Phase 2: Fix sema stub — add full function body type checking

- Iterate function body statements and call Sema_CheckStmt on each
- Add Sema_CheckBlock helper for child scope creation in if/while/for/loop blocks
- Extend Sema_CheckStmt to handle all statement kinds: skIf, skWhile, skDoWhile,
  skLoop, skFor, skMatch, skBreak, skContinue, skDecl, skReturn
- Extend Sema_CheckExpr with ekAssign, ekField, ekIndex, ekBlock, ekSelf
- Fix ekCall to resolve return type from function declaration
- Fix ekBinary to handle assignment operators (tkAssign, +=, -=, etc.)
- Add Sema_CollectGlobals handling for dkUse (imports), dkConst, dkEnum variants
- Add tkColonColon path parsing in parser (Color::Red)
- Add decl field to Symbol struct for function/type resolution
- Add Sema_ZeroInitSymbol helper to avoid garbage from uninitialized struct fields
- Fix concurrency.bux missing Task_Join import

All examples now pass buxc2 check. make selfhost succeeds.
This commit is contained in:
2026-06-05 14:07:38 +03:00
parent f0f94f30e1
commit 5ae85d5bd9
9 changed files with 667 additions and 20 deletions
+9
View File
@@ -170,6 +170,7 @@ func Cli_Check(srcPath: String) -> int {
PrintLine("Parse failed");
return 1;
}
PrintLine(" Parse done");
// Flatten module wrappers
var decl2: *Decl = mod.firstItem;
@@ -185,6 +186,14 @@ func Cli_Check(srcPath: String) -> int {
let sema: *Sema = Sema_Analyze(mod);
if Sema_HasError(sema) {
PrintLine("Sema errors found");
var i: int = 0;
while i < Sema_DiagCount(sema) {
Print(" line ");
PrintInt(sema.diags[i].line as int64);
Print(": ");
PrintLine(sema.diags[i].message);
i = i + 1;
}
return 1;
}
+13
View File
@@ -413,6 +413,19 @@ func parserParsePostfix(p: *Parser) -> *Expr {
continue;
}
// Path: expr::name
if kind == tkColonColon {
discard parserAdvance(p);
let line: uint32 = parserCurToken(p).line;
let col: uint32 = parserCurToken(p).column;
let name: LexToken = parserExpect(p, tkIdent, "expected path segment");
let e: *Expr = parserMakeExpr(ekField, line, col);
e.child1 = left;
e.strValue = name.text;
left = e;
continue;
}
// as, is
if kind == tkAs || kind == tkIs {
discard parserAdvance(p);
+2
View File
@@ -18,6 +18,7 @@ struct Symbol {
typeName: String;
isMutable: bool;
isPublic: bool;
decl: *Decl; // associated declaration (for funcs, structs, enums)
}
struct Scope {
@@ -90,4 +91,5 @@ func Scope_LookupLocal(scope: *Scope, name: String) -> Symbol {
func Scope_Free(scope: *Scope) {
bux_free(scope.symbols as *void);
}
}
+309 -10
View File
@@ -2,8 +2,6 @@
// Validates types, resolves identifiers, checks function calls.
module Sema {
// ---------------------------------------------------------------------------
// Sema context
// ---------------------------------------------------------------------------
@@ -15,6 +13,7 @@ struct Sema {
diagCount: int;
diags: *SemaDiag;
hasError: bool;
currentRetType: int; // return type of the function being checked
}
struct SemaDiag {
@@ -35,6 +34,20 @@ func Sema_EmitError(sema: *Sema, line: uint32, col: uint32, msg: String) {
sema.hasError = true;
}
// ---------------------------------------------------------------------------
// Symbol zero-init helper (bootstrap C backend does not zero-init structs)
// ---------------------------------------------------------------------------
func Sema_ZeroInitSymbol(sym: *Symbol) {
sym.kind = 0;
sym.name = "";
sym.typeKind = 0;
sym.typeName = "";
sym.isMutable = false;
sym.isPublic = false;
sym.decl = null as *Decl;
}
// ---------------------------------------------------------------------------
// Type resolution from TypeExpr → Type constants
// ---------------------------------------------------------------------------
@@ -105,6 +118,23 @@ func Sema_TypeName(kind: int) -> String {
return "?";
}
// ---------------------------------------------------------------------------
// Block checking helper
// ---------------------------------------------------------------------------
func Sema_CheckBlock(sema: *Sema, block: *Block) {
if block == null as *Block { return; }
var blockScope: Scope = Scope_NewChild(sema.scope);
let prevScope: *Scope = sema.scope;
sema.scope = &blockScope;
var stmt: *Stmt = block.firstStmt;
while stmt != null as *Stmt {
Sema_CheckStmt(sema, stmt);
stmt = stmt.nextStmt;
}
sema.scope = prevScope;
}
// ---------------------------------------------------------------------------
// Expression type checking
// ---------------------------------------------------------------------------
@@ -112,6 +142,10 @@ func Sema_TypeName(kind: int) -> String {
func Sema_CheckExpr(sema: *Sema, expr: *Expr) -> int {
if expr == null as *Expr { return tyUnknown; }
let kind: int = expr.kind;
// Debug: print expr kind
// Print("Sema_CheckExpr kind=");
// PrintInt(kind as int64);
// PrintLine("");
// Literal
if kind == ekLiteral {
@@ -141,12 +175,26 @@ func Sema_CheckExpr(sema: *Sema, expr: *Expr) -> int {
return sym.typeKind;
}
// self
if kind == ekSelf {
let sym: Symbol = Scope_Lookup(sema.scope, "self");
if sym.kind == 0 && !String_Eq(sym.name, "self") {
Sema_EmitError(sema, expr.line, expr.column, "self outside method");
return tyUnknown;
}
return sym.typeKind;
}
// Binary
if kind == ekBinary {
let left: int = Sema_CheckExpr(sema, expr.child1);
let right: int = Sema_CheckExpr(sema, expr.child2);
let op: int = expr.intValue;
// Assignment operators return target type
if op == tkAssign || (op >= tkPlusAssign && op <= tkShrAssign) {
return left;
}
// Comparison operators return bool
if op >= tkEq && op <= tkGe { return tyBool; }
// Logical operators return bool
@@ -169,12 +217,31 @@ func Sema_CheckExpr(sema: *Sema, expr: *Expr) -> int {
return operand;
}
// Assign
if kind == ekAssign {
let target: int = Sema_CheckExpr(sema, expr.child1);
let value: int = Sema_CheckExpr(sema, expr.child2);
// Basic type compatibility (permissive for now)
return target;
}
// Call
if kind == ekCall {
let callee: int = Sema_CheckExpr(sema, expr.child1);
if callee == tyUnknown { return tyUnknown; }
// Assume callee returns same type (simplified)
return callee;
discard Sema_CheckExpr(sema, expr.child1);
if expr.child2 != null as *Expr {
discard Sema_CheckExpr(sema, expr.child2);
}
if expr.child3 != null as *Expr {
discard Sema_CheckExpr(sema, expr.child3);
}
// Try to resolve return type from function declaration
if expr.child1.kind == ekIdent {
let sym: Symbol = Scope_Lookup(sema.scope, expr.child1.strValue);
if sym.kind == skFunc && sym.decl != null as *Decl && sym.decl.retType != null as *TypeExpr {
return Sema_ResolveType(sema, sym.decl.retType);
}
}
return tyUnknown;
}
// Ternary
@@ -201,6 +268,30 @@ func Sema_CheckExpr(sema: *Sema, expr: *Expr) -> int {
return tyNamed;
}
// Field access
if kind == ekField {
discard Sema_CheckExpr(sema, expr.child1);
return tyUnknown;
}
// Index
if kind == ekIndex {
let obj: int = Sema_CheckExpr(sema, expr.child1);
let idx: int = Sema_CheckExpr(sema, expr.child2);
if !Sema_IsNumeric(idx) && idx != tyUnknown {
Sema_EmitError(sema, expr.line, expr.column, "index must be integer");
}
return tyUnknown;
}
// Block expression
if kind == ekBlock {
if expr.refBlock != null as *Block {
Sema_CheckBlock(sema, expr.refBlock);
}
return tyVoid;
}
return tyUnknown;
}
@@ -226,6 +317,7 @@ func Sema_CheckStmt(sema: *Sema, stmt: *Stmt) {
}
sym.isMutable = stmt.boolValue;
sym.isPublic = false;
sym.decl = null as *Decl;
discard Scope_Define(sema.scope, sym);
return;
}
@@ -233,7 +325,21 @@ func Sema_CheckStmt(sema: *Sema, stmt: *Stmt) {
// Return
if kind == skReturn {
if stmt.child1 != null as *Expr {
discard Sema_CheckExpr(sema, stmt.child1);
let retType: int = Sema_CheckExpr(sema, stmt.child1);
if sema.currentRetType != tyUnknown && retType != tyUnknown {
if retType != sema.currentRetType {
// Be permissive: allow numeric widening, pointer/named mismatch
if !Sema_IsNumeric(retType) || !Sema_IsNumeric(sema.currentRetType) {
if retType != sema.currentRetType {
Sema_EmitError(sema, stmt.line, stmt.column, "return type mismatch");
}
}
}
}
} else {
if sema.currentRetType != tyVoid && sema.currentRetType != tyUnknown {
Sema_EmitError(sema, stmt.line, stmt.column, "missing return value");
}
}
return;
}
@@ -244,6 +350,8 @@ func Sema_CheckStmt(sema: *Sema, stmt: *Stmt) {
if !Sema_IsBool(condType) && condType != tyUnknown {
Sema_EmitError(sema, stmt.line, stmt.column, "if condition must be bool");
}
Sema_CheckBlock(sema, stmt.refStmtBlock);
Sema_CheckBlock(sema, stmt.refStmtElse);
return;
}
@@ -253,6 +361,54 @@ func Sema_CheckStmt(sema: *Sema, stmt: *Stmt) {
if !Sema_IsBool(condType) && condType != tyUnknown {
Sema_EmitError(sema, stmt.line, stmt.column, "while condition must be bool");
}
Sema_CheckBlock(sema, stmt.refStmtBlock);
return;
}
// Do-while
if kind == skDoWhile {
Sema_CheckBlock(sema, stmt.refStmtBlock);
let condType: int = Sema_CheckExpr(sema, stmt.child1);
if !Sema_IsBool(condType) && condType != tyUnknown {
Sema_EmitError(sema, stmt.line, stmt.column, "do-while condition must be bool");
}
return;
}
// Loop
if kind == skLoop {
Sema_CheckBlock(sema, stmt.refStmtBlock);
return;
}
// For
if kind == skFor {
discard Sema_CheckExpr(sema, stmt.child1);
var forScope: Scope = Scope_NewChild(sema.scope);
var loopSym: Symbol;
loopSym.kind = skVar;
loopSym.name = stmt.strValue;
loopSym.typeKind = tyUnknown;
loopSym.typeName = "";
loopSym.isMutable = true;
loopSym.isPublic = false;
loopSym.decl = null as *Decl;
discard Scope_Define(&forScope, loopSym);
let prevScope: *Scope = sema.scope;
sema.scope = &forScope;
Sema_CheckBlock(sema, stmt.refStmtBlock);
sema.scope = prevScope;
return;
}
// Match
if kind == skMatch {
discard Sema_CheckExpr(sema, stmt.child1);
return;
}
// Break / Continue
if kind == skBreak || kind == skContinue {
return;
}
@@ -261,6 +417,14 @@ func Sema_CheckStmt(sema: *Sema, stmt: *Stmt) {
discard Sema_CheckExpr(sema, stmt.child1);
return;
}
// Decl (nested)
if kind == skDecl {
if stmt.refStmtDecl != null as *Decl && stmt.refStmtDecl.kind == dkFunc {
Sema_EmitError(sema, stmt.line, stmt.column, "nested functions not yet supported");
}
return;
}
}
// ---------------------------------------------------------------------------
@@ -275,40 +439,158 @@ func Sema_CollectGlobals(sema: *Sema) {
// Function
if dk == dkFunc {
var sym: Symbol;
Sema_ZeroInitSymbol(&sym);
sym.kind = skFunc;
sym.name = decl.strValue;
sym.typeKind = tyFunc;
sym.isPublic = decl.isPublic;
sym.decl = decl;
discard Scope_Define(sema.scope, sym);
}
// Struct
if dk == dkStruct {
var sym: Symbol;
Sema_ZeroInitSymbol(&sym);
sym.kind = skType;
sym.name = decl.strValue;
sym.typeKind = tyNamed;
sym.isPublic = decl.isPublic;
sym.decl = decl;
discard Scope_Define(sema.scope, sym);
}
// Enum
if dk == dkEnum {
var sym: Symbol;
Sema_ZeroInitSymbol(&sym);
sym.kind = skType;
sym.name = decl.strValue;
sym.typeKind = tyNamed;
sym.isPublic = decl.isPublic;
sym.decl = decl;
discard Scope_Define(sema.scope, sym);
// Register enum variants as constants
var vi: int = 0;
while vi < decl.variantCount && vi < 8 {
var v: EnumVariant;
if vi == 0 { v = decl.variant0; }
else if vi == 1 { v = decl.variant1; }
else if vi == 2 { v = decl.variant2; }
else if vi == 3 { v = decl.variant3; }
else if vi == 4 { v = decl.variant4; }
else if vi == 5 { v = decl.variant5; }
else if vi == 6 { v = decl.variant6; }
else if vi == 7 { v = decl.variant7; }
let variantName: String = String_Concat(decl.strValue, "_");
let variantName2: String = String_Concat(variantName, v.name);
var vSym: Symbol;
vSym.kind = skConst;
vSym.name = variantName2;
vSym.typeKind = tyNamed;
vSym.typeName = String_Concat(decl.strValue, "_Tag");
vSym.isMutable = false;
vSym.isPublic = decl.isPublic;
vSym.decl = decl;
discard Scope_Define(sema.scope, vSym);
vi = vi + 1;
}
}
// Use (import)
if dk == dkUse {
if decl.useKind == 2 {
// Multi-import: names are comma-separated in useNames
// For now, register the whole useNames string as a single func name
// (permissive fallback — real fix needs string split)
let namesStr: String = decl.useNames;
if !String_Eq(namesStr, "") {
// Simple split by comma using available string functions
var start: uint = 0;
var pos: uint = 0;
let totalLen: uint = String_Len(namesStr);
while pos <= totalLen {
let atEnd: bool = pos == totalLen;
let isComma: bool = false;
if pos < totalLen {
// Check if character at pos is comma
let chStr: String = bux_str_slice(namesStr, pos, 1);
isComma = String_Eq(chStr, ",");
}
if atEnd || isComma {
let nameLen: uint = pos - start;
if nameLen > 0 {
let name: String = bux_str_slice(namesStr, start, nameLen);
var sym: Symbol;
sym.kind = skFunc;
sym.name = name;
sym.typeKind = tyUnknown;
sym.typeName = "";
sym.isMutable = false;
sym.isPublic = true;
sym.decl = null as *Decl;
discard Scope_Define(sema.scope, sym);
}
start = pos + 1;
}
pos = pos + 1;
}
}
} else {
// Single import or glob — add last path segment
let path: String = decl.usePath;
if !String_Eq(path, "") {
// Find last :: segment
var lastSeg: String = path;
let containsColons: int = bux_str_contains(path, "::");
if containsColons != 0 {
// Try to get last segment by finding last ::
// Use a simple heuristic: slice from various positions
let pathLen: uint = String_Len(path);
var tryPos: uint = 0;
while tryPos < pathLen {
let slice: String = bux_str_slice(path, tryPos, pathLen - tryPos);
if String_StartsWith(slice, "::") {
lastSeg = bux_str_slice(slice, 2, String_Len(slice) - 2);
}
tryPos = tryPos + 1;
}
}
var sym: Symbol;
sym.kind = skFunc;
sym.name = lastSeg;
sym.typeKind = tyUnknown;
sym.typeName = "";
sym.isMutable = false;
sym.isPublic = true;
sym.decl = null as *Decl;
discard Scope_Define(sema.scope, sym);
}
}
}
// Const
if dk == dkConst {
var sym: Symbol;
Sema_ZeroInitSymbol(&sym);
sym.kind = skConst;
sym.name = decl.strValue;
sym.typeKind = tyUnknown;
sym.isPublic = decl.isPublic;
sym.decl = decl;
discard Scope_Define(sema.scope, sym);
}
// Extern function
if dk == dkExternFunc {
var sym: Symbol;
Sema_ZeroInitSymbol(&sym);
sym.kind = skFunc;
sym.name = decl.strValue;
sym.typeKind = tyFunc;
sym.isPublic = true;
sym.decl = decl;
discard Scope_Define(sema.scope, sym);
}
@@ -332,6 +614,7 @@ func Sema_Analyze(mod: *Module) -> *Sema {
s.diags = bux_alloc(256 as uint * sizeof(SemaDiag)) as *SemaDiag;
s.typeTable = null as *void;
s.methodTable = null as *void;
s.currentRetType = tyVoid;
// First pass: collect globals
Sema_CollectGlobals(s);
@@ -349,6 +632,7 @@ func Sema_Analyze(mod: *Module) -> *Sema {
tpSym.kind = skType;
tpSym.name = decl.typeParam0;
tpSym.typeKind = tyTypeParam;
tpSym.decl = null as *Decl;
discard Scope_Define(&funcScope, tpSym);
}
if decl.typeParamCount >= 2 {
@@ -356,6 +640,7 @@ func Sema_Analyze(mod: *Module) -> *Sema {
tpSym2.kind = skType;
tpSym2.name = decl.typeParam1;
tpSym2.typeKind = tyTypeParam;
tpSym2.decl = null as *Decl;
discard Scope_Define(&funcScope, tpSym2);
}
@@ -370,6 +655,7 @@ func Sema_Analyze(mod: *Module) -> *Sema {
else if i == 4 { p = &decl.param4; }
else if i == 5 { p = &decl.param5; }
var pSym: Symbol;
Sema_ZeroInitSymbol(&pSym);
pSym.kind = skVar;
if p != null as *Param && p.refParamType != null as *TypeExpr {
pSym.typeKind = Sema_ResolveType(s, p.refParamType);
@@ -379,6 +665,8 @@ func Sema_Analyze(mod: *Module) -> *Sema {
pSym.typeName = "";
}
pSym.isMutable = false;
pSym.isPublic = false;
pSym.decl = null as *Decl;
if i == 0 { pSym.name = decl.param0.name; }
else if i == 1 { pSym.name = decl.param1.name; }
else if i == 2 { pSym.name = decl.param2.name; }
@@ -393,9 +681,19 @@ func Sema_Analyze(mod: *Module) -> *Sema {
let prevScope: *Scope = s.scope;
s.scope = &funcScope;
// Check body (simplified — just count statements)
var stmtCount: int = decl.refBody.stmtCount;
// In a full implementation, iterate statements and check each
// Set current function return type
if decl.retType != null as *TypeExpr {
s.currentRetType = Sema_ResolveType(s, decl.retType);
} else {
s.currentRetType = tyVoid;
}
// Check body statements
var stmt: *Stmt = decl.refBody.firstStmt;
while stmt != null as *Stmt {
Sema_CheckStmt(s, stmt);
stmt = stmt.nextStmt;
}
s.scope = prevScope;
}
@@ -419,4 +717,5 @@ func Sema_Free(sema: *Sema) {
bux_free(sema.diags as *void);
bux_free(sema as *void);
}
}