diff --git a/src/all_types.hpp b/src/all_types.hpp index 908c0e327c..fd66b77ad2 100644 --- a/src/all_types.hpp +++ b/src/all_types.hpp @@ -1038,6 +1038,7 @@ bool fn_type_id_eql(FnTypeId *a, FnTypeId *b); enum PtrLen { PtrLenUnknown, PtrLenSingle, + PtrLenC, }; struct ZigTypePointer { diff --git a/src/analyze.cpp b/src/analyze.cpp index 970d1cc382..e561050e0d 100644 --- a/src/analyze.cpp +++ b/src/analyze.cpp @@ -417,6 +417,18 @@ ZigType *get_promise_type(CodeGen *g, ZigType *result_type) { return entry; } +static const char *ptr_len_to_star_str(PtrLen ptr_len) { + switch (ptr_len) { + case PtrLenSingle: + return "*"; + case PtrLenUnknown: + return "[*]"; + case PtrLenC: + return "[*c]"; + } + zig_unreachable(); +} + ZigType *get_pointer_to_type_extra(CodeGen *g, ZigType *child_type, bool is_const, bool is_volatile, PtrLen ptr_len, uint32_t byte_alignment, uint32_t bit_offset_in_host, uint32_t host_int_bytes) @@ -466,7 +478,7 @@ ZigType *get_pointer_to_type_extra(CodeGen *g, ZigType *child_type, bool is_cons ZigType *entry = new_type_table_entry(ZigTypeIdPointer); - const char *star_str = ptr_len == PtrLenSingle ? "*" : "[*]"; + const char *star_str = ptr_len_to_star_str(ptr_len); const char *const_str = is_const ? "const " : ""; const char *volatile_str = is_volatile ? "volatile " : ""; buf_resize(&entry->name, 0); diff --git a/src/ir.cpp b/src/ir.cpp index 5d4013b4b9..76277f541b 100644 --- a/src/ir.cpp +++ b/src/ir.cpp @@ -169,6 +169,10 @@ static ConstExprValue *ir_resolve_const(IrAnalyze *ira, IrInstruction *value, Un static void copy_const_val(ConstExprValue *dest, ConstExprValue *src, bool same_global_refs); static Error resolve_ptr_align(IrAnalyze *ira, ZigType *ty, uint32_t *result_align); static void ir_add_alloca(IrAnalyze *ira, IrInstruction *instruction, ZigType *type_entry); +static IrInstruction *ir_analyze_int_to_ptr(IrAnalyze *ira, IrInstruction *source_instr, IrInstruction *target, + ZigType *ptr_type); +static IrInstruction *ir_analyze_bit_cast(IrAnalyze *ira, IrInstruction *source_instr, IrInstruction *value, + ZigType *dest_type); static ConstExprValue *const_ptr_pointee_unchecked(CodeGen *g, ConstExprValue *const_val) { assert(get_src_ptr_type(const_val->type) != nullptr); @@ -5019,10 +5023,23 @@ static IrInstruction *ir_lval_wrap(IrBuilder *irb, Scope *scope, IrInstruction * return ir_build_ref(irb, scope, value->source_node, value, false, false); } +static PtrLen star_token_to_ptr_len(TokenId token_id) { + switch (token_id) { + case TokenIdStar: + case TokenIdStarStar: + return PtrLenSingle; + case TokenIdBracketStarBracket: + return PtrLenUnknown; + case TokenIdBracketStarCBracket: + return PtrLenC; + default: + zig_unreachable(); + } +} + static IrInstruction *ir_gen_pointer_type(IrBuilder *irb, Scope *scope, AstNode *node) { assert(node->type == NodeTypePointerType); - PtrLen ptr_len = (node->data.pointer_type.star_token->id == TokenIdStar || - node->data.pointer_type.star_token->id == TokenIdStarStar) ? PtrLenSingle : PtrLenUnknown; + PtrLen ptr_len = star_token_to_ptr_len(node->data.pointer_type.star_token->id); bool is_const = node->data.pointer_type.is_const; bool is_volatile = node->data.pointer_type.is_volatile; AstNode *expr_node = node->data.pointer_type.op_expr; @@ -8538,6 +8555,20 @@ static bool ir_num_lit_fits_in_other_type(IrAnalyze *ira, IrInstruction *instruc } } } + if (other_type->id == ZigTypeIdPointer && other_type->data.pointer.ptr_len == PtrLenC && const_val_is_int) { + if (!bigint_fits_in_bits(&const_val->data.x_bigint, ira->codegen->pointer_size_bytes * 8, true) && + !bigint_fits_in_bits(&const_val->data.x_bigint, ira->codegen->pointer_size_bytes * 8, false)) + { + Buf *val_buf = buf_alloc(); + bigint_append_buf(val_buf, &const_val->data.x_bigint, 10); + + ir_add_error(ira, instruction, + buf_sprintf("integer value %s outside of pointer address range", + buf_ptr(val_buf))); + return false; + } + return true; + } const char *num_lit_str; Buf *val_buf = buf_alloc(); @@ -10811,6 +10842,37 @@ static IrInstruction *ir_analyze_vector_to_array(IrAnalyze *ira, IrInstruction * return ir_build_vector_to_array(ira, source_instr, vector, array_type); } +static IrInstruction *ir_analyze_int_to_c_ptr(IrAnalyze *ira, IrInstruction *source_instr, + IrInstruction *integer, ZigType *dest_type) +{ + IrInstruction *unsigned_integer; + if (instr_is_comptime(integer)) { + unsigned_integer = integer; + } else { + assert(integer->value.type->id == ZigTypeIdInt); + + if (integer->value.type->data.integral.bit_count > + ira->codegen->builtin_types.entry_usize->data.integral.bit_count) + { + ir_add_error(ira, source_instr, + buf_sprintf("integer type too big for implicit @intToPtr to type '%s'", buf_ptr(&dest_type->name))); + return ira->codegen->invalid_instruction; + } + + if (integer->value.type->data.integral.is_signed) { + ZigType *unsigned_int_type = get_int_type(ira->codegen, false, + integer->value.type->data.integral.bit_count); + unsigned_integer = ir_analyze_bit_cast(ira, source_instr, integer, unsigned_int_type); + if (type_is_invalid(unsigned_integer->value.type)) + return ira->codegen->invalid_instruction; + } else { + unsigned_integer = integer; + } + } + + return ir_analyze_int_to_ptr(ira, source_instr, unsigned_integer, dest_type); +} + static IrInstruction *ir_analyze_cast(IrAnalyze *ira, IrInstruction *source_instr, ZigType *wanted_type, IrInstruction *value) { @@ -11217,6 +11279,14 @@ static IrInstruction *ir_analyze_cast(IrAnalyze *ira, IrInstruction *source_inst return ir_analyze_array_to_vector(ira, source_instr, value, wanted_type); } + // casting to C pointers + if (wanted_type->id == ZigTypeIdPointer && wanted_type->data.pointer.ptr_len == PtrLenC) { + // cast from integer to C pointer + if (actual_type->id == ZigTypeIdInt || actual_type->id == ZigTypeIdComptimeInt) { + return ir_analyze_int_to_c_ptr(ira, source_instr, value, wanted_type); + } + } + // cast from undefined to anything if (actual_type->id == ZigTypeIdUndefined) { return ir_analyze_undefined_to_anything(ira, source_instr, value, wanted_type); @@ -20674,17 +20744,38 @@ static Error buf_read_value_bytes(IrAnalyze *ira, CodeGen *codegen, AstNode *sou zig_unreachable(); } -static IrInstruction *ir_analyze_instruction_bit_cast(IrAnalyze *ira, IrInstructionBitCast *instruction) { - Error err; - IrInstruction *dest_type_value = instruction->dest_type->child; - ZigType *dest_type = ir_resolve_type(ira, dest_type_value); - if (type_is_invalid(dest_type)) - return ira->codegen->invalid_instruction; +static bool type_can_bit_cast(ZigType *t) { + switch (t->id) { + case ZigTypeIdInvalid: + zig_unreachable(); + case ZigTypeIdMetaType: + case ZigTypeIdOpaque: + case ZigTypeIdBoundFn: + case ZigTypeIdArgTuple: + case ZigTypeIdNamespace: + case ZigTypeIdUnreachable: + case ZigTypeIdComptimeFloat: + case ZigTypeIdComptimeInt: + case ZigTypeIdUndefined: + case ZigTypeIdNull: + case ZigTypeIdPointer: + return false; + default: + // TODO list these types out explicitly, there are probably some other invalid ones here + return true; + } +} + +static IrInstruction *ir_analyze_bit_cast(IrAnalyze *ira, IrInstruction *source_instr, IrInstruction *value, + ZigType *dest_type) +{ + Error err; - IrInstruction *value = instruction->value->child; ZigType *src_type = value->value.type; - if (type_is_invalid(src_type)) - return ira->codegen->invalid_instruction; + assert(get_codegen_ptr_type(src_type) == nullptr); + assert(type_can_bit_cast(src_type)); + assert(get_codegen_ptr_type(dest_type) == nullptr); + assert(type_can_bit_cast(dest_type)); if ((err = type_resolve(ira->codegen, dest_type, ResolveStatusSizeKnown))) return ira->codegen->invalid_instruction; @@ -20692,60 +20783,11 @@ static IrInstruction *ir_analyze_instruction_bit_cast(IrAnalyze *ira, IrInstruct if ((err = type_resolve(ira->codegen, src_type, ResolveStatusSizeKnown))) return ira->codegen->invalid_instruction; - if (get_codegen_ptr_type(src_type) != nullptr) { - ir_add_error(ira, value, - buf_sprintf("unable to @bitCast from pointer type '%s'", buf_ptr(&src_type->name))); - return ira->codegen->invalid_instruction; - } - - switch (src_type->id) { - case ZigTypeIdInvalid: - case ZigTypeIdMetaType: - case ZigTypeIdOpaque: - case ZigTypeIdBoundFn: - case ZigTypeIdArgTuple: - case ZigTypeIdNamespace: - case ZigTypeIdUnreachable: - case ZigTypeIdComptimeFloat: - case ZigTypeIdComptimeInt: - case ZigTypeIdUndefined: - case ZigTypeIdNull: - ir_add_error(ira, dest_type_value, - buf_sprintf("unable to @bitCast from type '%s'", buf_ptr(&src_type->name))); - return ira->codegen->invalid_instruction; - default: - break; - } - - if (get_codegen_ptr_type(dest_type) != nullptr) { - ir_add_error(ira, dest_type_value, - buf_sprintf("unable to @bitCast to pointer type '%s'", buf_ptr(&dest_type->name))); - return ira->codegen->invalid_instruction; - } - - switch (dest_type->id) { - case ZigTypeIdInvalid: - case ZigTypeIdMetaType: - case ZigTypeIdOpaque: - case ZigTypeIdBoundFn: - case ZigTypeIdArgTuple: - case ZigTypeIdNamespace: - case ZigTypeIdUnreachable: - case ZigTypeIdComptimeFloat: - case ZigTypeIdComptimeInt: - case ZigTypeIdUndefined: - case ZigTypeIdNull: - ir_add_error(ira, dest_type_value, - buf_sprintf("unable to @bitCast to type '%s'", buf_ptr(&dest_type->name))); - return ira->codegen->invalid_instruction; - default: - break; - } uint64_t dest_size_bytes = type_size(ira->codegen, dest_type); uint64_t src_size_bytes = type_size(ira->codegen, src_type); if (dest_size_bytes != src_size_bytes) { - ir_add_error(ira, &instruction->base, + ir_add_error(ira, source_instr, buf_sprintf("destination type '%s' has size %" ZIG_PRI_u64 " but source type '%s' has size %" ZIG_PRI_u64, buf_ptr(&dest_type->name), dest_size_bytes, buf_ptr(&src_type->name), src_size_bytes)); @@ -20755,7 +20797,7 @@ static IrInstruction *ir_analyze_instruction_bit_cast(IrAnalyze *ira, IrInstruct uint64_t dest_size_bits = type_size_bits(ira->codegen, dest_type); uint64_t src_size_bits = type_size_bits(ira->codegen, src_type); if (dest_size_bits != src_size_bits) { - ir_add_error(ira, &instruction->base, + ir_add_error(ira, source_instr, buf_sprintf("destination type '%s' has %" ZIG_PRI_u64 " bits but source type '%s' has %" ZIG_PRI_u64 " bits", buf_ptr(&dest_type->name), dest_size_bits, buf_ptr(&src_type->name), src_size_bits)); @@ -20767,20 +20809,86 @@ static IrInstruction *ir_analyze_instruction_bit_cast(IrAnalyze *ira, IrInstruct if (!val) return ira->codegen->invalid_instruction; - IrInstruction *result = ir_const(ira, &instruction->base, dest_type); + IrInstruction *result = ir_const(ira, source_instr, dest_type); uint8_t *buf = allocate_nonzero(src_size_bytes); buf_write_value_bytes(ira->codegen, buf, val); - if ((err = buf_read_value_bytes(ira, ira->codegen, instruction->base.source_node, buf, &result->value))) + if ((err = buf_read_value_bytes(ira, ira->codegen, source_instr->source_node, buf, &result->value))) return ira->codegen->invalid_instruction; return result; } - IrInstruction *result = ir_build_bit_cast(&ira->new_irb, instruction->base.scope, - instruction->base.source_node, nullptr, value); + IrInstruction *result = ir_build_bit_cast(&ira->new_irb, source_instr->scope, + source_instr->source_node, nullptr, value); result->value.type = dest_type; return result; } +static IrInstruction *ir_analyze_instruction_bit_cast(IrAnalyze *ira, IrInstructionBitCast *instruction) { + IrInstruction *dest_type_value = instruction->dest_type->child; + ZigType *dest_type = ir_resolve_type(ira, dest_type_value); + if (type_is_invalid(dest_type)) + return ira->codegen->invalid_instruction; + + IrInstruction *value = instruction->value->child; + ZigType *src_type = value->value.type; + if (type_is_invalid(src_type)) + return ira->codegen->invalid_instruction; + + if (get_codegen_ptr_type(src_type) != nullptr) { + ir_add_error(ira, value, + buf_sprintf("unable to @bitCast from pointer type '%s'", buf_ptr(&src_type->name))); + return ira->codegen->invalid_instruction; + } + + if (!type_can_bit_cast(src_type)) { + ir_add_error(ira, dest_type_value, + buf_sprintf("unable to @bitCast from type '%s'", buf_ptr(&src_type->name))); + return ira->codegen->invalid_instruction; + } + + if (get_codegen_ptr_type(dest_type) != nullptr) { + ir_add_error(ira, dest_type_value, + buf_sprintf("unable to @bitCast to pointer type '%s'", buf_ptr(&dest_type->name))); + return ira->codegen->invalid_instruction; + } + + if (!type_can_bit_cast(dest_type)) { + ir_add_error(ira, dest_type_value, + buf_sprintf("unable to @bitCast to type '%s'", buf_ptr(&dest_type->name))); + return ira->codegen->invalid_instruction; + } + + return ir_analyze_bit_cast(ira, &instruction->base, value, dest_type); +} + +static IrInstruction *ir_analyze_int_to_ptr(IrAnalyze *ira, IrInstruction *source_instr, IrInstruction *target, + ZigType *ptr_type) +{ + assert(get_src_ptr_type(ptr_type) != nullptr); + assert(type_has_bits(ptr_type)); + + IrInstruction *casted_int = ir_implicit_cast(ira, target, ira->codegen->builtin_types.entry_usize); + if (type_is_invalid(casted_int->value.type)) + return ira->codegen->invalid_instruction; + + if (instr_is_comptime(casted_int)) { + ConstExprValue *val = ir_resolve_const(ira, casted_int, UndefBad); + if (!val) + return ira->codegen->invalid_instruction; + + IrInstruction *result = ir_const(ira, source_instr, ptr_type); + result->value.data.x_ptr.special = ConstPtrSpecialHardCodedAddr; + result->value.data.x_ptr.mut = ConstPtrMutRuntimeVar; + result->value.data.x_ptr.data.hard_coded_addr.addr = bigint_as_unsigned(&val->data.x_bigint); + return result; + } + + IrInstruction *result = ir_build_int_to_ptr(&ira->new_irb, source_instr->scope, + source_instr->source_node, nullptr, casted_int); + result->value.type = ptr_type; + return result; +} + static IrInstruction *ir_analyze_instruction_int_to_ptr(IrAnalyze *ira, IrInstructionIntToPtr *instruction) { Error err; IrInstruction *dest_type_value = instruction->dest_type->child; @@ -20802,30 +20910,12 @@ static IrInstruction *ir_analyze_instruction_int_to_ptr(IrAnalyze *ira, IrInstru return ira->codegen->invalid_instruction; } + IrInstruction *target = instruction->target->child; if (type_is_invalid(target->value.type)) return ira->codegen->invalid_instruction; - IrInstruction *casted_int = ir_implicit_cast(ira, target, ira->codegen->builtin_types.entry_usize); - if (type_is_invalid(casted_int->value.type)) - return ira->codegen->invalid_instruction; - - if (instr_is_comptime(casted_int)) { - ConstExprValue *val = ir_resolve_const(ira, casted_int, UndefBad); - if (!val) - return ira->codegen->invalid_instruction; - - IrInstruction *result = ir_const(ira, &instruction->base, dest_type); - result->value.data.x_ptr.special = ConstPtrSpecialHardCodedAddr; - result->value.data.x_ptr.mut = ConstPtrMutRuntimeVar; - result->value.data.x_ptr.data.hard_coded_addr.addr = bigint_as_unsigned(&val->data.x_bigint); - return result; - } - - IrInstruction *result = ir_build_int_to_ptr(&ira->new_irb, instruction->base.scope, - instruction->base.source_node, nullptr, casted_int); - result->value.type = dest_type; - return result; + return ir_analyze_int_to_ptr(ira, &instruction->base, target, dest_type); } static IrInstruction *ir_analyze_instruction_decl_ref(IrAnalyze *ira, diff --git a/src/parser.cpp b/src/parser.cpp index 1c1af87c51..160a7268b0 100644 --- a/src/parser.cpp +++ b/src/parser.cpp @@ -2779,6 +2779,7 @@ static AstNode *ast_parse_array_type_start(ParseContext *pc) { // <- ASTERISK // / ASTERISK2 // / LBRACKET ASTERISK RBRACKET +// / LBRACKET ASTERISK C RBRACKET static AstNode *ast_parse_ptr_type_start(ParseContext *pc) { Token *asterisk = eat_token_if(pc, TokenIdStar); if (asterisk != nullptr) { @@ -2804,6 +2805,13 @@ static AstNode *ast_parse_ptr_type_start(ParseContext *pc) { return res; } + Token *cptr = eat_token_if(pc, TokenIdBracketStarCBracket); + if (cptr != nullptr) { + AstNode *res = ast_create_node(pc, NodeTypePointerType, cptr); + res->data.pointer_type.star_token = cptr; + return res; + } + return nullptr; } diff --git a/src/tokenizer.cpp b/src/tokenizer.cpp index 3acd605748..9ff6ed3bbe 100644 --- a/src/tokenizer.cpp +++ b/src/tokenizer.cpp @@ -221,6 +221,7 @@ enum TokenizeState { TokenizeStateError, TokenizeStateLBracket, TokenizeStateLBracketStar, + TokenizeStateLBracketStarC, }; @@ -846,7 +847,6 @@ void tokenize(Buf *buf, Tokenization *out) { switch (c) { case '*': t.state = TokenizeStateLBracketStar; - set_token_id(&t, t.cur_tok, TokenIdBracketStarBracket); break; default: // reinterpret as just an lbracket @@ -857,6 +857,21 @@ void tokenize(Buf *buf, Tokenization *out) { } break; case TokenizeStateLBracketStar: + switch (c) { + case 'c': + t.state = TokenizeStateLBracketStarC; + set_token_id(&t, t.cur_tok, TokenIdBracketStarCBracket); + break; + case ']': + set_token_id(&t, t.cur_tok, TokenIdBracketStarBracket); + end_token(&t); + t.state = TokenizeStateStart; + break; + default: + invalid_char_error(&t, c); + } + break; + case TokenizeStateLBracketStarC: switch (c) { case ']': end_token(&t); @@ -1491,6 +1506,7 @@ void tokenize(Buf *buf, Tokenization *out) { case TokenizeStateLineStringContinue: case TokenizeStateLineStringContinueC: case TokenizeStateLBracketStar: + case TokenizeStateLBracketStarC: tokenize_error(&t, "unexpected EOF"); break; case TokenizeStateLineComment: @@ -1528,6 +1544,7 @@ const char * token_name(TokenId id) { case TokenIdBitShiftRightEq: return ">>="; case TokenIdBitXorEq: return "^="; case TokenIdBracketStarBracket: return "[*]"; + case TokenIdBracketStarCBracket: return "[*c]"; case TokenIdCharLiteral: return "CharLiteral"; case TokenIdCmpEq: return "=="; case TokenIdCmpGreaterOrEq: return ">="; diff --git a/src/tokenizer.hpp b/src/tokenizer.hpp index 17f36699b3..62117b5779 100644 --- a/src/tokenizer.hpp +++ b/src/tokenizer.hpp @@ -29,6 +29,7 @@ enum TokenId { TokenIdBitShiftRightEq, TokenIdBitXorEq, TokenIdBracketStarBracket, + TokenIdBracketStarCBracket, TokenIdCharLiteral, TokenIdCmpEq, TokenIdCmpGreaterOrEq, diff --git a/test/stage1/behavior/pointers.zig b/test/stage1/behavior/pointers.zig index 47b19700ee..6969e151df 100644 --- a/test/stage1/behavior/pointers.zig +++ b/test/stage1/behavior/pointers.zig @@ -42,3 +42,9 @@ test "double pointer parsing" { fn PtrOf(comptime T: type) type { return *T; } + +test "assigning integer to C pointer" { + var x: i32 = 0; + var ptr: [*c]u8 = 0; + var ptr2: [*c]u8 = x; +}