < body_end do
--- @type integer
local nul = buf:find("\0", up + 1, true) or (body_end + 1)
if nul > body_end or nul == up + 1 then up = nul break end
--- @type string
local path = buf:sub(up + 1, nul - 1)
up = nul
--- @type integer|nil, integer
local didx, up_next = M.read_uleb128_at(buf, up); up = up_next
--- @type integer|nil, integer
local _time, up_next2 = M.read_uleb128_at(buf, up); up = up_next2
--- @type integer|nil, integer
local _size, up_next3 = M.read_uleb128_at(buf, up); up = up_next3
--- @type integer
local idx = #unit_basenames + 1
--- @type string
local bs = path:match("[^/\\]+$") or path
unit_paths[idx] = path
unit_basenames[idx] = bs
@@ -957,25 +1295,42 @@ function M.read_line_unit_file_table(elf_path)
--- Parse one DWARF-version-5-style unit (DWARF5 line program; used by modern gcc with `-gdwarf-5`).
--- `content_start` is the first byte of program-header content (after the 8 fixed bytes version+addr_size+seg_size+header_length).
--- @return same shape as parse_dwarf3_unit
--- @param buf string
--- @param lstr_buf string
--- @param content_start integer
--- @param body_end integer
--- @return table<integer, string>
--- @return table<integer, string>
local function parse_dwarf5_unit(buf, lstr_buf, content_start, body_end)
--- @type integer
local up = content_start
-- 6 fixed bytes: min_insn, max_ops_per_insn, default_is, line_base, line_range, opcode_base
up = up + 6
--- @type integer
local opcode_base = buf:byte(content_start + 6)
up = up + (opcode_base - 1) -- std_opcode_lengths
-- directories
--- @type integer|nil, integer
local dir_format_count, after = M.read_uleb128_at(buf, up); up = after
--- @type integer[]
local dir_formats = {}
--- @type integer
for i = 1, dir_format_count do
--- @type integer|nil, integer
local f, a2 = M.read_uleb128_at(buf, up); up = a2
dir_formats[i] = f
end
--- @type integer|nil, integer
local dir_count, a3 = M.read_uleb128_at(buf, up); up = a3
--- @type string[]
local dirs = {}
--- @type integer
for i = 1, dir_count do
--- @type string
local combined = ""
--- @type integer
for j = 1, dir_format_count do
--- @type string|integer|nil, integer
local v, a4 = read_form(buf, lstr_buf, up, dir_formats[j])
up = a4
if j == 1 and type(v) == "string" then combined = v end
@@ -983,25 +1338,39 @@ function M.read_line_unit_file_table(elf_path)
dirs[i] = combined
end
-- file names
--- @type integer|nil, integer
local file_format_count, after2 = M.read_uleb128_at(buf, up); up = after2
--- @type integer[]
local file_formats = {}
--- @type integer
for i = 1, file_format_count do
--- @type integer|nil, integer
local f, a2 = M.read_uleb128_at(buf, up); up = a2
file_formats[i] = f
end
--- @type integer|nil, integer
local file_count, a3 = M.read_uleb128_at(buf, up); up = a3
--- @type table<integer, string> -- bag: 1-based unit file index -> basename
local unit_basenames = {}
--- @type table<integer, string> -- bag: 1-based unit file index -> full path
local unit_paths = {}
--- @type integer
for i = 1, file_count do
--- @type string
local combined = ""
--- @type integer
local didx = 0
--- @type integer
for j = 1, file_format_count do
--- @type string|integer|nil, integer
local v, a4 = read_form(buf, lstr_buf, up, file_formats[j])
up = a4
if j == 1 and type(v) == "string" then combined = v end
if j == 2 and type(v) == "number" then didx = v end
end
--- @type integer
local idx = #unit_basenames + 1
--- @type string
local bs = combined:match("[^/\\]+$") or combined
unit_paths[idx] = combined
unit_basenames[idx] = bs
@@ -1013,26 +1382,36 @@ function M.read_line_unit_file_table(elf_path)
end
--- Walk every line-program unit in the section.
--- @type integer
local p = 0
--- @type integer
local section_end = #line
while p + 4 <= section_end do
--- @type integer
local unit_length = M.read_u32_le(line, p)
if unit_length == 0xFFFFFFFF then
io.stderr:write("[elf_dwarf.read_line_unit_file_table] 64-bit DWARF (initial-length 0xFFFFFFFF); not supported\n")
return nil
end
--- @type integer
local body_start = p + 4
--- @type integer
local body_end = p + 4 + unit_length
if body_end > section_end then break end
--- @type integer
local version = M.read_u16_le(line, body_start)
--- @type table<integer, string>|nil, table<integer, string>|nil
local unit_basenames, unit_paths
if version >= 5 then
-- DWARF5 header: version(2) + addr_size(1) + seg_size(1) + header_length(4) + content
--- @type integer
local header_length_offset = body_start + 6 -- past version(2) + addr_size(1) + seg_size(1) - wait that's wrong; past hdr len is at +6
--- @type integer
local content_start = body_start + 8 -- past version(2) + addr_size(1) + seg_size(1) + header_length(4)
unit_basenames, unit_paths = parse_dwarf5_unit(line, lstr, content_start, body_end)
elseif version >= 2 then
-- DWARF2/3/4 header: version(2) + header_length(4) + content
--- @type integer
local content_start = body_start + 6 -- past version(2) + header_length(4)
unit_basenames, unit_paths = parse_dwarf3_unit(line, content_start, body_end)
else
@@ -1047,6 +1426,7 @@ function M.read_line_unit_file_table(elf_path)
-- For DWARF5 (crt0.s + C unit), each carries its own per-unit file-table map;
-- the atom-side DW_LNS_set_file(N) refers to the C unit's indices, NOT crt0.s's.
-- Since the C unit is the one with full include_directories + 12 entries, we can use it directly.
--- @type integer, string
for idx, bs in pairs(unit_basenames) do
basenames[idx] = bs
paths[idx] = unit_paths[idx]
scripts/passes/annotation.lua
+112 -56
View File
@@ -10,7 +10,9 @@
-- Bootstrap follows the entry scripts; `scripts/duffle_paths.lua` sets package.path and package.cpath. See `ps1_meta.lua` for the rationale.
-- `debug.getinfo(1, "S").source` locates this file for standalone and orchestrated runs, then `duffle_paths.lua` returns the loaded `duffle` module.
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- The annotation pass reads the source-derived registries from scan_source:
@@ -21,28 +23,8 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- Type declarations
-- ════════════════════════════════════════════════════════════════════════════
--- @class SourceFile
--- @field path string -- Absolute path to the source file
--- @field text string -- Full source text
--- @field dir string -- Directory containing the source
--- @field basename string -- Filename without extension
--- @field scan table -- Pre-scanned SourceScan payload (from duffle.scan_source)
--- @class PassCtx
--- @field sources SourceFile[]
--- @field metadata_path string
--- @field shared table
--- @field shared.word_counts table<string, integer>
--- @field out_root string
--- @field project_root string
--- @field upstream table<string, table>
--- @field flags table
--- @field verbose boolean
--- @class PassResult
--- @field outputs table[]
--- @field errors table[]
--- @field warnings table[]
-- SourceFile, PassCtx, PassResult, PassShared, Corpus: see ps1_meta.lua
-- SourceScan, AtomEntry, BindsEntry, RegTypeDefault, AtomViewEntry: see scan_source.lua
--- @class AtomAnnotation
--- @field atom_name string -- Atom name (scan.atom_infos row)
@@ -52,43 +34,51 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--- @field writes string[] -- R_* names (write targets)
--- @field errors string[]|nil -- Parse-time errors from scan_source (atom_info body malformed)
--- @class DebugSkipMarker -- Sub-shape of scan_source.lua's @class DebugSkipMarker
--- @field marker_kind string -- Exact marker ident read from source. Only "atom_dbg_skip" (bare) is positive.
--- @field marker_line integer
--- @field args string|nil -- Trimmed text inside the parens (nil when has_parens is false)
--- @field has_parens boolean
--- @field is_bare boolean -- true iff marker_kind == "atom_dbg_skip" AND has_parens == false (the only positive form)
--- @field pending boolean -- true while awaiting the following declaration
--- @field superseded_by_marker_line integer|nil -- Set on a marker that was bumped out of the pending slot
--- @field target_kind string|nil -- "atom" | "comp_bare" | "comp_proc" | "unrelated" once observed
--- @class Finding
--- @field line integer -- Source line (or 0 for pass-level)
--- @field msg string -- Finding message
--- @class RegTypeOccurrence
--- @field reg string
--- @field type_name string
--- @field source_line integer
--- @class Findings
--- @field errors Finding[]
--- @field warnings Finding[]
--- @field info Finding[]
--- @field errors PassFinding[]
--- @field warnings PassFinding[]
--- @field info PassFinding[]
--- @class PipeCtx
--- @field atom_index table<string, AtomEntry> -- raw_name or name -> scan.atoms row (kind atom/atom_proc)
--- @field binds_index table<string, BindsStruct> -- Name -> BindsStruct
--- @field annot_counts table<string, integer> -- Name -> annotation count (for unique_annotation check)
--- @field types table<string, RegTypeDefault> -- From scan_source
--- @field atom_views table<string, AtomViewEntry> -- From scan_source
--- @field seen_defaults table<string, integer> -- Duplicate atom_dbg_reg_default detection
--- @field seen_field table<string, integer> -- Binds_* -> count of fields (set/checked by check_binds_no_duplicate_fields)
--- @field _scan SourceScan -- Full scan payload (typed-view sub-calls live here)
--- @field atom_index table<string, AtomEntry> -- raw_name or name -> scan.atoms row (kind atom/atom_proc)
--- @field binds_index table<string, BindsEntry>
--- @field annot_counts table<string, integer> -- bag: atom name -> annotation count
--- @field types table<string, RegTypeDefault>
--- @field atom_views table<string, AtomViewEntry>
--- @field seen_defaults table<string, integer> -- bag: register ident -> occurrence count
--- @field seen_field table<string, integer> -- bag: leftover field-count slot
--- @field _scan SourceScan
--- @field word_counts WordCounts|nil
--- @field register_alias_registry table<string, AliasEntry>|nil
--- @field type_name_registry table<string, TypeNameEntry>|nil
--- @field type_occurrences RegTypeOccurrence[]|nil
--- @field atom_infos_list AtomInfoEntry[]|nil
--- @field binds_list BindsEntry[]|nil
--- @class AnnotatedResult
--- @field atoms AtomEntry[]
--- @field annots AtomAnnotation[]
--- @field macros MacroEntry[]
--- @field binds BindsEntry[]
--- @field errors Finding[]
--- @field warnings Finding[]
--- @field info Finding[]
--- @field errors PassFinding[]
--- @field warnings PassFinding[]
--- @field info PassFinding[]
--- @field source string|nil
--- @class CheckRule
--- @field per_annot (fun(item: AtomAnnotation, pipe_ctx: PipeCtx, findings: Findings): nil)|nil
--- @class SourceScan
--- @field type_occurrences RegTypeOccurrence[]|nil
--- @class AnnotationPass
--- @field validate fun(ctx: PassCtx, src: SourceFile, corpus_pipe_ctx: PipeCtx|nil): AnnotatedResult
--- @field run fun(ctx: PassCtx): PassResult
-- ════════════════════════════════════════════════════════════════════════════
-- Per-check functions (the CHECK_RULES table's payload)
@@ -100,6 +90,7 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--- @param info AtomAnnotation
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_atom_decl_exists(info, pipe_ctx, findings)
if not pipe_ctx.atom_index[info.atom_name] then
findings.errors[#findings.errors + 1] = {
@@ -111,9 +102,12 @@ end
--- Check: Every atom may have AT MOST ONE annotation.
--- Post-loop: Needs full-corpus `annot_counts` from pipe_ctx.
--- @param _item AtomAnnotation|nil
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_unique_annotation(_item, pipe_ctx, findings)
--- @type string, integer
for name, n in pairs(pipe_ctx.annot_counts) do
if n > 1 then
findings.errors[#findings.errors + 1] = {
@@ -129,6 +123,7 @@ end
--- @param info AtomAnnotation
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_binds_struct_exists(info, pipe_ctx, findings)
if not info.binds then return end
if pipe_ctx.binds_index[info.binds] then return end
@@ -142,11 +137,14 @@ end
--- Check: TAPE_WORDS(mac_X, N) ↔ WORD_COUNT(mac_X, N) drift.
--- Three outcomes: missing (error), mismatch (error), match (info).
--- @param m MacroEntry
--- @param wc table<string, integer> -- Shared word-count table (from ctx.shared.word_counts)
--- @param m MacroEntry
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_macro_word_drift(m, pipe_ctx, findings)
--- @type WordCounts
local wc = (pipe_ctx and pipe_ctx.word_counts) or {}
--- @type integer|nil
local declared = wc[m.name]
if not declared then
findings.errors[#findings.errors + 1] = {
@@ -173,9 +171,12 @@ end
--- @param _src SourceFile -- unused (kept for the per_source shape)
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_semantic_reg_defaults(_src, pipe_ctx, findings)
-- Detect duplicate defaults using the ordered occurrence list (the out.types hash only retains the last declaration).
--- @type table<string, integer> -- bag: register ident -> first source line
local seen_first_line = {}
--- @type integer, RegTypeOccurrence
for _, occ in ipairs(pipe_ctx.type_occurrences or {}) do
if seen_first_line[occ.reg] == nil then
seen_first_line[occ.reg] = occ.source_line
@@ -188,8 +189,11 @@ local function check_semantic_reg_defaults(_src, pipe_ctx, findings)
}
end
end
--- @type table<string, AliasEntry>
local reg_registry = pipe_ctx.register_alias_registry or {}
--- @type table<string, TypeNameEntry>
local type_registry = pipe_ctx.type_name_registry or {}
--- @type string, RegTypeDefault
for reg, def in pairs(pipe_ctx.types or {}) do
if not reg_registry[reg] then
findings.errors[#findings.errors + 1] = {
@@ -223,11 +227,16 @@ end
--- @param _src SourceFile
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_atom_reg_types(_src, pipe_ctx, findings)
--- @type table<string, AliasEntry>
local reg_registry = pipe_ctx.register_alias_registry or {}
--- @type table<string, TypeNameEntry>
local type_registry = pipe_ctx.type_name_registry or {}
--- @type integer, AtomInfoEntry
for _, ai in ipairs(pipe_ctx.atom_infos_list or {}) do
if ai.reg_type_overrides then
--- @type string, RegTypeOverride
for reg, ov in pairs(ai.reg_type_overrides) do
if not reg_registry[reg] then
findings.errors[#findings.errors + 1] = {
@@ -254,11 +263,14 @@ end
--- @param _src SourceFile
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_atom_view_layout(_src, pipe_ctx, findings)
--- @type string, AtomViewEntry
for atom_name, view in pairs(pipe_ctx.atom_views or {}) do
if not view.binds_name then
-- The atom had atom_reg_types but no atom_view; no layout check needed.
else
--- @type BindsEntry|nil
local bs = pipe_ctx.binds_index[view.binds_name]
if not bs then
findings.errors[#findings.errors + 1] = {
@@ -280,15 +292,20 @@ local function check_atom_view_layout(_src, pipe_ctx, findings)
end
--- Check: Binds_* structs require unique field names because atom_view uses those names for typed-field lookup in gdb.
--- @param _src SourceFile
--- @param _src SourceFile
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_binds_no_duplicate_fields(_src, pipe_ctx, findings)
--- @type integer, BindsEntry
for _, bs in ipairs(pipe_ctx.binds_list or {}) do
--- @type table<string, integer> -- bag: field name -> occurrence count
local seen = {}
--- @type integer, TypeField
for _, f in ipairs(bs.fields or {}) do
seen[f.name] = (seen[f.name] or 0) + 1
end
--- @type string, integer
for name, count in pairs(seen) do
if count > 1 then
findings.errors[#findings.errors + 1] = {
@@ -312,11 +329,14 @@ end
--- 5. pending + no target_kind -> dangling (no following declaration)
--- 6. unsupported target_kind -> marker precedes an unrelated declaration
--- Valid markers stamp `debug_skip` on whole-atom, bare-component, and proc-component declaration records in scan_source.lua.
--- @param marker DebugSkipMarker
--- @param marker DebugSkipMarker
--- @param _pipe_ctx PipeCtx -- Unused; kept for consistency with per_annot
--- @param findings Findings
--- @param findings Findings
--- @return nil
local function check_skip_marker(marker, _pipe_ctx, findings)
--- @type string
local kind = marker.marker_kind
--- @type integer
local line = marker.marker_line
-- Left `scan.debug_skip_markers` with production records for `atom_dbg_skip` only; other identifiers take the walker's unrelated branch.
@@ -372,12 +392,16 @@ end
--- @param _src SourceFile
--- @param pipe_ctx PipeCtx
--- @param findings Findings
--- @return nil
local function check_wave_context_migration(_src, pipe_ctx, findings)
if not (pipe_ctx.types and next(pipe_ctx.types)) then return end
if not (pipe_ctx.atom_infos_list) then return end
--- @type table<string, AliasEntry>
local reg_registry = pipe_ctx.register_alias_registry or {}
--- @type integer, AtomInfoEntry
for _, ai in ipairs(pipe_ctx.atom_infos_list) do
if ai.reg_type_overrides then
--- @type string, RegTypeOverride
for reg, _ in pairs(ai.reg_type_overrides) do
if not reg_registry[reg] then
findings.warnings[#findings.warnings + 1] = {
@@ -405,6 +429,7 @@ end
--
-- Adding a new check = 1 row here + 1 function above. The `validate()` dispatch loop never needs editing.
--- @type CheckRule[]
local CHECK_RULES = {
{ name = "atom_decl_exists", per_annot = check_atom_decl_exists },
{ name = "binds_struct_exists", per_annot = check_binds_struct_exists },
@@ -428,8 +453,11 @@ local CHECK_RULES = {
--- @param ctx PassCtx
--- @return PipeCtx
local function build_corpus_pipe_ctx(ctx)
--- @type PipeCtx
local view = duffle.corpus_view(ctx)
--- @type table<string, integer> -- bag: atom name -> annotation count
local annot_counts = {}
--- @type integer, AtomInfoEntry
for _, info in ipairs(view.atom_infos) do
if info and info.atom_name then
annot_counts[info.atom_name] = (annot_counts[info.atom_name] or 0) + 1
@@ -448,12 +476,16 @@ end
--- @return AnnotatedResult
local function validate(ctx, src, corpus_pipe_ctx)
corpus_pipe_ctx = corpus_pipe_ctx or build_corpus_pipe_ctx(ctx)
--- @type SourceScan
local scan = src.scan
-- Build a per-source pipe_ctx: shared lookups come from `corpus_pipe_ctx`, while declarations, bodies, types, views, defaults, and occurrences come from `src.scan`.
--- @type table<string, integer> -- bag: register ident -> occurrence count
local seen_defaults = {}; for reg, _ in pairs (scan.types or {}) do seen_defaults[reg] = (seen_defaults[reg] or 0) + 1 end
--- @type AtomInfoEntry[]
local atom_infos_list = {}; for _, ai in ipairs(scan.atom_infos or {}) do atom_infos_list[#atom_infos_list + 1] = ai end
--- @type PipeCtx
local pipe_ctx = {
atom_index = {},
binds_index = {},
@@ -468,22 +500,28 @@ local function validate(ctx, src, corpus_pipe_ctx)
register_alias_registry = corpus_pipe_ctx.register_alias_registry,
type_name_registry = corpus_pipe_ctx.type_name_registry,
}
--- @type AtomEntry[]
local atoms = {}
--- @type integer, AtomEntry
for _, a in ipairs(scan.atoms) do
if a.kind == "atom" or a.kind == "atom_proc" then
atoms[#atoms + 1] = a
pipe_ctx.atom_index[a.raw_name or a.name] = a
end
end
--- @type integer, BindsEntry
for _, b in ipairs(scan.binds) do pipe_ctx.binds_index[b.name] = b end
-- Findings live in a single struct with three lists (errors / warnings / info).
-- Each check writes to the list appropriate for its severity.
--- @type Findings
local findings = { errors = {}, warnings = {}, info = {} }
-- Lift parse-time errors already recorded in scan_source's atom_info payload into this pass's findings list.
--- @type integer, AtomInfoEntry
for _, info in ipairs(scan.atom_infos) do
if info.errors then
--- @type integer, string
for _, msg in ipairs(info.errors) do
findings.errors[#findings.errors + 1] = {
line = info.info_line,
@@ -494,6 +532,7 @@ local function validate(ctx, src, corpus_pipe_ctx)
end
-- THE per-annotation pipeline. ONE loop. CHECK_RULES dispatches per_annot rules.
--- @type integer, AtomInfoEntry
for _, info in ipairs(scan.atom_infos) do
duffle.run_check_rules(CHECK_RULES, "per_annot", info, pipe_ctx, findings)
end
@@ -503,13 +542,16 @@ local function validate(ctx, src, corpus_pipe_ctx)
-- scan_source records each marker in scan.debug_skip_markers; this loop validates each record independently and emits at most one error per marker.
-- Valid markers stamp `debug_skip = true` on the following atom or component declaration, which downstream consumers read directly.
--- @type DebugSkipMarker[]
local skip_markers = scan.debug_skip_markers or {}
--- @type integer, DebugSkipMarker
for _, marker in ipairs(skip_markers) do
duffle.run_check_rules(CHECK_RULES, "per_skip_marker", marker, pipe_ctx, findings)
end
-- Per-macro rules (TAPE_WORDS vs WORD_COUNT drift).
pipe_ctx.word_counts = corpus_pipe_ctx.word_counts
--- @type integer, MacroEntry
for _, m in ipairs(scan.macros) do
duffle.run_check_rules(CHECK_RULES, "per_macro", m, pipe_ctx, findings)
end
@@ -540,8 +582,7 @@ end
-- M.run — orchestrator entry
-- ════════════════════════════════════════════════════════════════════════════
--- @class M
--- @type AnnotationPass
local M = {}
-- Expose `validate` for downstream passes (e.g. report.lua) that need to re-render the per-source results into a per-MODULE report.
@@ -550,31 +591,46 @@ M.validate = validate
--- @param ctx PassCtx
--- @return PassResult
function M.run(ctx)
--- @type PassOutputEntry[]
local outputs = {}
--- @type PassFinding[]
local errors = {}
--- @type PassFinding[]
local warnings = {}
-- Build the shared pipe_ctx once for this run; every validate() call sees the same cross-source registries.
-- The corpus owns the canonical cross-source registries; per-source scans retain body / declaration ownership.
--- @type PipeCtx
local corpus_pipe_ctx = build_corpus_pipe_ctx(ctx)
--- @type Corpus
local corpus = ctx.shared.corpus
-- Group `corpus.sources_by_dir` by module, validate every source in each bucket, and emit one errors.h per directory.
--- @type table<string, SourceFile[]>
local by_dir = (corpus and corpus.sources_by_dir) or {}
--- @type string, SourceFile[]
for dir, dir_sources in pairs(by_dir) do
--- @type string
local dir_basename = dir:match("([^/\\]+)$") or dir
--- @type integer
local dir_atoms = 0
--- @type PassFinding[]
local dir_errors = {}
--- @type PassFinding[]
local dir_warnings = {}
--- @type integer, SourceFile
for _, src in ipairs(dir_sources) do
--- @type AnnotatedResult
local result = validate(ctx, src, corpus_pipe_ctx)
result.source = src.path -- tag for downstream rendering
dir_atoms = dir_atoms + #result.atoms
--- @type integer, PassFinding
for _, e in ipairs(result.errors) do
dir_errors[#dir_errors + 1] = { line = e.line, msg = e.msg, source = src.path }
errors [#errors + 1] = { line = e.line, msg = e.msg }
end
--- @type integer, PassFinding
for _, w in ipairs(result.warnings) do
dir_warnings[#dir_warnings + 1] = { line = w.line, msg = w.msg }
warnings [#warnings + 1] = { line = w.line, msg = w.msg }
scripts/passes/atoms_source_map.lua
+149 -26
View File
@@ -37,8 +37,11 @@
-- Bootstrap: load `duffle_paths.lua` via `debug.getinfo(1, "S").source`
-- (works both standalone + when require'd). `duffle_paths.lua` sets package.path then returns `require("duffle")`
-- at the bottom, so the dofile value IS the duffle module.
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--- @type ElfDwarfMod
local elf_dwarf = require("elf_dwarf")
-- ════════════════════════════════════════════════════════════════════════════
@@ -47,6 +50,7 @@ local elf_dwarf = require("elf_dwarf")
-- Format version emitted as the first line. Bump + add a migration test if the format changes;
-- the gdb runtime loader rejects mismatches (E2).
--- @type integer
local FORMAT_VERSION = 1
-- ════════════════════════════════════════════════════════════════════════════
@@ -54,29 +58,71 @@ local FORMAT_VERSION = 1
-- ════════════════════════════════════════════════════════════════════════════
--- @class AtomSourceMapCtx
--- @field shared table -- `ctx.shared`
--- @field shared.corpus table -- source-order registry; single writer is build_ctx
--- @field shared.word_counts table
--- @field out_root string -- output root (e.g. "build/gen")
--- @field flags table -- `ctx.flags`; reads `flags.gdb_runtime` + `flags.elf_path`
--- @field shared PassShared
--- @field out_root string
--- @field flags PassFlags
--- @field project_root string|nil
--- @class WordMapEntry
--- @field pos integer
--- @field line integer
--- @field text string
--- @field body_line integer
--- @field gpr_keys string[]|nil
--- @field invocation InvocationRecord|nil
--- @class NmAddr
--- @field [1] integer -- st_value
--- @field [2] integer -- st_size
--- @class GdbAtomRecord
--- @field idx integer|nil
--- @field name string
--- @field src_path string
--- @field file_base string
--- @field addr integer
--- @field size_bytes integer
--- @field words integer
--- @field entries WordMapEntry[]
--- @class ElfDwarfMod
--- @field read_nm fun(elf_path: Path): table<string, NmAddr>
--- @class AtomSourceMapPass
--- @field render_source_map fun(src: SourceFile): string
--- @field render_provenance fun(src: SourceFile, wc: WordCounts): string
--- @field render_atom_source_map fun(atom: AtomEntry): string
--- @field render_atom_provenance fun(atom: AtomEntry, wc: WordCounts, rel_path: string): string
--- @field run fun(ctx: PassCtx): PassResult
--- @class AtomEntry
--- @field paths AtomPaths|nil
-- ════════════════════════════════════════════════════════════════════════════
-- Atom-path renderers
-- ════════════════════════════════════════════════════════════════════════════
--- Join word boundaries (from `items`) to per-word call text + source lines (from `word_events`).
--- @param atom table
--- @return table[], integer
--- @param atom AtomEntry
--- @return WordMapEntry[]
--- @return integer
local function canonical_word_entries(atom)
--- @type AtomPaths
local paths = atom.paths or {}
--- @type WordEvent[]
local events = paths.word_events or {}
--- @type EmissionItem[]
local word_items = {}
--- @type integer, EmissionItem
for _, item in ipairs(paths.items or {}) do
if item.kind == "word" then word_items[#word_items + 1] = item end
end
--- @type WordMapEntry[]
local entries = {}
--- @type integer, WordEvent
for index, event in ipairs(events) do
--- @type EmissionItem
local item = word_items[index] or {}
entries[#entries + 1] = {
pos = event.i or (index - 1),
@@ -97,18 +143,25 @@ end
--- `WORD N CALL <src-path>:<src-line> RAW` (raw `.word` outside any mac_* component)
--- Component identity comes from the outermost invocation record; the count-table lookup confirms the component was declared in `corpus.word_counts`
--- (populated by word_count_eval + components passes).
--- @param src table
--- @param atom table
--- @param wc table -- identity alias of corpus.word_counts
--- @return string[], integer
--- @param src SourceFile
--- @param atom AtomEntry
--- @param wc WordCounts
--- @return string[]
--- @return integer
local function emit_provenance_stanza(src, atom, wc)
--- @type string[]
local lines = {}
--- @type string
local rel_path = src.path:gsub("\\\\", "/")
--- @type WordMapEntry[], integer
local entries, total = canonical_word_entries(atom)
lines[#lines + 1] = string.format('ATOM %s "%s" 0', atom.raw_name or atom.name, rel_path)
--- @type integer, WordMapEntry
for _, entry in ipairs(entries) do
--- @type InvocationRecord|nil
local inv = entry.invocation
--- @type integer|nil
local macro_count = inv and wc["mac_" .. inv.component_name]
if inv and macro_count ~= nil then
lines[#lines + 1] = string.format('WORD %d CALL %s:%d MACRO %s "%s:%d" BODY %d'
@@ -125,10 +178,11 @@ local function emit_provenance_stanza(src, atom, wc)
end
--- Render the full provenance file content for one source.
--- @param src table
--- @param wc table
--- @param src SourceFile
--- @param wc WordCounts
--- @return string
local function render_provenance(src, wc)
--- @type string[]
local lines = {}
lines[#lines + 1] = "# FORMAT_VERSION 1"
lines[#lines + 1] = "# auto-generated by ps1_meta.lua (passes/atoms_source_map.lua) — DO NOT EDIT"
@@ -138,13 +192,19 @@ local function render_provenance(src, wc)
lines[#lines + 1] = "# dwarf_injection to synthesize DW_TAG_inlined_subroutine instances + per-word"
lines[#lines + 1] = "# line program rows for native source-level step into component bodies."
--- @param atom AtomEntry
--- @return nil
local function append(atom)
--- @type string[]
local stanza = emit_provenance_stanza(src, atom, wc)
--- @type integer, string
for _, line in ipairs(stanza) do lines[#lines + 1] = line end
end
--- @type integer, AtomEntry
for _, atom in ipairs(src.scan.atoms or {}) do
if atom.paths then append(atom) end
end
--- @type integer, AtomEntry
for _, atom in ipairs(src.scan.raw_atoms or {}) do
if atom.paths then append(atom) end
end
@@ -154,16 +214,20 @@ end
--- Render one atom's stanza for the sourcemap.txt form (ATOM header line, N WORD lines, ENDATOM marker).
--- Returns (lines, total_words).
--- @param src table
--- @param atom table
--- @param wc table
--- @return string[], integer
--- @param src SourceFile
--- @param atom AtomEntry
--- @return string[]
--- @return integer
local function emit_atom_stanza(src, atom)
--- @type string[]
local lines = {}
--- @type string
local rel_path = src.path:gsub("\\\\", "/")
--- @type WordMapEntry[], integer
local entries, total = canonical_word_entries(atom)
lines[#lines + 1] = string.format('ATOM %s "%s" 0', atom.raw_name or atom.name, rel_path)
--- @type integer, WordMapEntry
for _, entry in ipairs(entries) do
lines[#lines + 1] = string.format("WORD %d LINE %d TEXT %s",
entry.pos, entry.line, entry.text)
@@ -175,21 +239,27 @@ end
--- Render the full source map file content for one source (one .atoms.sourcemap.txt per source). Mirrors offsets.lua's
--- `project_atoms` shape: scan.atoms + scan.raw_atoms, no kind filter.
--- @param src table
--- @param wc table
--- @param src SourceFile
--- @return string
local function render_source_map(src)
--- @type string[]
local lines = {}
lines[#lines + 1] = "# FORMAT_VERSION " .. FORMAT_VERSION
lines[#lines + 1] = "# auto-generated by ps1_meta.lua (passes/atoms_source_map.lua) — DO NOT EDIT"
--- @param atom AtomEntry
--- @return nil
local function append(atom)
--- @type string[]
local stanza = emit_atom_stanza(src, atom)
--- @type integer, string
for _, line in ipairs(stanza) do lines[#lines + 1] = line end
end
--- @type integer, AtomEntry
for _, atom in ipairs(src.scan.atoms or {}) do
if atom.paths then append(atom) end
end
--- @type integer, AtomEntry
for _, atom in ipairs(src.scan.raw_atoms or {}) do
if atom.paths then append(atom) end
end
@@ -211,19 +281,29 @@ end
--- Build the list of atoms with addresses + word entries. Shared helper for the gdb-runtime file emission.
--- @param ctx PassCtx
--- @return table[] -- list of {idx, name, src_path, file_base, addr, size_bytes, words, entries}
--- @return GdbAtomRecord[]
local function build_atom_table(ctx)
--- @type table<string, NmAddr>
local addrs = elf_dwarf.read_nm(ctx.flags.elf_path)
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
--- @type GdbAtomRecord[]
local matched = {}
--- @type integer, SourceFile
for _, src in ipairs(corpus.source_order or {}) do
--- @type string
local file_base = src.path:match("([^/\\\\]+)$") or src.path
--- @param atom AtomEntry
--- @return nil
local function append(atom)
if not atom.paths then return end
--- @type string
local name = atom.raw_name or atom.name
--- @type NmAddr|nil
local info = addrs[name]
if not info then return end
--- @type WordMapEntry[], integer
local entries, total = canonical_word_entries(atom)
matched[#matched + 1] = {
name = name,
@@ -235,12 +315,18 @@ local function build_atom_table(ctx)
entries = entries,
}
end
--- @type integer, AtomEntry
for _, atom in ipairs((src.scan or {}).atoms or {}) do append(atom) end
--- @type integer, AtomEntry
for _, atom in ipairs((src.scan or {}).raw_atoms or {}) do append(atom) end
end
-- Deterministic order: sort by address (matches `nm` output ordering).
--- @param a GdbAtomRecord
--- @param b GdbAtomRecord
--- @return boolean
table.sort(matched, function(a, b) return a.addr < b.addr end)
--- @type integer, GdbAtomRecord
for i, a in ipairs(matched) do a.idx = i - 1 end
return matched
end
@@ -252,12 +338,14 @@ end
---
--- Why hardcoded per-atom: gdb's `$` substitution doesn't concat inside var names — `$__atom_name_$__i` in a `while`
--- loop resolves to one literal identifier, not `name_i`. Compile-time emission is the only path.
--- @param lines table -- output line buffer (mutated in place)
--- @param matched table -- list of atom records from `build_atom_table`
--- @param lines string[]
--- @param matched GdbAtomRecord[]
--- @return nil
local function append_gdb_commands(lines, matched)
-- ── tape_atoms ──
-- Hardcoded one printf per atom. No loop.
lines[#lines + 1] = "define tape_atoms"
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
-- gdb 12.1 quirk: literals in printf args require an attached target.
-- Use the per-atom convenience vars set above as printf args.
@@ -273,6 +361,7 @@ local function append_gdb_commands(lines, matched)
-- ── break_atom (generic) + per-atom break_atom_X ──
lines[#lines + 1] = "define break_atom"
lines[#lines + 1] = ' echo "Usage: break_atom_<exact_name> (pick from the list below)"'
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
lines[#lines + 1] = string.format(' printf " break_atom_%%-32s\\n", $__atom_name_%d', a.idx)
end
@@ -282,6 +371,7 @@ local function append_gdb_commands(lines, matched)
lines[#lines + 1] = "end"
lines[#lines + 1] = ""
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
lines[#lines + 1] = string.format("define break_atom_%s", a.name)
lines[#lines + 1] = string.format(" break *$__atom_addr_%d", a.idx)
@@ -296,6 +386,7 @@ local function append_gdb_commands(lines, matched)
-- ── step_atom / next_atom ──
-- Hardcoded one tbreak per atom. No loop.
lines[#lines + 1] = "define step_atom"
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
lines[#lines + 1] = string.format(" tbreak *$__atom_addr_%d", a.idx)
end
@@ -319,6 +410,7 @@ local function append_gdb_commands(lines, matched)
lines[#lines + 1] = "define where_in_atom"
lines[#lines + 1] = " set $__pc = (unsigned int)$pc"
lines[#lines + 1] = " set $__matched = 0"
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
-- Precompute end_addr (gdb 12.1's expression evaluator chokes on `addr + words*4`).
lines[#lines + 1] = string.format(" set $__end_%d = $__atom_addr_%d + $__atom_words_%d * 4", a.idx, a.idx, a.idx)
@@ -328,14 +420,17 @@ local function append_gdb_commands(lines, matched)
lines[#lines + 1] = string.format(" set $__word = ($__pc - $__atom_addr_%d) / 4", a.idx)
lines[#lines + 1] = string.format(' printf "word: %%d/%%d\\n", $__word, $__atom_words_%d', a.idx)
-- One inner-if per WORD entry. Each word's line + text hardcoded.
--- @type integer, WordMapEntry
for _, we in ipairs(a.entries) do
lines[#lines + 1] = string.format(" if $__word == %d", we.pos)
-- Escape TEXT for printf format string.
--- @type string
local escaped_text = we.text:gsub("%%", "%%%%"):gsub('"', '\\"')
lines[#lines + 1] = string.format(' printf "source: %%s:%%d %%s\\n", $__atom_file_%d, %d, "%s"', a.idx, we.line, escaped_text)
lines[#lines + 1] = " end"
end
-- Fallback for words beyond the source map (shouldn't happen if nm matches).
--- @type integer
local max_word = 0
if #a.entries > 0 then max_word = a.entries[#a.entries].pos end
lines[#lines + 1] = string.format(' if $__word > %d', max_word)
@@ -361,6 +456,7 @@ local function append_gdb_commands(lines, matched)
lines[#lines + 1] = " set $__in_atom = 0"
lines[#lines + 1] = " set $__did_step = 0"
lines[#lines + 1] = " set $__pc = (unsigned int)$pc"
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
-- Precompute end_addr in the convenience var (single expression gdb handles).
lines[#lines + 1] = string.format(" set $__end_%d = $__atom_addr_%d + $__atom_words_%d * 4", a.idx, a.idx, a.idx)
@@ -395,8 +491,10 @@ end
--- Emit the gdb-runtime file (post-link). Pure gdb scripting — addresses come from `mipsel-none-elf-nm -S`, get embedded
--- in `<ctx.out_root>/gdb_tape_atoms_runtime.gdb`, and load via `set $var = ...` + `define ... end` blocks at gdb source-time.
--- @param ctx PassCtx
--- @return nil
local function emit_gdb_runtime(ctx)
if not (ctx.flags and ctx.flags.gdb_runtime) then return end
--- @type string|nil
local elf_path = ctx.flags.elf_path
if not elf_path or elf_path == "" then
io.stderr:write("[atoms_source_map] --gdb-runtime requires --elf <elf>\n")
@@ -408,12 +506,14 @@ local function emit_gdb_runtime(ctx)
return
end
--- @type GdbAtomRecord[]
local matched = build_atom_table(ctx)
if #matched == 0 then
io.stderr:write("[atoms_source_map] --gdb-runtime: no atoms matched against nm symbols (stale scan?).\n")
return
end
--- @type string[]
local lines = {}
lines[#lines + 1] = "# Auto-generated by ps1_meta.lua (passes/atoms_source_map.lua)"
lines[#lines + 1] = "# DO NOT EDIT — re-run ps1_meta.lua --atoms-source-map --gdb-runtime to regenerate"
@@ -435,6 +535,7 @@ local function emit_gdb_runtime(ctx)
-- Per-atom convenience vars (used as printf args; literals aren't accepted
-- without an attached target on gdb 12.1).
--- @type integer, GdbAtomRecord
for _, a in ipairs(matched) do
lines[#lines + 1] = string.format('set $__atom_name_%d = "%s"', a.idx, gdb_escape(a.name))
lines[#lines + 1] = string.format("set $__atom_addr_%d = 0x%x", a.idx, a.addr)
@@ -451,10 +552,13 @@ local function emit_gdb_runtime(ctx)
-- Confirmation line for the source operator.
lines[#lines + 1] = 'printf "[gdb_tape_atoms] runtime loaded %d atoms from %s\\n", $__atom_count, $__elf_path'
--- @type string
local out_path
-- Move out of `<out_root>/gdb_tape_atoms_runtime.gdb` to `<out_root>/../gdb_tape_atoms_runtime.gdb` when the conventional `<out_root>` is `<build>/gen`
-- (any equivalent spelling — relative, absolute backslash, absolute forward-slash, trailing-separator variants).
-- This puts the gdb runtime alongside the ELF at `build/` rather than under the report subdir.
--- @param p string
--- @return boolean
local function ends_with_gen_dir(p)
if type(p) ~= "string" then return false end
return p:match("[/\\]gen[/\\]?$") ~= nil or p == "build/gen" or p == "build\\gen"
@@ -462,6 +566,7 @@ local function emit_gdb_runtime(ctx)
if ends_with_gen_dir(ctx.out_root) then
-- Strip the trailing `/gen` segment, then write the runtime script under `build/`.
-- e.g. "C:/projects/Pikuma/ps1/build/gen" -> "C:/projects/Pikuma/ps1/build".
--- @type string
local parent = ctx.out_root:gsub("[/\\]gen[/\\]?$", "")
out_path = parent .. "/gdb_tape_atoms_runtime.gdb"
else
@@ -476,6 +581,7 @@ end
-- M — module exports
-- ════════════════════════════════════════════════════════════════════════════
--- @type AtomSourceMapPass
local M = {}
-- Expose the pure render functions so `report.lua` and the focused tests can call them directly without triggering the file-emit path.
@@ -483,19 +589,26 @@ M.render_source_map = render_source_map
M.render_provenance = render_provenance
--- Render ONE atom's sourcemap stanza.
--- @param atom table -- atom record (must have `atom.paths` populated)
--- @param atom AtomEntry
--- @return string
function M.render_atom_source_map(atom)
assert(type(atom) == "table", "render_atom_source_map: atom must be a table")
assert(type(atom.paths) == "table", "render_atom_source_map: atom.paths must be a table")
--- @type WordMapEntry[], integer
local entries, total = canonical_word_entries(atom)
--- @type string[]
local lines = {}
lines[#lines + 1] = string.format("ATOM %s %d", (atom.raw_name or atom.name), total)
--- @type integer, WordMapEntry
for _, entry in ipairs(entries) do
--- @type string
local word_line = string.format("WORD %d LINE %d TEXT %s",
entry.pos, entry.line, entry.text)
--- @type string[]
local keys = {}
--- @type integer
for pos = 1, 16 do
--- @type string|nil
local k = entry.gpr_keys and entry.gpr_keys[pos]
if type(k) == "string" and k:sub(1, 7) == "reguse:" then
keys[#keys + 1] = k
@@ -514,19 +627,24 @@ end
---
--- `rel_path` is the source path (forward-slashes) embedded in every `CALL` line.
--- The .md caller (report.lua) is expected to derive this once per `## <source>` heading and pass it down for each atom in that source.
--- @param atom table -- atom record (must have `atom.paths` populated)
--- @param wc table -- identity alias of `corpus.word_counts`
--- @param rel_path string -- source path (forward-slashes) for `CALL` fields
--- @param atom AtomEntry
--- @param wc WordCounts
--- @param rel_path string
--- @return string
function M.render_atom_provenance(atom, wc, rel_path)
assert(type(atom) == "table", "render_atom_provenance: atom must be a table")
assert(type(atom.paths) == "table", "render_atom_provenance: atom.paths must be a table")
assert(type(rel_path) == "string", "render_atom_provenance: rel_path must be a string")
--- @type WordMapEntry[], integer
local entries, total = canonical_word_entries(atom)
--- @type string[]
local lines = {}
lines[#lines + 1] = string.format("ATOM %s %d", (atom.raw_name or atom.name), total)
--- @type integer, WordMapEntry
for _, entry in ipairs(entries) do
--- @type InvocationRecord|nil
local inv = entry.invocation
--- @type integer|nil
local macro_count = inv and wc and wc["mac_" .. inv.component_name]
if inv and macro_count ~= nil then
lines[#lines + 1] = string.format('WORD %d CALL %s:%d MACRO %s "%s:%d" BODY %d'
@@ -546,16 +664,21 @@ end
--- @param ctx PassCtx
--- @return PassResult
function M.run(ctx)
--- @type PassOutputEntry[]
local outputs = {}
--- @type PassFinding[]
local errors = {}
--- @type PassFinding[]
local warnings = {}
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" or type(corpus.source_order) ~= "table" then
error("atoms_source_map.run requires ctx.shared.corpus.source_order (canonical corpus).", 0)
end
-- Word counts come from `corpus.word_counts` (populated by word_count_eval + components passes).
--- @type WordCounts
local wc = corpus.word_counts or {}
if not next(wc) then
warnings[#warnings + 1] = {
scripts/passes/auto_reg.lua
+104 -4
View File
@@ -18,12 +18,26 @@
--- Pool exhaustion: If a phase declares more `R_<Sym>` mappings than the 10-register pool can hold,
--- emit `phase_register_pool_exhausted` as a build-stopping error.
--- @class AutoRegResult
--- @field outputs table[] -- {kind=, path=} entries
--- @field errors table[] -- {line=, msg=} entries (build-stops)
--- @field warnings table[] -- {line=, msg=} entries (build-continues)
--- @alias GprIdent string
--- @class GprAllocMap
--- @field [string] GprIdent -- bag: auto-reg symbol -> physical GPR
--- @class AutoRegOutput
--- @field auto_reg_h string
--- @class AutoRegResult
--- @field outputs AutoRegOutput[]
--- @field errors PassFinding[]
--- @field warnings PassFinding[]
--- @class AutoRegPass
--- @field run fun(ctx: PassCtx): AutoRegResult
--- @field POOL GprIdent[]
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--- ════════════════════════════════════════════════════════════════════════════
@@ -42,6 +56,7 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--- R_K0/K1 (codes 26-27) — Kernel / interrupt handler reserves. Never touched by user code.
--- R_GP/SP/FP/RA (codes 28-31) — R_SP/R_FP/R_RA are tape-runtime carriers between tape_enter and tape_exit; R_GP stays the host global pointer.
---
--- @type GprIdent[]
local POOL = {
"R_V0", "R_V1",
"R_T0", "R_T1", "R_T2", "R_T3",
@@ -57,6 +72,7 @@ local POOL = {
-- Only the POOL entries matter for auto_reg — non-pool aliases
-- (R_AT=1, R_A0..A3=4..7, R_T8=24, R_T9=25, R_K0/K1=26..27, R_GP/SP/FP/RA=28..31)
-- are deliberately omitted — see the comment block above for the WHY of each exclusion.
--- @type table<integer, GprIdent> -- bag: MIPS GPR code -> POOL ident
local INT_CODE_TO_POOL_GPR = {
[2] = "R_V0", [3] = "R_V1",
[4] = "R_A0", [5] = "R_A1", [6] = "R_A2", [7] = "R_A3",
@@ -68,8 +84,12 @@ local INT_CODE_TO_POOL_GPR = {
}
-- Stable sort for deterministic allocation order.
--- @param tbl table<string, string> -- bag: key set only; values unused
--- @return string[]
local function stable_sort_keys(tbl)
--- @type string[]
local keys = {}
--- @type string
for k in pairs(tbl) do keys[#keys + 1] = k end
table.sort(keys)
return keys
@@ -77,15 +97,25 @@ end
-- Allocate one phase's auto-reg mappings.
-- Returns (allocated_map, errors). On pool exhaustion, errors is populated and the function halts.
--- @param phase_label string
--- @param decls table<string, string> -- bag: auto-reg symbol -> decl payload
--- @return GprAllocMap
--- @return PassFinding[]
local function allocate_phase(phase_label, decls)
-- Deep-copy POOL into a fresh sequence table. The original `table.unpack and table.unpack(POOL) or { unpack(POOL) }`
-- idiom wraps the unpacked values in a single inner table under LuaJIT 5.1 (`table.unpack` is nil; the `or` returns one value),
-- which corrupts the pool into `{ {R_T0, R_T1, ...} }` — making `table.remove(pool, 1)` return the inner table on iteration.
--- @type GprIdent[]
local pool = {}
--- @type integer
for i = 1, #POOL do pool[i] = POOL[i] end
--- @type GprAllocMap
local result = {}
--- @type PassFinding[]
local errors = {}
--- @type integer, string
for _, sym in ipairs(stable_sort_keys(decls)) do
--- @type GprIdent|nil
local next_gpr = table.remove(pool, 1)
if not next_gpr then
errors[#errors + 1] = {
@@ -110,12 +140,19 @@ end
-- Each entry's `code` is the integer MIPS GPR number (0..31); INT_CODE_TO_POOL_GPR translates it back to the physical GPR ident.
-- Aliases whose `code` points to a non-POOL GPR (e.g. R_S0, R_T8, R_K1) are ignored —
-- they don't affect the auto_reg pool, and they're already excluded from POOL above.
--- @param corpus Corpus
--- @return table<GprIdent, boolean>
--- @return table<string, GprIdent>
local function build_user_pins(corpus)
--- @type table<GprIdent, boolean> -- bag: pinned physical GPR -> true
local user_pinned = {}
--- @type table<string, GprIdent> -- bag: alias ident -> physical GPR
local alias_to_gpr = {}
if not corpus.register_alias_registry then return user_pinned, alias_to_gpr end
--- @type string, AliasEntry
for alias_name, alias_entry in pairs(corpus.register_alias_registry) do
if alias_entry.has_atom_reg and alias_entry.code then
--- @type GprIdent|nil
local gpr = INT_CODE_TO_POOL_GPR[alias_entry.code]
if gpr then
user_pinned[gpr] = true
@@ -133,22 +170,32 @@ end
--- Clash-detection and source-pool-exclusion logic only needs the presence of each GPR (boolean test),
--- but keeping count preserves the original find_hardcoded_rn shape so callers can switch without churn.
--- The alias pattern is sorted lexicographically to keep the regex deterministic.
--- @param body_text string
--- @param alias_to_gpr table<string, GprIdent> -- bag: alias ident -> physical GPR
--- @return table<GprIdent, integer>
local function find_used_gprs(body_text, alias_to_gpr)
--- @type table<GprIdent, integer> -- bag: physical GPR -> hit count
local found = {}
-- (a) Hardcoded physical GPRs (R_T0..R_T7, R_V0..R_V1, R_A0..R_A3, R_S0..R_S7).
--- @type GprIdent
for gpr in body_text:gmatch("(R_T%d+|R_V%d+|R_A%d+|R_S%d+)") do
found[gpr] = (found[gpr] or 0) + 1
end
-- (b) Alias references (R_<Alias>) resolved to physical GPRs via the registry.
-- Sorted by name so the regex is byte-stable across runs.
if alias_to_gpr and next(alias_to_gpr) then
--- @type string[]
local aliases = {}
--- @type string
for alias_name in pairs(alias_to_gpr) do
aliases[#aliases + 1] = alias_name
end
table.sort(aliases)
--- @type string
local pattern = "(" .. table.concat(aliases, "|") .. ")"
--- @type string
for alias_name in body_text:gmatch(pattern) do
--- @type GprIdent|nil
local gpr = alias_to_gpr[alias_name]
if gpr and not found[gpr] then
found[gpr] = 1
@@ -159,10 +206,17 @@ local function find_used_gprs(body_text, alias_to_gpr)
end
-- Emit one gen/auto_reg.h header per directory.
--- @param out_dir string
--- @param dir string
--- @param sources SourceFile[]
--- @param mappings GprAllocMap
--- @return string|nil
local function emit_auto_reg_h(out_dir, dir, sources, mappings)
if not mappings or next(mappings) == nil then return end
--- @type string
local out_path = out_dir .. "/" .. "auto_reg.h"
duffle.ensure_dir(out_dir)
--- @type string[]
local lines = {
"#ifdef INTELLISENSE_DIRECTIVES",
"#pragma once",
@@ -170,14 +224,18 @@ local function emit_auto_reg_h(out_dir, dir, sources, mappings)
"// Auto-generated by ps1_meta.lua (passes/auto_reg.lua) — DO NOT EDIT",
"// Directory: " .. dir:gsub("/", "\\"),
}
--- @type integer, SourceFile
for _, src in ipairs(sources) do
lines[#lines + 1] = "// source: " .. src.path
end
lines[#lines + 1] = "// Per-phase register allocations resolved by the lua pass."
lines[#lines + 1] = "// R_<Sym>_Code = <chosen GPR's _Code constant> for every marker in this directory."
lines[#lines + 1] = ""
--- @type integer, string
for _, sym in ipairs(stable_sort_keys(mappings)) do
--- @type GprIdent
local gpr = mappings[sym]
--- @type string
local gpr_code = gpr .. "_Code"
lines[#lines + 1] = "#define " .. sym .. "_Code " .. gpr_code
end
@@ -191,15 +249,20 @@ end
-- Pass entry
-- ════════════════════════════════════════════════════════════════════════════
--- @type AutoRegPass
local M = {}
--- @param ctx PassCtx
--- @return AutoRegResult
function M.run(ctx)
--- @type AutoRegOutput[]
local outputs = {}
--- @type PassFinding[]
local errors = {}
--- @type PassFinding[]
local warnings = {}
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" then
error("auto_reg.run requires ctx.shared.corpus", 0)
@@ -210,16 +273,22 @@ function M.run(ctx)
-- MUST NOT be allocated to any auto-reg marker — they're preserved across atoms by the wave-context discipline.
-- The corpus's register_alias_registry is the source of truth for these opt-in pins.
-- Body references to those aliases (via alias_to_gpr) are also excluded on a per-atom basis in step 2 below.
--- @type table<GprIdent, boolean>, table<string, GprIdent>
local user_pinned, alias_to_gpr = build_user_pins(corpus)
-- 1. Allocate phase pools first (phase declarations take precedence over per-atom declarations).
--- @type table<string, GprAllocMap> -- bag: phase_label -> alloc map
local phase_allocations = {}
--- @type string, table<string, string>
for phase_label, decls in pairs(corpus.phase_auto_regs or {}) do
--- @type GprAllocMap, PassFinding[]
local mapping, errs = allocate_phase(phase_label, decls)
--- @type string, GprIdent
for sym, gpr in pairs(mapping) do
phase_allocations[phase_label] = phase_allocations[phase_label] or {}
phase_allocations[phase_label][sym] = gpr
end
--- @type integer, PassFinding
for _, e in ipairs(errs) do
errors[#errors + 1] = e
end
@@ -229,15 +298,21 @@ function M.run(ctx)
-- Otherwise, allocate a private pool for the atom.
-- The phase membership is in `corpus.atom_phases[phase_label].atoms` (an array of atom names declared via `atom_phase(<phase>)`
-- in the atom's `atom_info` line). Build a reverse map `atom_name -> phase_label` so the lookup is O(1) per atom scope.
--- @type table<AtomName, string> -- bag: atom name -> phase label
local atom_name_to_phase = {}
--- @type string, AtomPhaseGroup
for phase_label, entry in pairs(corpus.atom_phases or {}) do
--- @type integer, AtomName
for _, atom_name in ipairs(entry.atoms or {}) do
atom_name_to_phase[atom_name] = phase_label
end
end
--- @type table<AtomName, GprAllocMap> -- bag: atom scope -> alloc map
local atom_allocations = {}
--- @type AtomName, table<string, string>
for atom_scope, decls in pairs(corpus.atom_auto_regs or {}) do
--- @type string|nil
local phase_label = atom_name_to_phase[atom_scope]
-- Build the atom's source pool: start with the full POOL, subtract:
-- (a) every GPR already committed (phase allocations + prior atom allocations)
@@ -248,25 +323,36 @@ function M.run(ctx)
-- the original `source_pool = phase_allocations[phase_label]` form used the phase
-- allocation MAP as a pool, but that map has no array part, so `table.remove(source_pool, 1)`
-- returned nil and every atom-with-phase marker errored with `phase_register_pool_exhausted`.
--- @type table<GprIdent, boolean> -- bag: committed or body-referenced GPR -> true
local used = {}
--- @type integer, GprAllocMap
for _, m in pairs(phase_allocations) do for _, gpr in pairs(m) do used[gpr] = true end end
--- @type integer, GprAllocMap
for _, m in pairs(atom_allocations) do for _, gpr in pairs(m) do used[gpr] = true end end
-- (c) Body references — scan the atom body for hardcoded + alias-resolved GPRs.
-- Folded into `used` so the source_pool exclusion is a single check.
--- @type AtomEntry|nil
local atom = corpus.atoms_by_name and corpus.atoms_by_name[atom_scope]
if atom and atom.body then
--- @type table<GprIdent, integer>
local body_used = find_used_gprs(atom.body, alias_to_gpr)
--- @type GprIdent
for gpr in pairs(body_used) do used[gpr] = true end
end
--- @type GprIdent[]
local source_pool = {}
--- @type integer, GprIdent
for _, gpr in ipairs(POOL) do
-- Exclude (a) prior commitments, (b) USER-PINNED GPRs (wave-context carriers declared via atom_reg + _Code defs, preserved across atoms globally).
if not used[gpr] and not user_pinned[gpr] then
source_pool[#source_pool + 1] = gpr
end
end
--- @type GprAllocMap
local result = {}
--- @type integer, string
for _, sym in ipairs(stable_sort_keys(decls)) do
--- @type GprIdent|nil
local next_gpr = table.remove(source_pool, 1)
if not next_gpr then
errors[#errors + 1] = {
@@ -289,10 +375,14 @@ function M.run(ctx)
-- This warning is kept as a defensive safety net for cases the body scanner might miss
-- (e.g. macros that expand to register references the scanner cannot resolve).
-- For each resolved (scope, sym) -> R_Tn mapping, scan the atom body source for used GPRs.
--- @type AtomName, GprAllocMap
for atom_scope, decls in pairs(atom_allocations) do
--- @type AtomEntry|nil
local atom = corpus.atoms_by_name and corpus.atoms_by_name[atom_scope]
if atom and atom.body then
--- @type table<GprIdent, integer>
local used_in_body = find_used_gprs(atom.body, alias_to_gpr)
--- @type string, GprIdent
for sym, allocated_gpr in pairs(decls) do
if used_in_body[allocated_gpr] and used_in_body[allocated_gpr] > 0 then
warnings[#warnings + 1] = {
@@ -308,26 +398,36 @@ function M.run(ctx)
-- 4. Emit per-directory gen/auto_reg.h.
-- For each source directory that has atom_auto_regs or phase_auto_regs entries, emit one header.
--- @type table<string, SourceFile[]>
local sources_by_dir = corpus.sources_by_dir or {}
--- @type string, SourceFile[]
for dir, sources in pairs(sources_by_dir) do
--- @type GprAllocMap
local per_dir_mappings = {}
--- @type integer, SourceFile
for _, src in ipairs(sources) do
-- Collect every (sym -> gpr) entry that originated from a source in this directory.
-- `src.scan.atom_auto_regs` is keyed by ATOM SCOPE NAME; `pairs(t)` iterates KEYS so `scope_name` here is the scope ident (e.g. "cube_g4_face").
-- The previous `for _, scan_atom_auto` form silently assigned the VALUE (a `{sym = sym}` table) to the variable,
-- which made `atom_allocations[scan_atom_auto]` a table-indexed lookup that never resolved.
--- @type string
for scope_name in pairs(src.scan and src.scan.atom_auto_regs or {}) do
--- @type string, GprIdent
for sym, gpr in pairs(atom_allocations[scope_name] or {}) do
per_dir_mappings[sym] = gpr
end
end
--- @type string
for scope_name in pairs(src.scan and src.scan.phase_auto_regs or {}) do
--- @type string, GprIdent
for sym, gpr in pairs(phase_allocations[scope_name] or {}) do
per_dir_mappings[sym] = gpr
end
end
end
--- @type string
local out_dir = dir .. "/gen"
--- @type string|nil
local out_path = emit_auto_reg_h(out_dir, dir, sources, per_dir_mappings)
if out_path then outputs[#outputs + 1] = { auto_reg_h = out_path } end
end
scripts/passes/components.lua
+203 -53
View File
@@ -22,7 +22,9 @@
-- Uses `debug.getinfo` to find this file's own directory, so it works both standalone and when require'd from the orchestrator.
-- Bootstrap: load `duffle_paths.lua` via `debug.getinfo(1, "S").source` (works both standalone + when require'd).
-- duffle_paths.lua sets package.path then returns `require("duffle")` at the bottom, so the dofile value IS the duffle module.
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
@@ -30,62 +32,75 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
-- Atom component declaration identifiers.
--- @type string
local ATOM_COMP_PROC = "MipsAtomComp_Proc_"
--- @type string
local MIPS_ATOM = "Slice_MipsCode" -- prefix on the function declaration that wraps an AtomComp_Proc_
-- Component-name prefixes.
--- @type string
local AC_PREFIX = "ac_" -- arg to MipsAtomComp_(ac_X); the X is the atom name
--- @type integer
local AC_PREFIX_LEN = 3
--- @type string
local MAC_PREFIX = "mac_" -- prefix on generated macros; the rest is the atom name
--- @type integer
local MAC_PREFIX_LEN = 4
-- ASCII byte values used in tokenization.
--- @type integer
local BYTE_NEWLINE = 10
--- @type integer
local BYTE_SLASH = 47
-- Output gen subdirectory + filename (per-directory aggregation; the directory name is the namespace).
--- @type string
local GEN_SUBDIR = "gen"
--- @type string
local MACS_FILENAME = "macs.h"
-- ════════════════════════════════════════════════════════════════════════════
-- Type declarations
-- ════════════════════════════════════════════════════════════════════════════
--- @class SourceFile
--- @field path string -- Absolute path to the source file
--- @field text string -- Full source text
--- @field dir string -- Directory containing the source
--- @field basename string -- Filename without extension
--- @field scan table -- Pre-scanned SourceScan payload (from duffle.scan_source)
--- @class PassCtx
--- @field sources SourceFile[] -- All source files in the build
--- @field metadata_path string -- Path to word_count.metadata.h
--- @field shared table -- Cross-pass shared state
--- @field out_root string -- Output root (e.g. "build/gen")
--- @field project_root string -- Project root (e.g. "code/")
--- @field upstream table<string, table> -- Per-pass upstream outputs
--- @field flags table -- CLI flags
--- @field verbose boolean -- Log diagnostic info
--- @class PassResult
--- @field outputs table[] -- {kind=, path=} entries describing emit files
--- @field errors table[] -- {line=, msg=} entries; build-stops
--- @field warnings table[] -- {line=, msg=} entries; build-succeeds
-- SourceFile, PassCtx, PassResult: see ps1_meta.lua
-- DuffleExport: see duffle.lua
-- SourceScan, AtomEntry, CorpusCollision, CollisionSite: see scan_source.lua
-- BodyToken: see emission_model.lua
-- WordCounts: see word_count_eval.lua
-- ComponentBodyEntry: see duffle_emit.lua
-- InstructionRow, GteCommandRow: see duffle_isa.lua
--- @class Component
--- @field name string -- Atom name (without `ac_` prefix)
--- @field body string -- Brace-delimited body (without the braces)
--- @field args string|nil -- Function-args string (function form only)
--- @field line integer -- Source line of the declaration
--- @field comment string|nil -- Scanner-owned `declaration_comment`; the components pass reads it from the scanner record
--- @field kind string -- "comp_bare" | "comp_proc" (atom_proc is NOT a component — see `project_components`)
--- @field debug_skip boolean -- Mirror of `a.debug_skip` (scanner-owned); true iff a bare `atom_dbg_skip` marker immediately preceded the declaration
--- @field name string -- Atom name (without `ac_` prefix)
--- @field body string -- Brace-delimited body (without the braces)
--- @field body_off integer|nil -- Byte offset of body[1] in source
--- @field body_tokens BodyToken[]|nil
--- @field args string|nil -- Function-args string (function form only)
--- @field arg_names string[]|nil -- Formal names with leading `ab` dropped
--- @field line integer -- Source line of the declaration
--- @field comment string|nil -- Scanner-owned `declaration_comment`; the components pass reads it from the scanner record
--- @field kind string -- "comp_bare" | "comp_proc" (atom_proc is NOT a component — see `project_components`)
--- @field debug_skip boolean -- Mirror of `a.debug_skip` (scanner-owned); true iff a bare `atom_dbg_skip` marker immediately preceded the declaration
--- @class ComponentMeta
--- @field cycle_cost integer
--- @field gp0_contrib integer
--- @class ComponentMetaMap
--- @field [string] ComponentMeta -- bag: bare component name -> meta
--- @class MacsOutput
--- @field macs_h string
--- @class ComponentsPass
--- @field run fun(ctx: PassCtx): PassResult
-- ════════════════════════════════════════════════════════════════════════════
-- Local helpers (file I/O + path normalization)
-- ════════════════════════════════════════════════════════════════════════════
--- @type ComponentsPass
local M = {}
-- ════════════════════════════════════════════════════════════════════════════
@@ -108,6 +123,7 @@ local M = {}
--- @param before_pos integer
--- @return string|nil
local function find_function_args_for(source, name, before_pos)
--- @type string|nil, string|nil
local _, args_inner = duffle.find_function_decl_for(source, before_pos, #MIPS_ATOM)
return args_inner
end
@@ -124,9 +140,13 @@ end
--- @return string[]|nil
local function extract_arg_names(args_str)
if not args_str or args_str == "" then return nil end
--- @type string[]
local names = {}
--- @type string[]
local tokens = duffle.split_top_level_commas(args_str)
--- @type integer, string
for _, tok in ipairs(tokens) do
--- @type string
local trimmed = duffle.trim(tok)
if trimmed ~= "" then
-- Strip trailing block comment (/* ... */) from the token, if present.
@@ -134,13 +154,16 @@ local function extract_arg_names(args_str)
-- not block comments embedded WITHIN a token between a parameter and a trailing comma.
-- Without this strip, the identifier-walk below stops at the `/` of `*/` and returns
-- the wrong name (or nothing). See `test_extract_arg_names_handles_trailing_block_comments`.
--- @type integer
local trimmed_end = #trimmed
if trimmed_end >= 2 and trimmed:sub(trimmed_end - 1, trimmed_end) == "*/" then
-- Find the matching `/*` that opens the trailing comment.
-- Walk back from the `*/` looking for `/*` (whitespace + `/*`).
--- @type integer
local close_pos = trimmed_end - 1 -- position of the second-to-last char
-- Walk back: skip trailing whitespace, then look for the `/*` opener.
while close_pos > 1 do
--- @type string
local ch = trimmed:sub(close_pos, close_pos)
if ch == " " or ch == "\t" or ch == "\n" or ch == "\r" then
close_pos = close_pos - 1
@@ -149,7 +172,9 @@ local function extract_arg_names(args_str)
end
end
-- Now scan back from close_pos for the `/*` opener (slashes are at close_pos-1 and close_pos-2).
--- @type integer|nil
local opener_pos = nil
--- @type integer
local scan = close_pos - 3
while scan >= 1 do
if trimmed:sub(scan, scan + 1) == "/*" then
@@ -169,8 +194,10 @@ local function extract_arg_names(args_str)
trimmed_end = #trimmed
if trimmed_end >= 4 and trimmed:sub(trimmed_end, trimmed_end) == "]" then
-- Walk back: skip digits, expect `[`.
--- @type integer
local bracket_pos = trimmed_end - 1
while bracket_pos > 1 do
--- @type string
local ch = trimmed:sub(bracket_pos, bracket_pos)
if ch >= "0" and ch <= "9" then
bracket_pos = bracket_pos - 1
@@ -185,8 +212,10 @@ local function extract_arg_names(args_str)
if trimmed == "" then goto continue end
-- Find the identifier at the end: walk back over trailers (whitespace + `*` + `[]`),
-- then walk back over the identifier chars (alnum + `_`).
--- @type integer
local ident_end = #trimmed
while ident_end > 0 do
--- @type string
local ch = trimmed:sub(ident_end, ident_end)
if ch == " " or ch == "\t" or ch == "*" or ch == "]" or ch == "[" then
ident_end = ident_end - 1
@@ -194,8 +223,10 @@ local function extract_arg_names(args_str)
break
end
end
--- @type integer
local ident_start = ident_end
while ident_start > 0 do
--- @type string
local ch = trimmed:sub(ident_start, ident_start)
if duffle.is_alnum_byte(string.byte(ch)) or ch == "_" then
ident_start = ident_start - 1
@@ -204,6 +235,7 @@ local function extract_arg_names(args_str)
end
end
ident_start = ident_start + 1
--- @type string
local name = trimmed:sub(ident_start, ident_end)
if name ~= "" then names[#names + 1] = name end
::continue::
@@ -213,7 +245,10 @@ local function extract_arg_names(args_str)
return names
end
--- @param args_str string|nil
--- @return string[]|nil
local function formal_arg_names(args_str)
--- @type string[]|nil
local names = extract_arg_names(args_str)
if not names then return nil end
if names[1] == "ab" then table.remove(names, 1) end
@@ -233,10 +268,12 @@ end
--- Carries the scanner-owned `debug_skip` flag forward so the generated projection can emit `/* atom_dbg_skip */`
--- before the authored comment and so `update_canonical_components` can mirror the same field onto `corpus.components[name]`.
--- @param source string -- the full source text (needed for backward lookups)
--- @param scan table -- SourceScan from duffle.scan_source
--- @param scan SourceScan
--- @return Component[]
local function project_components(source, scan)
--- @type Component[]
local out = {}
--- @type integer, AtomEntry
for _, a in ipairs(scan.atoms) do
-- Only `MipsAtomComp_(ac_X)` (kind="comp_bare") and `MipsAtomComp_Proc_(ac_X, ...)` (kind="comp_proc")
-- are COMPONENTS — they get inlined via `mac_<name>` aliases inside atom bodies.
@@ -248,9 +285,11 @@ local function project_components(source, scan)
-- Function-args lookup is meaningful for `MipsAtomComp_Proc_` components
-- (the macro sits inside `FI_ Slice_MipsCode ac_X(...)`); the alias expansion
-- discards the `ab` (atom-builder) arg the same way both forms do.
--- @type string|nil
local args = find_function_args_for(source, a.raw_name, a.ident_pos)
-- Comment ownership: scan_source.lua stamps `declaration_comment` on the record by walking backward past any associated bare marker.
-- The pass reads `declaration_comment` directly.
--- @type string
local comment = a.declaration_comment or ""
out[#out + 1] = {
line = a.line,
@@ -282,22 +321,30 @@ end
--- @param s string
--- @return string
local function convert_line_comments_to_block(s)
--- @type string
local result = s
--- @type integer
local pos = 1
--- @type integer
local len = #result
while pos <= len do
--- @type boolean
local is_double_slash = result:byte(pos) == BYTE_SLASH
and pos + 1 <= len and result:byte(pos + 1) == BYTE_SLASH
if not is_double_slash then
pos = pos + 1
else
-- Find end of line.
--- @type integer
local eol = pos
while eol <= len and result:byte(eol) ~= BYTE_NEWLINE do
eol = eol + 1
end
--- @type string
local before = result:sub(1, pos - 1)
--- @type string
local comment = result:sub(pos + 2, eol - 1) -- skip the `//`
--- @type string
local after
if eol <= len and result:byte(eol) == BYTE_NEWLINE then
after = " */" .. result:sub(eol) -- keep the newline
@@ -335,10 +382,13 @@ end
--- @param tok string
--- @return string
local function strip_leading_delay_marker(tok)
--- @type string|nil
local ident = duffle.read_ident(tok, 1)
if not ident or not duffle.DELAY_MARKERS[ident] then return tok end
--- @type string
local rest = tok:sub(#ident + 1):match("^%s*(.*)$") or ""
while rest:sub(1, 2) == "/*" do
--- @type integer|nil
local close = rest:find("*/", 3, true)
if not close then return "" end
rest = rest:sub(close + 2):match("^%s*(.*)$") or ""
@@ -350,22 +400,29 @@ end
--- in a single source's `count_all_components` pass; the in-progress -1 sentinel detects cycles (A -> B -> A).
--- @param name string -- the component name (without `mac_`)
--- @param comp_by_name table<string, Component>
--- @param wc table<string, integer>
--- @param cache table<string, integer>
--- @param wc WordCounts
--- @param cache table<string, integer> -- bag: name -> count; -1 in-progress sentinel
--- @return integer
local function word_count_rec(name, comp_by_name, wc, cache)
if cache[name] ~= nil then return cache[name] end
cache[name] = -1 -- mark in-progress (cycle detection)
--- @type Component|nil
local cc = comp_by_name[name]
--- @type integer
local n
if cc then
n = 0
--- @type BodyToken[]
local tokens = cc.body_tokens
--- @type integer, BodyToken
for _, t in ipairs(tokens) do
--- @type string
local trimmed = t.tok
if trimmed ~= "" then
--- @type string
local work = trimmed
while true do
--- @type string|nil
local marker = duffle.read_ident(work, 1)
if marker and duffle.DELAY_MARKERS[marker] then
work = strip_leading_delay_marker(work)
@@ -375,6 +432,7 @@ local function word_count_rec(name, comp_by_name, wc, cache)
end
end
if work ~= "" then
--- @type string|nil
local lookup = strip_mac_prefix(duffle.read_ident(work, 1))
if lookup == "atom_label" or lookup == "atom_offset" then
-- Pure metaprogram anchors; emit zero words.
@@ -405,13 +463,18 @@ end
--- references hit memoized values instead of re-walking the body.
--- Cycle detection (A -> B -> A) is preserved via the in-progress `-1` sentinel in `cache`.
--- @param components Component[]
--- @param wc table<string, integer>
--- @return table<string, integer> -- map of component name (without `mac_`) -> word count
--- @param wc WordCounts
--- @return table<string, integer> -- bag: bare component name -> word count
local function count_all_components(components, wc)
--- @type table<string, Component>
local comp_by_name = {}
--- @type integer, Component
for _, cc in ipairs(components) do comp_by_name[cc.name] = cc end
--- @type table<string, integer> -- bag: memo; -1 in-progress sentinel
local cache = {}
--- @type table<string, integer> -- bag: bare name -> word count
local counts = {}
--- @type integer, Component
for _, c in ipairs(components) do
counts[c.name] = word_count_rec(c.name, comp_by_name, wc, cache)
end
@@ -435,28 +498,39 @@ end
--- Missing component: cycle 1, gp0 0.
--- @param name string -- component bare name (e.g. "yield", "pack_color_word")
--- @param comp_by_name table<string, Component>
--- @param latency table<string, integer>
--- @param cache table<string, {cycle_cost=integer, gp0_contrib=integer}>
--- @return {cycle_cost=integer, gp0_contrib=integer}
--- @param latency table<string, integer> -- bag: ident -> cycle cost
--- @param cache ComponentMetaMap
--- @return ComponentMeta
local function component_meta_rec(name, comp_by_name, latency, cache)
if cache[name] ~= nil then return cache[name] end
cache[name] = { cycle_cost = -1, gp0_contrib = -1 }
--- @type Component|nil
local cc = comp_by_name[name]
--- @type integer
local cycle_cost
--- @type integer
local gp0_contrib
if cc then
--- @type boolean
local skip_cycle = (name == "yield")
--- @type boolean
local skip_gp0 = name:match("^insert_ot_tag") ~= nil
cycle_cost = 0
gp0_contrib = 0
if not skip_cycle or not skip_gp0 then
--- @type BodyToken[]
local tokens = cc.body_tokens
--- @type integer, BodyToken
for _, t in ipairs(tokens) do
--- @type string
local trimmed = t.tok
if trimmed ~= "" then
--- @type string|nil
local ident = duffle.read_ident(trimmed, 1)
if ident and ident:sub(1, MAC_PREFIX_LEN) == MAC_PREFIX then
--- @type string
local nested = ident:sub(MAC_PREFIX_LEN + 1)
--- @type ComponentMeta
local nested_meta = component_meta_rec(nested, comp_by_name, latency, cache)
if not skip_cycle then
cycle_cost = cycle_cost + nested_meta.cycle_cost
@@ -466,7 +540,9 @@ local function component_meta_rec(name, comp_by_name, latency, cache)
end
else
if not skip_cycle then
--- @type InstructionRow|nil
local isa = duffle.instr(ident)
--- @type GteCommandRow|nil
local gte = duffle.gte(ident)
cycle_cost = cycle_cost + ((isa and isa.cycles) or (gte and gte.cycles) or latency[ident] or 1)
end
@@ -499,13 +575,18 @@ end
--- Compute `cycle_cost` + `gp0_contrib` for every component in `components` in a single pass.
--- One memoization cache; a nested `mac_Y` inside a `mac_X` body computes both fields once.
--- @param components Component[]
--- @param latency table<string, integer>
--- @return table<string, {cycle_cost=integer, gp0_contrib=integer}>
--- @param latency table<string, integer> -- bag: ident -> cycle cost
--- @return ComponentMetaMap
local function compute_components_metadata(components, latency)
--- @type table<string, Component>
local comp_by_name = {}
--- @type integer, Component
for _, cc in ipairs(components) do comp_by_name[cc.name] = cc end
--- @type ComponentMetaMap
local cache = {}
--- @type ComponentMetaMap
local out = {}
--- @type integer, Component
for _, c in ipairs(components) do
out[c.name] = component_meta_rec(c.name, comp_by_name, latency, cache)
end
@@ -521,10 +602,14 @@ end
--- @param s string
--- @return string[]
local function split_comment_lines(s)
--- @type string[]
local out = {}
--- @type integer
local pos = 1
--- @type integer
local s_len = #s
while pos <= s_len do
--- @type integer|nil
local nl = s:find("\n", pos, true)
if not nl then
out[#out + 1] = s:sub(pos)
@@ -544,6 +629,7 @@ end
--- @param args_str string|nil
--- @return string
local function signature_from_args(args_str)
--- @type string[]|nil
local names = formal_arg_names(args_str)
if names then
return table.concat(names, ", ")
@@ -553,7 +639,10 @@ end
--- Strip the trailing `" \"` (space + backslash) line continuation from the last body line.
--- The last 2 chars are always that pair.
--- @param lines string[]
--- @return nil
local function strip_trailing_continuation(lines)
--- @type string
local last = lines[#lines]
if last:sub(-2) == " \\" then
lines[#lines] = last:sub(1, -3)
@@ -580,12 +669,15 @@ end
--- @param tok string -- a single token from split_top_level_commas (already trimmed at the start, may contain trailing whitespace + block comment)
--- @return boolean
local function is_pure_delay_marker_token(tok)
--- @type table<string, boolean> -- bag: delay-marker ident -> true
local markers = duffle.DELAY_MARKERS
if type(markers) ~= "table" then return false end
-- Identify a leading delay-marker identifier (e.g. `GteDelay_`).
--- @type integer
local ident_end = 1
while ident_end <= #tok do
--- @type string
local ch = tok:sub(ident_end, ident_end)
if ch:match("[%w_]") then
ident_end = ident_end + 1
@@ -593,16 +685,20 @@ local function is_pure_delay_marker_token(tok)
break
end
end
--- @type string
local ident = tok:sub(1, ident_end - 1)
if not markers[ident] then return false end
-- Walk the remainder: only whitespace and block comments are allowed.
--- @type integer
local scan = ident_end
while scan <= #tok do
--- @type string
local ch = tok:sub(scan, scan)
if ch:match("%s") then
scan = scan + 1
elseif ch == "/" and tok:sub(scan + 1, scan + 1) == "*" then
--- @type integer|nil
local close = tok:find("*/", scan + 2, true)
if not close then return false end
scan = close + 2
@@ -641,14 +737,22 @@ end
--- the annotation IS preserved in the generated header
--- (so the comment + marker remain visible to anyone reading `gen/macs.h`), but the C preprocessor expands the marker to empty, so leaving the `,`
--- separator out is what stops the `,,` syntax error. See `token_skips_leading_comma` for the contract.
--- @param lines string[]
--- @param c Component
--- @param sig string
--- @param tokens string[]
--- @return nil
local function emit_macro_body(lines, c, sig, tokens)
--- @type integer
for tok_idx = 1, #tokens do
tokens[tok_idx] = convert_line_comments_to_block(tokens[tok_idx])
end
if #tokens == 0 then return end
lines[#lines + 1] = "#define mac_" .. c.name .. "(" .. sig .. ") \\"
lines[#lines + 1] = "\t" .. tokens[1] .. " \\"
--- @type integer
for tok_idx = 2, #tokens do
--- @type string
local sep = token_skips_leading_comma(tokens[tok_idx]) and "\t" or ",\t"
lines[#lines + 1] = sep .. tokens[tok_idx] .. " \\"
end
@@ -660,11 +764,11 @@ end
--- For skipped components, a `/* atom_dbg_skip */` marker comment is emitted immediately before the authored comment block.
--- The marker is a single line, the comment comes next, and the `#define` line follows. The `debug_skip` stamp is scanner-owned
--- (`a.debug_skip == true` on the declaration record); the components pass projects it directly.
--- @param c Component
--- @param components Component[]
--- @param wc table<string, integer>
--- @param c Component
--- @param counts table<string, integer> -- bag: bare component name -> word count
--- @return string[] -- list of lines for this component
local function build_component_lines(c, counts)
--- @type string[]
local lines = {}
-- Marker comment: emitted once for every skipped component.
@@ -675,15 +779,20 @@ local function build_component_lines(c, counts)
end
if c.comment and c.comment ~= "" then
--- @type integer, string
for _, line in ipairs(split_comment_lines(c.comment)) do
lines[#lines + 1] = line
end
end
--- @type string[]
local tokens = duffle.split_top_level_commas(c.body)
--- @type integer
for i = 1, #tokens do tokens[i] = duffle.trim(tokens[i]) end
--- @type string
local sig = signature_from_args(c.args)
-- Direct lookup against the per-source precomputed `counts` table (built once by count_all_components).
--- @type integer
local n = counts[c.name]
if n > 0 then
@@ -707,10 +816,13 @@ end
--- @param sources SourceFile[] -- Sources contributing to this directory (for the header comment)
--- @return string[]
local function header_boilerplate(dir, sources)
--- @type string[]
local source_lines = { "// Directory: " .. duffle.to_absolute_path(dir) .. "/" }
--- @type integer, SourceFile
for _, src in ipairs(sources) do
source_lines[#source_lines + 1] = "// source: " .. duffle.to_absolute_path(src.path)
end
--- @type string
local source_blob = table.concat(source_lines, "\n")
return {
-- #pragma once wrapped in #ifdef INTELLISENSE_DIRECTIVES, matching the convention in lottes_tape.h.
@@ -739,7 +851,9 @@ end
--- @return string -- Output directory
--- @return string -- Full output path
local function compute_macs_h_path(dir)
--- @type string
local out_dir = dir .. "/" .. GEN_SUBDIR
--- @type string
local out_path = out_dir .. "/" .. MACS_FILENAME
return out_dir, out_path
end
@@ -750,19 +864,24 @@ end
--- @param dir string -- Absolute source directory
--- @param sources SourceFile[] -- Sources contributing to this directory (for the header comment)
--- @param components Component[] -- Aggregated components from all sources in this directory
--- @param counts table<string, integer> -- Precomputed word counts (from count_all_components)
--- @param counts table<string, integer> -- bag: bare component name -> word count
--- @return string|nil -- Path to the written file (nil if no components)
local function emit_component_macros_h(ctx, dir, sources, components, counts)
if #components == 0 then return nil end
--- @type string, string
local out_dir, out_path = compute_macs_h_path(dir)
--- @type string[]
local lines = header_boilerplate(dir, sources)
--- @type integer, Component
for _, c in ipairs(components) do
--- @type integer, string
for _, l in ipairs(build_component_lines(c, counts)) do
lines[#lines + 1] = l
end
end
--- @type string
local content = table.concat(lines, "\n") .. "\n"
duffle.ensure_dir(out_dir)
duffle.write_file_lf(out_path, content)
@@ -776,12 +895,16 @@ end
--- (internal) Extend `corpus.word_counts` with this source's component macros so offsets sees them without re-reading the file.
--- First declaration wins: a later caller's count is dropped (the existing entry from the first source is preserved).
--- @param corpus table -- the corpus
--- @param corpus Corpus
--- @param components Component[]
--- @param counts table<string, integer> -- precomputed word counts (from count_all_components)
--- @param counts table<string, integer> -- bag: bare component name -> word count
--- @return nil
local function update_canonical_word_counts(corpus, components, counts)
--- @type WordCounts
local wc = corpus.word_counts
--- @type integer, Component
for _, c in ipairs(components) do
--- @type string
local key = "mac_" .. c.name
if wc[key] == nil then
wc[key] = counts[c.name]
@@ -790,27 +913,33 @@ local function update_canonical_word_counts(corpus, components, counts)
end
--- @class ComponentDef
--- @field name string -- Bare name (without ac_/mac_ prefix)
--- @field line integer -- Definition source line (line of `MipsAtomComp_(ac_X)` / `MipsAtomComp_Proc_(ac_X, ...)`)
--- @field path string -- Absolute source path of the definition
--- @field kind string -- "comp_bare" | "comp_proc" (atom_proc is NOT a component)
--- @field debug_skip boolean -- Mirror of the scanner-owned `a.debug_skip`; consumers read this directly
--- @field name string -- Bare name (without ac_/mac_ prefix)
--- @field line integer -- Definition source line (line of `MipsAtomComp_(ac_X)` / `MipsAtomComp_Proc_(ac_X, ...)`)
--- @field path string -- Absolute source path of the definition
--- @field kind string -- "comp_bare" | "comp_proc" (atom_proc is NOT a component)
--- @field debug_skip boolean -- Mirror of the scanner-owned `a.debug_skip`; consumers read this directly
--- @field cycle_cost integer|nil -- From metadata[c.name]; nil when the body was not costed
--- @field gp0_contrib integer|nil -- From metadata[c.name]; nil when the body was not costed
--- (internal) Populate `corpus.components` with this source's components-by-name map.
--- First declaration wins; later declarations of the same bare name are dropped and recorded as a collision via `corpus.collisions` (kind = "component").
--- The pass does NOT write to `ctx.shared.components`.
--- No parallel skip map is built here; consumers that need the per-component skip state read `corpus.components[name].debug_skip` directly.
--- The `cycle_cost` + `gp0_contrib` fields are populated from `metadata[c.name]` (computed by `compute_components_metadata` against the original `MipsAtomComp_` body).
--- @param corpus table -- the corpus
--- @param corpus Corpus
--- @param src SourceFile
--- @param components Component[]
--- @param metadata table<string, {cycle_cost=integer, gp0_contrib=integer}>
--- @param metadata ComponentMetaMap
--- @return nil
local function update_canonical_components(corpus, src, components, metadata)
--- @type string
local rel_path = src.path:gsub("\\", "/")
--- @type integer, Component
for _, c in ipairs(components) do
-- Keyed by bare name (e.g. `yield`, `load_tri_indices`).
-- The atoms_source_map pass looks up components by bare name from the corpus;
-- `mac_` prefix lives at the call-site identifier and is stripped before lookup.
--- @type ComponentMeta|nil
local m = metadata and metadata[c.name] or nil
if corpus.components[c.name] == nil then
corpus.components[c.name] = {
@@ -825,9 +954,12 @@ local function update_canonical_components(corpus, src, components, metadata)
else
-- A second declaration of the same bare name: record a typed collision so static-analysis + the report can surface it.
-- Identical-shape declarations (same path + line) reuse the first-wins entry without a collision record.
--- @type ComponentDef
local existing = corpus.components[c.name]
if existing.path ~= rel_path or existing.line ~= c.line then
--- @type string
local kind = c.kind or "comp_bare"
--- @type string
local first_kind = existing.kind or "comp_bare"
corpus.collisions[#corpus.collisions + 1] = {
kind = "component",
@@ -845,12 +977,15 @@ end
--- (internal) Populate `corpus.component_body_index` with this source's body index entries.
--- First declaration wins; later declarations are dropped (no separate collision record: the components collision is already surfaced by `update_canonical_components`).
--- The pass writes to `corpus.component_body_index` only (the corpus owns this projection).
--- @param corpus table -- the corpus
--- @param corpus Corpus
--- @param src SourceFile
--- @param components Component[]
--- @param scan table -- the SourceScan payload (for line_of)
--- @param scan SourceScan
--- @return nil
local function update_canonical_component_body_index(corpus, src, components, scan)
--- @type (fun(pos: integer): integer)|nil
local line_of = scan and scan.line_of
--- @type integer, Component
for _, c in ipairs(components) do
if corpus.component_body_index[c.name] == nil then
corpus.component_body_index[c.name] = {
@@ -869,11 +1004,15 @@ end
--- @param ctx PassCtx
--- @return PassResult
function M.run(ctx)
--- @type MacsOutput[]
local outputs = {}
--- @type PassFinding[]
local errors = {}
--- @type PassFinding[]
local warnings = {}
-- Corpus ownership gate.
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" then
error("components.run requires ctx.shared.corpus.", 0)
@@ -895,14 +1034,21 @@ function M.run(ctx)
-- Per-directory aggregation: every source in the same directory contributes to one `gen/macs.h`.
-- The directory itself is the namespace. `corpus.sources_by_dir` preserves source-order within each bucket (matches `corpus.source_order`).
--- @type table<string, SourceFile[]>
local sources_by_dir = corpus.sources_by_dir or duffle.group_sources_by_dir(corpus.source_order)
--- @type string, SourceFile[]
for dir, sources in pairs(sources_by_dir) do
-- Aggregate components from every source in this directory.
-- `project_components` returns nil for sources with no `MipsAtomComp_` declarations; we skip those.
--- @type Component[]
local aggregated_components = {}
--- @type table<SourceFile, ComponentMetaMap>
local metadata_per_source = {}
--- @type integer, SourceFile
for _, src in ipairs(sources) do
--- @type Component[]
local per_source = project_components(src.text, src.scan) or {}
--- @type integer, Component
for _, c in ipairs(per_source) do
aggregated_components[#aggregated_components + 1] = c
end
@@ -913,13 +1059,17 @@ function M.run(ctx)
if #aggregated_components > 0 then
-- Compute word counts across the aggregated set. `corpus.word_counts` carries the
-- same-source + prior-directory entries so the recursive lookup sees both.
--- @type table<string, integer> -- bag: bare name -> word count
local counts = count_all_components(aggregated_components, corpus.word_counts)
--- @type string|nil
local macs_path = emit_component_macros_h(ctx, dir, sources, aggregated_components, counts)
if macs_path then
outputs[#outputs + 1] = { macs_h = macs_path }
-- Populate the projections AFTER disk emission (byte-identical `.macs.h` contract).
update_canonical_word_counts(corpus, aggregated_components, counts)
--- @type integer, SourceFile
for _, src in ipairs(sources) do
--- @type Component[]
local per_source = project_components(src.text, src.scan) or {}
if #per_source > 0 then
update_canonical_components(corpus, src, per_source, metadata_per_source[src])
scripts/passes/dwarf_injection.lua
+707 -46
View File
File diff suppressed because it is too large Load Diff
scripts/passes/emission_model.lua
+153
View File
@@ -26,12 +26,106 @@
---
--- `passes.scan_source` strips its private `_code_macros` / `_code_macro_bodies` tables before this pass runs.
--- @class BodyToken
--- @field tok string
--- @field rel integer
--- @class EmissionItem
--- @field kind string
--- @field encoder string|nil
--- @field args string[]|nil
--- @field i integer|nil
--- @field word_count integer|nil
--- @field line integer|nil
--- @field call_text string|nil
--- @field root_call_text string|nil
--- @field invocation_ids integer[]|nil
--- @field outermost_invocation_id integer|nil
--- @field gpr_keys string[]|nil
--- @field ident string|nil
--- @field isa_kind string|nil
--- @field nop_words integer|nil
--- @field is_yield boolean|nil
--- @field is_load boolean|nil
--- @field is_branch boolean|nil
--- @field is_unconditional_jump boolean|nil
--- @field is_terminal_jump boolean|nil
--- @field gp0_shape string|nil
--- @field name string|nil
--- @field target string|nil
--- @field word_index integer|nil
--- @field consuming_encoder string|nil
--- @field consuming_arg_pos integer|nil
--- @field invocation_id integer|nil
--- @class WordEvent
--- @field i integer
--- @field encoder string
--- @field args string[]
--- @field def_path string
--- @field def_line integer
--- @field call_text string|nil
--- @field root_call_text string|nil
--- @field invocation_ids integer[]
--- @field outermost_invocation_id integer
--- @field word_count integer
--- @field gpr_keys string[]|nil
--- @field ident string
--- @field kind string
--- @field nop_words integer
--- @field is_yield boolean
--- @field is_load boolean
--- @field is_branch boolean
--- @field is_unconditional_jump boolean
--- @field is_terminal_jump boolean
--- @field gp0_shape string|nil
--- @field body_line integer|nil
--- @field call_line integer|nil
--- @field call_path string|nil
--- @class EmissionMarker
--- @field kind string
--- @field name string
--- @field line integer
--- @field word_index integer
--- @field target string|nil
--- @field consuming_encoder string|nil
--- @field consuming_arg_pos integer|nil
--- @class EmitError
--- @field kind string
--- @field line integer|nil
--- @field msg string
--- @field source string|nil
--- @field schema_name string|nil
--- @class EmitWarning
--- @field kind string
--- @field line integer|nil
--- @field msg string
--- @class AtomPaths
--- @field tokens BodyToken[]
--- @field line_in_body table<integer, integer> -- bag: body byte offset -> 1-based line
--- @field items EmissionItem[]
--- @field word_events WordEvent[]
--- @field markers EmissionMarker[]
--- @field invocations InvocationRecord[]
--- @field errors EmitError[]
--- @field warnings EmitWarning[]
--- @class EmissionModelPass
--- @field run fun(ctx: PassCtx): PassResult
--- @type EmissionModelPass
local M = {}
-- ─────────────────────────────────────────────────────────────────────────
-- Bootstrap: load `duffle_paths.lua` via debug.getinfo so the module works standalone (run as `luajit passes/emission_model.lua`) and when require'd from the orchestrator.
-- ─────────────────────────────────────────────────────────────────────────
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ─────────────────────────────────────────────────────────────────────────
@@ -49,7 +143,13 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
--
-- After this function, every `inv.call_line` is physical. DWARF and provenance output read it directly.
-- The word-event loop forwards the already-physical `outer_inv.call_line` into `we.call_line` for words inside an invocation.
--- @param projection EmissionProjection
--- @param atom_record AtomEntry
--- @param src SourceFile
--- @param corpus Corpus
--- @return nil
local function stamp_root_provenance(projection, atom_record, src, corpus)
--- @type (fun(pos: integer): integer)|nil
local root_line_of = src.scan and src.scan.line_of
assert(type(root_line_of) == "function"
, "emission_model: src.scan.line_of is required (canonical LineIndex closure over the source text) to stamp physical provenance")
@@ -58,10 +158,14 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- `root_body_line` is the physical source line of the ATOM HEADER byte containing the opening `{`; that byte is one byte BEFORE `atom_record.body_off`.
-- The walker assigns line 2 to the body's first content line because line 1 is the trailing `\n` after `{`. Body-text line k therefore maps to `root_body_line + (k - 1)`.
-- `body_off - 1` points at the opening `{`, whose line index identifies the header line. `body_off` points after `{` and would shift every word row forward by one line.
--- @type integer
local root_body_line = root_line_of(atom_record.body_off - 1) or atom_record.line or 0
--- @type table<string, ComponentBodyEntry>
local component_index = corpus.component_body_index or {}
--- @type EmissionItem[]
local word_items = {}
--- @type integer, EmissionItem
for _, item in ipairs(projection.items) do
if item.kind == "word" then word_items[#word_items + 1] = item end
end
@@ -69,14 +173,21 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- Resolve one word's physical body line, where the byte containing that word appears in source.
-- * Component expansions carry `invocation_ids`; the component's full-file `line_of` leaves `item.line` physical.
-- * Raw tokens in the root atom body carry an empty `invocation_ids` list and a body-relative `item.line`; convert them here.
--- @param event WordEvent
--- @param item EmissionItem
--- @return integer
local function body_line_for(event, item)
--- @type integer[]
local ids = event.invocation_ids or {}
-- The innermost open invocation identifies which line index the walker used.
-- A component `line_of` makes `item.line` physical; the atom's `body_text` line index makes it body-relative.
if ids and #ids > 0 then
--- @type integer
local inner_id = ids[#ids]
--- @type InvocationRecord|nil
local inner_inv = inner_id and projection.invocations[inner_id]
if inner_inv then
--- @type ComponentBodyEntry|nil
local component = component_index[inner_inv.component_name]
if component and component.line_of then
-- Walker used `comp.line_of`, which is the source's physical LineIndex. item.line is already physical.
@@ -92,7 +203,9 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- Stamp the root source path onto invocation records whose `call_path` the walker left empty.
-- The walker passes `body_entry.source` to `emit_invoke_begin`; `M.project_emission` creates the root `body_entry` with source `""`, leaving its `call_path` empty.
-- This stamp gives every invocation a physical `call_path` matching `passes/atoms_source_map.lua`'s in-memory provenance projection.
--- @type string
local root_path = src.path or ""
--- @type integer, InvocationRecord
for _, inv in ipairs(projection.invocations) do
if inv.call_path == nil or inv.call_path == "" then
inv.call_path = root_path
@@ -102,6 +215,7 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- Normalize `inv.call_line` to a physical source line.
-- * ROOT invocations (`parent_id == 0`) carry body-relative `call_line` values from `M.LineIndex(body_text)`; convert them once with `root_body_line`.
-- * INNER invocations (`parent_id ~= 0`) carry physical `call_line` values from the component's `line_of`; retain them unchanged.
--- @type integer, InvocationRecord
for _, inv in ipairs(projection.invocations) do
if inv.parent_id == 0 then
inv.call_line = (root_body_line or 0) + (inv.call_line or 1) - 1
@@ -111,13 +225,20 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- Build `body_lines` for each invocation.
-- `atoms_source_map` and `dwarf_injection` read `inv.body_lines[k]` directly from the invocation record created here.
-- Component words already carry physical `item.line` values from the walker's COMPONENT line index, so `body_line_for` returns them unchanged.
--- @type integer, InvocationRecord
for _, inv in ipairs(projection.invocations) do
--- @type integer
local sw = inv.start_word
--- @type integer
local ew = inv.end_word
--- @type integer[]
local bls = {}
--- @type integer
for i = sw, ew do
--- @type EmissionItem|nil
local it = projection.items and projection.items[i]
if it and it.kind == "word" then
--- @type WordEvent
local fake_event = { invocation_ids = { inv.id } }
bls[#bls + 1] = body_line_for(fake_event, it) or 0
end
@@ -128,14 +249,20 @@ local function stamp_root_provenance(projection, atom_record, src, corpus)
-- Resolve each `word_event`'s physical `body_line` and `call_line`.
-- For words inside an invocation, `we.call_line` identifies the OUTER atom source line containing the `mac_X(...)` token that triggered expansion.
-- The root-invocation conversion above makes every `inv.call_line` physical; forward it directly and use each raw word's `body_line` as the fallback.
--- @type integer, WordEvent
for index, we in ipairs(projection.word_events) do
--- @type EmissionItem
local item = word_items[index] or {}
--- @type integer
local body_line = body_line_for(we, item)
item.line = body_line
we.body_line = body_line
--- @type integer
local call_line = body_line
--- @type integer
local outer_id = we.outermost_invocation_id or 0
--- @type InvocationRecord|nil
local outer_inv = projection.invocations[outer_id]
if outer_inv then
-- `outer_inv.call_line` is physical after the conversion loop above, so use it directly.
@@ -151,15 +278,24 @@ end
-- Project one atom record into `atom.paths`.
-- Mutates the atom record in-place and returns the projection (for pass-level error/warning accumulation).
--- @param atom_record AtomEntry
--- @param src SourceFile
--- @param corpus Corpus
--- @return EmissionProjection
local function project_atom(atom_record, src, corpus)
--- @type string
local body = atom_record.body or ""
--- @type WordCounts
local wc = corpus.word_counts or {}
--- @type table<string, ComponentBodyEntry>
local cbi = corpus.component_body_index or {}
--- @type RegUseSchema|nil
local schema = nil
if atom_record.reg_use_schema_name then
schema = corpus.reg_use_schemas and corpus.reg_use_schemas[atom_record.reg_use_schema_name]
end
-- That construction site stamps `invocation.debug_skip` while appending each record to `proj.invocations`.
--- @type EmissionProjection
local proj = duffle.project_emission(body, cbi, wc, corpus.components, {
reg_use_schema = schema,
reg_use_param = atom_record.reg_use_param_name,
@@ -172,11 +308,13 @@ local function project_atom(atom_record, src, corpus)
msg = string.format("RegUse schema %q is missing", atom_record.reg_use_schema_name),
}
end
--- @type integer, EmitError
for _, err in ipairs(corpus.reg_use_errors or {}) do
if err.schema_name == atom_record.reg_use_schema_name then
proj.errors[#proj.errors + 1] = err
end
end
--- @type AtomPaths
local paths = {
tokens = atom_record.body_tokens or {},
line_in_body = duffle.build_body_line_index(body),
@@ -199,23 +337,33 @@ end
--- @param ctx PassCtx -- { shared = { corpus = ... }, out_root, ... }
--- @return PassResult
function M.run(ctx)
--- @type PassOutputEntry[]
local outputs = {}
--- @type EmitError[]
local errors = {}
--- @type EmitWarning[]
local warnings = {}
--- @type Corpus|nil
local corpus = ctx and ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" then error("emission_model: ctx.shared.corpus is required (canonical projection)", 0) end
if type(corpus.source_order) ~= "table" then error("emission_model: ctx.shared.corpus.source_order is required", 0) end
-- Project once, collect errors + warnings for one atom.
-- Kind must be one of: atom | atom_proc | raw_atom | comp_bare | comp_proc.
--- @param atom AtomEntry
--- @param src SourceFile
--- @return nil
local function process_atom(atom, src)
if not (atom and atom.body) then return end
--- @type string
local kind = atom.kind
if kind ~= "atom" and kind ~= "atom_proc" and kind ~= "raw_atom" and kind ~= "comp_bare" and kind ~= "comp_proc" then
return
end
--- @type EmissionProjection
local proj = project_atom(atom, src, corpus)
--- @type integer, EmitError
for _, e in ipairs(proj.errors) do
-- Preserve `kind` (cycle / count_mismatch / unbalanced) so readers dispatch on the diagnostic class and leave the message string as display text.
errors[#errors + 1] = {
@@ -225,6 +373,7 @@ function M.run(ctx)
source = e.source or src.path,
}
end
--- @type integer, EmitWarning
for _, w in ipairs(proj.warnings) do
warnings[#warnings + 1] = {
kind = w.kind,
@@ -237,11 +386,15 @@ function M.run(ctx)
-- Walk `corpus.source_order`; within each source, visit atoms followed by raw_atoms.
-- Recognized kinds (atom | atom_proc | raw_atom | comp_bare | comp_proc) each receive the atom.paths projection via duffle.project_emission.
-- Components are macros inlined into atom bodies; focused tests and isolated component analyses consume atom.paths directly.
--- @type integer, SourceFile
for _, src in ipairs(corpus.source_order) do
--- @type SourceScan
local scan = src.scan or {}
--- @type integer, AtomEntry
for _, atom in ipairs(scan.atoms or {}) do
process_atom(atom, src)
end
--- @type integer, AtomEntry
for _, atom in ipairs(scan.raw_atoms or {}) do
process_atom(atom, src)
end
scripts/passes/offsets.lua
+88 -29
View File
@@ -20,7 +20,9 @@
-- Uses `debug.getinfo` to find this file's own directory, so it works both standalone and when require'd from the orchestrator.
-- Bootstrap: load `duffle_paths.lua` via `debug.getinfo(1, "S").source` (works both standalone + when require'd).
-- duffle_paths.lua sets package.path then returns `require("duffle")` at the bottom, so the dofile value IS the duffle module.
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
@@ -28,33 +30,20 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
-- Offset macro/enum naming prefixes (the emitted header uses these).
--- @type string
local OFFSET_MACRO_PREFIX = "_atom_offset_"
--- @type string
local OFFSET_ENUM_PREFIX = "atom_offset_"
-- Column width for the `#define _atom_offset_F_T = N` alignment.
--- @type integer
local OFFSET_MACRO_COL = 44
-- ════════════════════════════════════════════════════════════════════════════
-- Type declarations
-- ════════════════════════════════════════════════════════════════════════════
--- @class SourceFile
--- @field path string -- Absolute path to the source file
--- @field text string -- Full source text
--- @field dir string -- Directory containing the source
--- @field basename string -- Filename without extension
--- @field scan table -- Pre-scanned SourceScan payload (from duffle.scan_source)
--- @class PassCtx
--- @field shared table -- Cross-pass shared state
--- @field shared.corpus table -- Corpus projection
--- @field shared.word_counts table
--- @field out_root string -- Output root (e.g. "build/gen")
--- @class PassResult
--- @field outputs table[] -- {kind=, path=} entries describing emit files
--- @field errors table[] -- {line=, msg=} entries; build-stops
--- @field warnings table[] -- {line=, msg=} entries; build-succeeds
-- SourceFile, PassCtx, PassResult: see ps1_meta.lua
--- @class BranchOffset
--- @field tag string -- Marker tag (e.g. "F" in `atom_offset(F, T)`)
@@ -69,6 +58,32 @@ local OFFSET_MACRO_COL = 44
--- @field total_words integer -- Total word count of the atom body
--- @field offsets BranchOffset[] -- Per-branch offset list
--- @class OffsetBranch
--- @field tag string
--- @field target string
--- @field branch_word integer
--- @field consuming_encoder string|nil
--- @field consuming_arg_pos integer|nil
--- @field line integer|nil
--- @class MarkerProjectState
--- @field labels table<string, integer> -- bag: label name -> word index
--- @field branches OffsetBranch[]
--- @class OffsetConst
--- @field macro_name string
--- @field enum_name string
--- @field value integer
--- @class OffsetOutput
--- @field offsets_h string
--- @class OffsetsPass
--- @field run fun(ctx: PassCtx): PassResult
--- @class AtomEntry
--- @field paths AtomPaths|nil
-- ════════════════════════════════════════════════════════════════════════════
-- Canonical marker projection
-- ════════════════════════════════════════════════════════════════════════════
@@ -76,10 +91,17 @@ local OFFSET_MACRO_COL = 44
-- MARKER_PROJECTORS is the marker-kind data table.
-- The emission-model pass already records marker word positions + consuming-instruction context;
-- this pass only projects those records into the label/branch lookup shape needed by offset computation.
--- @type table<string, fun(state: MarkerProjectState, marker: EmissionMarker): nil>
local MARKER_PROJECTORS = {
--- @param state MarkerProjectState
--- @param marker EmissionMarker
--- @return nil
label = function(state, marker)
state.labels[marker.name] = marker.word_index
end,
--- @param state MarkerProjectState
--- @param marker EmissionMarker
--- @return nil
offset = function(state, marker)
state.branches[#state.branches + 1] = {
tag = marker.name,
@@ -93,11 +115,15 @@ local MARKER_PROJECTORS = {
--- Project canonical marker records into the two lookup tables used by the offset renderer.
--- No source text, body text, or body token is inspected.
--- @param markers table[] -- atom.paths.markers
--- @return table<string, integer>, table[]
--- @param markers EmissionMarker[]
--- @return table<string, integer>
--- @return OffsetBranch[]
local function project_markers(markers)
--- @type MarkerProjectState
local state = { labels = {}, branches = {} }
--- @type integer, EmissionMarker
for _, marker in ipairs(markers or {}) do
--- @type (fun(state: MarkerProjectState, marker: EmissionMarker): nil)|nil
local project = MARKER_PROJECTORS[marker.kind]
if project then project(state, marker) end
end
@@ -119,12 +145,15 @@ end
--- `jump_reg` / `call_reg` / `jump_link` -> ERROR. Register-form jumps have no offset field; `atom_offset` is invalid.
--- missing `consuming_encoder` -> ERROR. A lone top-level `atom_offset` is not a branch.
--- @param labels table<string, integer>
--- @param branches table[]
--- @param errors table[]
--- @param branches OffsetBranch[]
--- @param errors PassFinding[]
--- @return BranchOffset[]
local function compute_offsets(labels, branches, errors)
--- @type BranchOffset[]
local results = {}
--- @type integer, OffsetBranch
for _, br in ipairs(branches) do
--- @type integer|nil
local target = labels[br.target]
if not target then
errors[#errors + 1] = {
@@ -132,6 +161,7 @@ local function compute_offsets(labels, branches, errors)
msg = "Branch target '" .. br.target .. "' has no atom_label (at word " .. br.branch_word .. ")",
}
else
--- @type string|nil
local consuming = br.consuming_encoder
if consuming == nil or consuming == "" then
errors[#errors + 1] = {
@@ -171,7 +201,7 @@ end
--- (internal) Build a constant-table entry `{macro_name, enum_name, value}` from a BranchOffset.
--- @param bo BranchOffset
--- @return table
--- @return OffsetConst
local function make_offset_const(bo)
return {
macro_name = OFFSET_MACRO_PREFIX .. bo.tag .. "_" .. bo.target,
@@ -183,19 +213,24 @@ end
--- (internal) Emit one atom's offset constants + enum into the lines buffer.
--- @param add fun(s: string)
--- @param atom AtomData
--- @return nil
local function emit_atom_offsets(add, atom)
if #atom.offsets == 0 then return end
add("// --- atom: " .. atom.name .. " (" .. atom.total_words .. " words) ---")
add("")
--- @type OffsetConst[]
local consts = {}
--- @type integer, BranchOffset
for _, r in ipairs(atom.offsets) do
consts[#consts + 1] = make_offset_const(r)
end
--- @type integer, OffsetConst
for _, c in ipairs(consts) do
add("#define " .. pad_right(c.macro_name, OFFSET_MACRO_COL) .. " " .. c.value)
end
add("")
add("enum {")
--- @type integer, OffsetConst
for _, c in ipairs(consts) do
add(" " .. c.enum_name .. " = " .. c.macro_name .. ",")
end
@@ -204,18 +239,23 @@ local function emit_atom_offsets(add, atom)
end
--- Generate the per-directory .offsets.h header.
--- @param dir string -- the absolute source directory
--- @param sources table[] -- sources contributing to this directory (for the header comment)
--- @param atoms_data AtomData[]
--- @param dir string
--- @param sources SourceFile[]
--- @param atoms_data AtomData[]
--- @return string
local function generate_header(dir, sources, atoms_data)
--- @type string
local dir_basename = duffle.basename_no_ext(dir)
--- @type string[]
local lines = {}
--- @param s string
--- @return nil
local function add(s) lines[#lines + 1] = s end
add("// Auto-generated by ps1_meta.lua (passes/offsets.lua) — DO NOT EDIT")
add("// Directory: " .. dir:gsub("/", "\\") .. "\\")
--- @type integer, SourceFile
for _, src in ipairs(sources) do
add("// source: " .. src.path:gsub("/", "\\"))
end
@@ -224,6 +264,7 @@ local function generate_header(dir, sources, atoms_data)
add("#pragma region " .. dir_basename)
add("")
add("")
--- @type integer, AtomData
for _, atom in ipairs(atoms_data) do
emit_atom_offsets(add, atom)
end
@@ -232,21 +273,27 @@ local function generate_header(dir, sources, atoms_data)
return table.concat(lines, "\n") .. "\n"
end
--- @type OffsetsPass
local M = {}
--- (internal) Aggregate atoms from every source in one directory, render the per-directory `offsets.h`.
--- Returns the offsets_h path if a header was written, or nil.
--- @param ctx PassCtx
--- @param dir string -- the absolute source directory
--- @param sources SourceFile[] -- sources in this directory
--- @param errors table[]
--- @return string|nil -- the offsets_h path
--- @param dir string
--- @param sources SourceFile[]
--- @param errors PassFinding[]
--- @return string|nil
local function process_directory(ctx, dir, sources, errors)
--- @type AtomData[]
local atoms_data = {}
--- @param atom AtomEntry
--- @return nil
local function append_atom(atom)
--- @type AtomPaths|nil
local paths = atom and atom.paths
if not paths then return end
--- @type table<string, integer>, OffsetBranch[]
local labels, branches = project_markers(paths.markers)
atoms_data[#atoms_data + 1] = {
name = atom.raw_name or atom.name,
@@ -255,13 +302,18 @@ local function process_directory(ctx, dir, sources, errors)
}
end
--- @type integer, SourceFile
for _, src in ipairs(sources) do
--- @type SourceScan
local scan = src.scan or {}
--- @type integer, AtomEntry
for _, atom in ipairs(scan.atoms or {}) do append_atom(atom) end
--- @type integer, AtomEntry
for _, atom in ipairs(scan.raw_atoms or {}) do append_atom(atom) end
end
if #atoms_data == 0 then return nil end
--- @type string
local out_path = dir .. "/gen/offsets.h"
duffle.ensure_dir(duffle.dirname(out_path))
duffle.write_file(out_path, generate_header(dir, sources, atoms_data))
@@ -274,10 +326,14 @@ end
--- @param ctx PassCtx
--- @return PassResult
function M.run(ctx)
--- @type OffsetOutput[]
local outputs = {}
--- @type PassFinding[]
local errors = {}
--- @type PassFinding[]
local warnings = {}
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" then
error("offsets.run requires ctx.shared.corpus", 0)
@@ -287,8 +343,11 @@ function M.run(ctx)
end
-- Per-directory aggregation: every source in the same directory contributes to one `gen/offsets.h`.
--- @type table<string, SourceFile[]>
local sources_by_dir = corpus.sources_by_dir or duffle.group_sources_by_dir(corpus.source_order)
--- @type string, SourceFile[]
for dir, sources in pairs(sources_by_dir) do
--- @type string|nil
local out_path = process_directory(ctx, dir, sources, errors)
if out_path then
outputs[#outputs + 1] = { offsets_h = out_path }
scripts/passes/report.lua
+401 -43
View File
File diff suppressed because it is too large Load Diff
scripts/passes/scan_source.lua
+922 -32
View File
File diff suppressed because it is too large Load Diff
scripts/passes/static_analysis.lua
+1195 -28
View File
File diff suppressed because it is too large Load Diff
scripts/passes/word_count_eval.lua
+14 -22
View File
@@ -23,7 +23,9 @@
-- Bootstrap: load `scripts/duffle_paths.lua` (sets package.path + package.cpath).
-- Uses `debug.getinfo` to find this file's own directory, so it works both standalone and when require'd from the orchestrator.
-- duffle_paths.lua sets package.path then returns `require("duffle")` at the bottom, so the dofile value IS the duffle module.
--- @type string
local _bootstrap_dir = debug.getinfo(1, "S").source:match("^@?(.*[/\\])") or "./"
--- @type DuffleExport
local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
@@ -31,35 +33,20 @@ local duffle = dofile(_bootstrap_dir .. "../duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
--- @class WordCounts
--- @field [string] integer -- macro name -> word count
--- @field [string] integer -- bag: macro name -> word count
--- @class SourceFile
--- @field path string -- absolute path to the source file
--- @field text string -- the full source text
--- @field dir string -- the directory containing the source
--- @field basename string -- filename without extension
--- @class WordCountEval
--- @field count_token_words fun(token: string, wc: WordCounts): integer
--- @field run fun(ctx: PassCtx): PassResult
--- @class PassCtx
--- @field sources SourceFile[] -- all source files in the build
--- @field metadata_path string -- path to word_count.metadata.h
--- @field shared table -- cross-pass shared state
--- @field shared.corpus table -- canonical corpus (required)
--- @field shared.corpus.word_counts WordCounts -- canonical count table (populated by this pass)
--- @field out_root string -- output root (e.g. "build/gen")
--- @field project_root string -- project root (e.g. "code/")
--- @field upstream table<string, table> -- per-pass upstream outputs
--- @field flags table -- CLI flags
--- @field verbose boolean -- if true, log diagnostic info
--- @class PassResult
--- @field outputs table[] -- {kind=, path=} entries describing emit files
--- @field errors table[] -- {line=, msg=} entries; build-stops
--- @field warnings table[] -- {line=, msg=} entries; build-succeeds
-- SourceFile, PassCtx, PassResult: see ps1_meta.lua
-- DuffleExport: see duffle.lua (facade returned by duffle_paths.lua)
-- ════════════════════════════════════════════════════════════════════════════
-- Module exports
-- ════════════════════════════════════════════════════════════════════════════
--- @type WordCountEval
local M = {}
-- ┌────────────────────────────────────────────────────────────────────┐
@@ -74,11 +61,14 @@ local M = {}
--- @param wc WordCounts -- the shared word-count table
--- @return integer
function M.count_token_words(token, wc)
--- @type string
local s = duffle.trim(token)
if s == "" then return 0 end
--- @type string|nil, integer
local name, after = duffle.read_ident(s, 1)
if not name then return 1 end
if wc[name] then return wc[name] end
--- @type integer
local paren_pos = duffle.skip_ws_and_cmt(s, after)
if s:sub(paren_pos, paren_pos) == "(" then
io.stderr:write(" warning: unknown macro '" .. name .. "', assuming 1 word\n")
@@ -105,6 +95,7 @@ end
--- @return PassResult
function M.run(ctx)
-- 1. Canonical-corpus ownership gate.
--- @type Corpus|nil
local corpus = ctx.shared and ctx.shared.corpus
if type(corpus) ~= "table" then
error("word_count_eval.run requires ctx.shared.corpus (canonical corpus). The fixture must install the corpus before running this pass.", 0)
@@ -117,6 +108,7 @@ function M.run(ctx)
-- 3. Load authored metadata. Generated .macs.h files are NOT scanned
-- (the pass computes their counts from the just-built bodies after disk emission; see passes/components.lua).
--- @type WordCounts
local wc = duffle.load_word_counts(ctx.metadata_path)
-- 4. Assign the count table. ONE assignment, no copy. The assignment creates no secondary alias.
scripts/ps1_meta.lua
+209 -23
View File
@@ -19,7 +19,9 @@
-- fall back to `debug.getinfo(1, "S").source` when this file is being dofile()'d or require()'d (in which case `arg[0]` is the *caller's* path).
-- That single statement: (a) sets `package.path` + `package.cpath`, (b) at the bottom returns `require("duffle")`.
-- So the dofile's return value is the duffle module.
--- @type boolean
local _is_entry_script = arg and arg[0] and arg[0]:match("ps1_meta%.lua$") ~= nil
--- @type string
local _bootstrap_src
if _is_entry_script then
_bootstrap_src = arg[0]
@@ -28,6 +30,7 @@ else
-- strip the leading "@" so the directory match works in both cases.
_bootstrap_src = debug.getinfo(1, "S").source:sub(2)
end
--- @type DuffleExport
local duffle = dofile((_bootstrap_src:match("(.*[/\\])") or "./") .. "duffle_paths.lua")
-- ════════════════════════════════════════════════════════════════════════════
@@ -35,18 +38,24 @@ local duffle = dofile((_bootstrap_src:match("(.*[/\\])") or "./") .. "duffle_pat
-- ════════════════════════════════════════════════════════════════════════════
-- Exit codes (per the --help text and the post-build summary convention).
--- @type integer
local EXIT_OK = 0
--- @type integer
local EXIT_VALIDATION_ERRORS = 1
--- @type integer
local EXIT_INTERNAL_ERROR = 2
-- Default --out-root value if not provided.
--- @type string
local DEFAULT_OUT_ROOT = "build/gen"
-- Sentinel for "all passes" in `PASS_FLAG_TO_NAME`. Distinguishes `--all` from the per-pass flags (which map to individual pass names).
--- @type string
local ALL_PASSES_SENTINEL = "__all__"
-- Sentinel key for the pass-flag dispatcher in `FLAG_HANDLERS`.
-- The actual pass names are looked up via `PASS_FLAG_TO_NAME`, not direct dispatch, so this key never matches a real flag.
--- @type string
local PASS_FLAG_DISPATCH_KEY = "__pass__"
-- ════════════════════════════════════════════════════════════════════════════
@@ -60,38 +69,84 @@ local PASS_FLAG_DISPATCH_KEY = "__pass__"
--- @field deps string[] -- Names of upstream passes
--- @field groups string[]? -- OPTIONAL build-phase groups this pass is a root of (e.g. { "pre-link" }, { "post-link" }); absent ⇒ dependency-only
--- @class SourceFile
--- @field path string -- Absolute path to the source file
--- @field text string -- Full source text
--- @field dir string -- Directory containing the source
--- @field basename string -- Filename without extension
--- @class Corpus
--- @field unity_root string|nil
--- @field project_root string
--- @field code_root string
--- @field source_order SourceFile[]
--- @field sources_by_path table<Path, SourceFile>
--- @field sources_by_dir table<string, SourceFile[]>
--- @field atoms_by_name table<AtomName, AtomEntry>
--- @field binds_by_name table<string, BindsEntry>
--- @field atom_infos AtomInfoEntry[]
--- @field register_alias_registry table<string, AliasEntry>
--- @field type_name_registry table<string, TypeNameEntry>
--- @field atom_views table<AtomName, AtomViewEntry>
--- @field atom_ctxs table<AtomName, AtomCtxEntry>
--- @field atom_phases table<string, AtomPhaseGroup>
--- @field word_counts WordCounts
--- @field components table<string, ComponentDef>
--- @field component_body_index table<string, ComponentBodyEntry>
--- @field collisions CorpusCollision[]
--- @field resolver SourceResolver
--- @field component_atom_infos AtomInfoEntry[]|nil
--- @field atom_auto_regs table<AtomName, table<string, string>>|nil
--- @field phase_auto_regs table<string, table<string, string>>|nil
--- @field reg_use_schemas table<string, RegUseSchema>|nil
--- @field reg_use_errors RegUseError[]|nil
--- @field static_analysis_results table<string, AtomAnalysis>|nil
--- @field tape_chains table<string, TapeChain>|nil
--- @class PassShared
--- @field corpus Corpus
--- @class PassFlags
--- @field gdb_runtime boolean|nil
--- @field dwarf_injection boolean|nil
--- @field elf_path string|nil
--- @class PassCtx
--- @field metadata_path string -- Path to word_count.metadata.h
--- @field shared table -- Cross-pass shared state
--- @field shared.corpus table -- Authored-source/project projection
--- @field out_root string -- Output root (e.g. "build/gen")
--- @field project_root string -- PS1 repository root
--- @field flags table -- CLI flags + per-pass stash
--- @field verbose boolean -- If true, log diagnostic info
--- @field metadata_path string -- Path to word_count.metadata.h
--- @field shared PassShared -- Cross-pass shared state
--- @field out_root string -- Output root (e.g. "build/gen")
--- @field project_root string -- PS1 repository root
--- @field flags PassFlags -- CLI flags + per-pass stash
--- @field verbose boolean -- If true, log diagnostic info
--- @class Finding
--- @class PassFinding
--- @field line integer -- Source line (or 0 for pass-level)
--- @field msg string -- Finding message
--- @class PassOutputEntry
--- @field kind string
--- @field path string
--- @class PassResult
--- @field outputs PassOutputEntry[] -- Emitted file paths
--- @field errors Finding[] -- Build-stops (per-pass kind policy)
--- @field warnings Finding[] -- Informational
--- @field outputs PassOutputEntry[]
--- @field errors PassFinding[] -- Build-stops (per-pass kind policy)
--- @field warnings PassFinding[] -- Informational
--- @field info CheckFinding[]|nil -- static_analysis only
--- @class ParsedArgs
--- @field requested_set string[] -- Pass names to run (explicit --all expanded)
--- @field sources string[] -- Exact --source values, retained in CLI order
--- @field unity_root string|nil -- --unity-root value; mutually exclusive with sources
--- @field metadata string -- --metadata value
--- @field out_root string -- --out-root value (default "build/gen")
--- @field project_root string -- PS1 repository root (derived from metadata by default)
--- @field verbose boolean -- If true, log diagnostic info
--- @field requested_set string[] -- Pass names to run (explicit --all expanded)
--- @field sources string[] -- Exact --source values, retained in CLI order
--- @field unity_root string|nil -- --unity-root value; mutually exclusive with sources
--- @field metadata string -- --metadata value
--- @field out_root string -- --out-root value (default "build/gen")
--- @field project_root string -- PS1 repository root (derived from metadata by default)
--- @field verbose boolean -- If true, log diagnostic info
--- @field flags PassFlags|nil -- Per-pass stash; copied onto PassCtx.flags
--- @alias FlagHandler fun(args: ParsedArgs, argv: string[]|nil, arg_idx: integer|nil): integer|nil
--- @class PassModule
--- @field run fun(ctx: PassCtx): PassResult
--- @class Ps1MetaMod
--- @field PASSES table<string, PassDescriptor>
--- @field PASS_KIND_STOP_ON_ERROR table<string, boolean>
--- @field parse_args fun(argv: string[]): ParsedArgs
--- @field build_ctx fun(args: ParsedArgs): PassCtx
-- ════════════════════════════════════════════════════════════════════════════
-- PASSES Table
@@ -104,6 +159,7 @@ local PASS_FLAG_DISPATCH_KEY = "__pass__"
-- A row without a `groups` entry is dependency-only: it runs only when a transitive dep requests it,
-- but it remains directly requestable through its explicit CLI flag (e.g. --atoms-source-map, --scan-source).
--- @type table<string, PassDescriptor>
local PASSES = {
["scan-source"] = {
module = "passes.scan_source",
@@ -174,9 +230,12 @@ local PASSES = {
--- @param group_name string -- Build-phase group ("pre-link" | "post-link")
--- @return string[] -- Sorted root pass names belonging to that group
local function roots_for_group(group_name)
--- @type string[]
local names = {}
--- @type string, PassDescriptor
for name, pass in pairs(PASSES) do
if pass.groups then
--- @type integer, string
for _, g in ipairs(pass.groups) do
if g == group_name then
names[#names + 1] = name
@@ -194,12 +253,15 @@ end
--- cannot silently fall through to pre-link (or any other default) and dispatch nothing.
--- @param args ParsedArgs
--- @param group_name string
--- @return nil
local function request_roots_for_group(args, group_name)
--- @type string[]
local roots = roots_for_group(group_name)
if #roots == 0 then
error(string.format("ps1_meta: build-phase group %q has zero roots in PASSES; check PASSES rows for a `groups = { %q }` field"
, group_name, group_name))
end
--- @type integer, string
for _, name in ipairs(roots) do
args.requested_set[#args.requested_set + 1] = name
end
@@ -208,6 +270,7 @@ end
-- Pass-kind taxonomy: findings always print. No pass kind stops the build.
-- Report severity is independent from process exit policy.
-- Adding a new pass kind requires listing it here explicitly; an unknown kind must not silently fall back to "true".
--- @type table<string, boolean> -- bag: pass kind -> stop-on-error
local PASS_KIND_STOP_ON_ERROR = {
["shared"] = false,
["header-output"] = false,
@@ -220,6 +283,7 @@ local PASS_KIND_STOP_ON_ERROR = {
-- Per-pass flags (e.g. --word-counts); phase flags (--pre-link, --post-link, --all) are within FLAG_HANDLERS because they own side effects or invoke group-derivation logic.
-- dwarf-injection is *also* a per-pass opt-in flag, but its selection + opt-in state are both owned by the explicit FLAG_HANDLERS entry below
-- (it sets args.flags.dwarf_injection and appends "dwarf-injection" to requested_set), so it is intentionally absent from this table.
--- @type table<string, string> -- bag: CLI flag -> pass name or ALL_PASSES_SENTINEL
local PASS_FLAG_TO_NAME = {
["--word-counts"] = "word-counts",
["--components"] = "components",
@@ -235,10 +299,14 @@ local PASS_FLAG_TO_NAME = {
--- Append every pass name to args.requested_set.
--- Names are derived from PASSES (no parallel name list); used by --all and by any caller that wants the full closure.
--- @param args ParsedArgs
--- @return nil
local function request_all_passes(args)
--- @type string[]
local names = {}
--- @type string
for name in pairs(PASSES) do names[#names + 1] = name end
table.sort(names)
--- @type integer, string
for _, n in ipairs(names) do
args.requested_set[#args.requested_set + 1] = n
end
@@ -246,6 +314,7 @@ end
-- Per-flag handlers. Each handler takes (args, argv, arg_idx) and returns the new arg_idx (so multi-arg flags like --source FILE advance it).
-- Returning nil + os.exit() handles termination flags (--help).
--- @type table<string, FlagHandler>
local FLAG_HANDLERS = {}
-- ════════════════════════════════════════════════════════════════════════════
@@ -253,6 +322,7 @@ local FLAG_HANDLERS = {}
-- ════════════════════════════════════════════════════════════════════════════
--- Print the CLI usage to stdout and exit 0.
--- @return nil
local function print_help()
io.write([[
ps1_meta.lua - Tape-atom metaprogram orchestrator
@@ -302,6 +372,7 @@ EXAMPLES:
]])
end
--- @type table<string, string> -- bag: flag -> value metavar
local FLAG_VALUE_NAMES = {
["--source"] = "FILE",
["--unity-root"] = "FILE",
@@ -311,8 +382,15 @@ local FLAG_VALUE_NAMES = {
["--elf"] = "PATH",
}
--- @param argv string[]
--- @param arg_idx integer
--- @param flag string
--- @return string
--- @return integer
local function require_flag_value(argv, arg_idx, flag)
--- @type string|nil
local value = argv[arg_idx + 1]
--- @type boolean
local next_known = type(value) == "string"
and (FLAG_HANDLERS[value] ~= nil or PASS_FLAG_TO_NAME[value] ~= nil)
if value == nil or next_known then
@@ -327,30 +405,59 @@ end
-- Populated AFTER print_help so the --help handler can reference it as an upvalue (Lua resolves locals at closure-call time,
-- but if the closure is defined before the local, it falls back to _G).
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--help"] = function(args) print_help(); os.exit(0) end
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--verbose"] = function(args) args.verbose = true end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--source"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--source")
args.sources[#args.sources + 1] = value
return value_idx
end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--unity-root"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--unity-root")
args.unity_root = value
return value_idx
end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--metadata"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--metadata")
args.metadata = value
return value_idx
end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--out-root"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--out-root")
args.out_root = value
return value_idx
end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--project-root"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--project-root")
args.project_root = value
return value_idx
@@ -358,11 +465,18 @@ end
-- Per-pass stash flags. Read by `passes/atoms_source_map.lua` to opt into the post-link gdb-runtime emission.
-- Same shape as the existing per-flag handlers. mutates `args.flags` (which propagates into `ctx.flags`).
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--gdb-runtime"] = function(args)
args.flags = args.flags or {}
args.flags.gdb_runtime = true
end
--- @param args ParsedArgs
--- @param argv string[]
--- @param arg_idx integer
--- @return integer
FLAG_HANDLERS["--elf"] = function(args, argv, arg_idx)
--- @type string, integer
local value, value_idx = require_flag_value(argv, arg_idx, "--elf")
args.flags = args.flags or {}
args.flags.elf_path = value
@@ -370,6 +484,8 @@ FLAG_HANDLERS["--elf"] = function(args, argv, arg_idx)
end
-- Enable DWARF injection (default OFF). Opts in to the post-link pass and sets the flag in one shot.
-- The explicit handler below owns both selection and opt-in state, so --dwarf-injection is intentionally absent from PASS_FLAG_TO_NAME.
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--dwarf-injection"] = function(args)
args.flags = args.flags or {}
args.flags.dwarf_injection = true
@@ -377,12 +493,16 @@ FLAG_HANDLERS["--dwarf-injection"] = function(args)
end
-- Build-phase flags: --pre-link and --post-link request the roots of their declared groups (see roots_for_group).
-- topo_sort closes transitive deps from those roots; dispatch_passes runs every pass in the resolved closure without phase-filtering.
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--pre-link"] = function(args)
request_roots_for_group(args, "pre-link")
end
-- Batch post-link phase: gdb-runtime + dwarf-injection in one luajit cold start.
-- Sets the same opt-in flags as --gdb-runtime + --dwarf-injection and selects the post-link build-phase group.
-- elf is required; parse_args enforces it after all flags are parsed.
--- @param args ParsedArgs
--- @return nil
FLAG_HANDLERS["--post-link"] = function(args)
args.flags = args.flags or {}
args.flags.gdb_runtime = true
@@ -393,7 +513,11 @@ end
-- `--dwarf-injection` also emits atom-local debug data.
-- Pass-flag handler. Reads the closed-set table, expands --all, appends to requested_set.
--- @param args ParsedArgs
--- @param a string
--- @return nil
FLAG_HANDLERS[PASS_FLAG_DISPATCH_KEY] = function(args, a)
--- @type string|nil
local name = PASS_FLAG_TO_NAME[a]
if name == ALL_PASSES_SENTINEL then
request_all_passes(args)
@@ -406,6 +530,7 @@ end
--- @param argv string[]
--- @return ParsedArgs
local function parse_args(argv)
--- @type ParsedArgs
local args = {
requested_set = {},
sources = {},
@@ -416,9 +541,12 @@ local function parse_args(argv)
verbose = false,
}
--- @type integer
local pos = 1
while pos <= #argv do
--- @type string
local a = argv[pos]
--- @type FlagHandler|nil
local handler = FLAG_HANDLERS[a]
if handler then
pos = handler(args, argv, pos) or pos
@@ -444,13 +572,16 @@ local function parse_args(argv)
-- `<repo>/code/duffle/word_count.metadata.h` is the canonical metadata location.
-- `project_root` names `<repo>`; the resolver derives `<project_root>/code` separately.
if not args.project_root then
--- @type string
local metadata_dir = duffle.dirname(duffle.normalize_path(args.metadata))
--- @type string
local code_root = duffle.dirname(metadata_dir)
args.project_root = duffle.dirname(code_root)
else
args.project_root = duffle.normalize_path(args.project_root)
end
--- @type boolean
local has_unity = type(args.unity_root) == "string" and args.unity_root ~= ""
if has_unity and #args.sources > 0 then
io.stderr:write("ps1_meta: --unity-root FILE and --source FILE are mutually exclusive\n")
@@ -464,9 +595,13 @@ local function parse_args(argv)
-- Post-link opt-ins (--gdb-runtime, --dwarf-injection) write output that depends on the linked ELF.
-- Without --elf the metaprogram can't satisfy those requests, so refuse loud and early.
-- This covers the explicit --post-link batch, --dwarf-injection by itself, and --gdb-runtime by itself.
--- @type PassFlags
local flags = args.flags or {}
--- @type string|nil
local elf_path = flags.elf_path
--- @type boolean
local has_elf = type(elf_path) == "string" and #elf_path > 0
--- @type boolean
local post_links = flags.gdb_runtime or flags.dwarf_injection
if post_links and not has_elf then
io.stderr:write("ps1_meta: --elf PATH is required for post-link output\n")
@@ -486,8 +621,11 @@ end
--- @param args ParsedArgs
--- @return PassCtx
local function build_ctx(args)
--- @type string
local normalized_project_root = duffle.normalize_path(args.project_root)
--- @type string
local project_root = normalized_project_root
--- @type boolean
local project_root_is_absolute = normalized_project_root:match("^%a:/")
or normalized_project_root:sub(1, 2) == "//"
or normalized_project_root:sub(1, 1) == "/"
@@ -499,8 +637,10 @@ local function build_ctx(args)
-- Do not route POSIX/UNC/drive-absolute paths through to_absolute_path.
duffle.canonical_path_key(project_root)
end
--- @type Corpus
local resolution
if args.unity_root then
--- @type boolean, Corpus|string
local ok_resolve, resolved = pcall(duffle.resolve_source_corpus, {
unity_root = args.unity_root,
project_root = project_root,
@@ -511,27 +651,36 @@ local function build_ctx(args)
end
resolution = resolved
else
--- @type SourceFile[]
local source_order = {}
--- @type table<Path, SourceFile>
local sources_by_path = {}
--- @type SourceResolver
local resolver = {
resolved = {},
skipped = {},
shadowed = {},
}
--- @type integer, string
for _, input_path in ipairs(args.sources) do
--- @type string
local path = duffle.normalize_path(input_path)
--- @type boolean, string
local key_ok, key_or_error = pcall(duffle.canonical_path_key, path)
if not key_ok then
error("ps1_meta: invalid --source " .. input_path .. ": " .. tostring(key_or_error), 0)
end
--- @type file*|nil
local file = io.open(path, "r")
if not file then
io.stderr:write("ps1_meta: cannot open --source " .. input_path .. "\n")
os.exit(EXIT_INTERNAL_ERROR)
end
--- @type string
local text = file:read("*a")
file:close()
--- @type SourceFile
local source = {
path = path,
text = text,
@@ -539,6 +688,7 @@ local function build_ctx(args)
basename = duffle.basename_no_ext(path),
}
source_order[#source_order + 1] = source
--- @type string
local key = key_or_error
if not sources_by_path[key] then sources_by_path[key] = source end
resolver.resolved[#resolver.resolved + 1] = {
@@ -563,6 +713,7 @@ local function build_ctx(args)
}
end
--- @type Corpus
local corpus = {
unity_root = resolution.unity_root,
project_root = resolution.project_root,
@@ -584,6 +735,7 @@ local function build_ctx(args)
collisions = {},
resolver = resolution.resolver,
}
--- @type PassCtx
local ctx = {
metadata_path = args.metadata,
shared = { corpus = corpus },
@@ -613,14 +765,20 @@ end
--- Keeping these blocks local makes the topological sort self-contained.
local function topo_sort(passes, requested_set)
-- Dependency closure: include every pass transitively required by `requested_set`.
--- @type table<string, boolean> -- bag: pass name -> needed
local needed = {}
--- @type integer, string
for _, name in ipairs(requested_set) do needed[name] = true end
--- @type boolean
local changed = true
while changed do
changed = false
--- @type string, boolean
for name, _ in pairs(needed) do
--- @type PassDescriptor
local pass = passes[name]
if not pass then error("unknown pass '" .. name .. "' requested") end
--- @type integer, string
for _, dep in ipairs(pass.deps) do
if not needed[dep] then
needed[dep] = true
@@ -631,9 +789,13 @@ local function topo_sort(passes, requested_set)
end
-- In-degree calculation: count each needed pass's needed dependencies.
--- @type table<string, integer> -- bag: pass name -> in-degree
local in_degree = {}
--- @type string, boolean
for name, _ in pairs(needed) do in_degree[name] = 0 end
--- @type string, boolean
for name, _ in pairs(needed) do
--- @type integer, string
for _, dep in ipairs(passes[name].deps) do
if needed[dep] then
in_degree[name] = in_degree[name] + 1
@@ -642,7 +804,9 @@ local function topo_sort(passes, requested_set)
end
-- Ready-queue seeding: add zero-in-degree passes in deterministic order.
--- @type string[]
local ready = {}
--- @type string, integer
for name, deg in pairs(in_degree) do
if deg == 0 then ready[#ready + 1] = name end
end
@@ -650,12 +814,16 @@ local function topo_sort(passes, requested_set)
-- Ready-queue drain: decrement dependents when each pass is emitted.
-- Newly-zero-degree passes are inserted back into the ready queue (kept sorted).
--- @type string[]
local order = {}
while #ready > 0 do
--- @type string
local just_finished = table.remove(ready, 1)
order[#order + 1] = just_finished
--- @type string, boolean
for name, _ in pairs(needed) do
if name ~= just_finished then
--- @type integer, string
for _, dep in ipairs(passes[name].deps) do
if dep == just_finished then
in_degree[name] = in_degree[name] - 1
@@ -672,9 +840,12 @@ local function topo_sort(passes, requested_set)
-- Cycle detection: if `order` doesn't include all needed passes, some are stuck with in_degree > 0
-- (the cycle closed on itself before Kahn could process them).
-- Without this check, a fully-closed cycle (e.g. A -> B -> A) would silently return an empty order list, leaving the orchestrator to dispatch nothing.
--- @type integer
local needed_count = 0
--- @type string
for _ in pairs(needed) do needed_count = needed_count + 1 end -- count hash entries; Lua's #t doesn't work
if #order ~= needed_count then
--- @type string, integer
for name, deg in pairs(in_degree) do
if deg > 0 then
error("dependency cycle detected involving pass '" .. name .. "'")
@@ -696,8 +867,10 @@ end
--- @param result PassResult
--- @return boolean
local function report_validation_errors(pass_name, pass, result)
--- @type boolean
local has_errors = result.errors and #result.errors > 0
if not has_errors then return false end
--- @type integer, PassFinding
for _, e in ipairs(result.errors) do
io.stderr:write(string.format("[%s] line %d: %s\n", pass_name, e.line or 0, e.msg or ""))
end
@@ -709,10 +882,15 @@ end
--- @param order string[]
--- @return boolean -- true if any validation errors were reported
local function dispatch_passes(ctx, order)
--- @type boolean
local had_errors = false
--- @type integer, string
for _, pass_name in ipairs(order) do
--- @type PassDescriptor
local pass = PASSES[pass_name]
--- @type PassModule
local mod = require(pass.module)
--- @type PassResult
local result = mod.run(ctx)
if report_validation_errors(pass_name, pass, result) then
had_errors = true
@@ -723,14 +901,21 @@ end
--- Main entry point. Runs the requested passes in dep-topological order.
--- @param argv string[]
--- @return nil
local function main(argv)
--- @type boolean, string|nil
local ok, err = pcall(function()
--- @type ParsedArgs
local args = parse_args(argv)
--- @type PassCtx
local ctx = build_ctx(args)
--- @type string[]
local requested = args.requested_set
--- @type string[]
local closed = topo_sort(PASSES, requested)
--- @type boolean
local had_errors = dispatch_passes(ctx, closed)
if had_errors then os.exit(EXIT_VALIDATION_ERRORS) end
end)
@@ -746,6 +931,7 @@ end
-- Module export for in-process consumers (tests that dofile this script).
-- The conditional `main(...)` call below only fires when this file is invoked as the entry script (arg[0] ends in "ps1_meta.lua");
-- in dofile() mode (test's arg[0] does not match), main() is skipped and the chunk returns `_M` to the caller.
--- @type Ps1MetaMod
local _M = {
PASSES = PASSES,
PASS_KIND_STOP_ON_ERROR = PASS_KIND_STOP_ON_ERROR,