From 1708134f4bf3118d25ed3e95e3037b3a6eb479e3 Mon Sep 17 00:00:00 2001 From: Kyle Lin Date: Thu, 20 Mar 2025 17:25:36 +0800 Subject: [PATCH] Fix dereference behavior on mixed operators Previously, using subscript operator (literally "[]") with arrow operator (literally "->") or dot operator (literally ".") would cause incorrect dereference result due to inappropriate delayed dereference strategy. For example, considering "data->raw[0]", this would evaluate to "data[0]", same for "data.raw[0]". In this patch, by adding dereference instruction when encountered subscript operator after parsed either arrow or dot operator, in other word, after accessed struct's member, this corrects the final evaluated address. Close #164, close #165, close #181. --- src/globals.c | 4 +--- src/parser.c | 47 ++++++++++++++++++++++++++++++++--------------- tests/driver.sh | 43 +++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 76 insertions(+), 18 deletions(-) diff --git a/src/globals.c b/src/globals.c index 18bee333..f3cb4b12 100644 --- a/src/globals.c +++ b/src/globals.c @@ -250,9 +250,7 @@ void hashmap_free(hashmap_t *map) next = cur->next; free(cur->key); free(cur->val); - /* FIXME: Remove this if-clause will cause double free error */ - if (cur != map->buckets[0]) - free(cur); + free(cur); cur = next; } } diff --git a/src/parser.c b/src/parser.c index 1c341868..4b3f9c3b 100644 --- a/src/parser.c +++ b/src/parser.c @@ -750,7 +750,7 @@ void read_lvalue(lvalue_t *lvalue, var_t *var, block_t *parent, basic_block_t **bb, - int eval, + bool eval, opcode_t op); /* Maintain a stack of expression values and operators, depending on next @@ -805,7 +805,7 @@ void read_expr_operand(block_t *parent, basic_block_t **bb) lex_peek(T_identifier, token); var_t *var = find_var(token, parent); - read_lvalue(&lvalue, var, parent, bb, 0, OP_generic); + read_lvalue(&lvalue, var, parent, bb, false, OP_generic); if (!lvalue.is_reference) { ph1_ir = add_ph1_ir(OP_address_of); @@ -825,7 +825,7 @@ void read_expr_operand(block_t *parent, basic_block_t **bb) int open_bracket = lex_accept(T_open_bracket); lex_peek(T_identifier, token); var_t *var = find_var(token, parent); - read_lvalue(&lvalue, var, parent, bb, 1, OP_generic); + read_lvalue(&lvalue, var, parent, bb, true, OP_generic); if (open_bracket) lex_expect(T_close_bracket); @@ -962,7 +962,7 @@ void read_expr_operand(block_t *parent, basic_block_t **bb) } else if (var) { /* evalue lvalue expression */ lvalue_t lvalue; - read_lvalue(&lvalue, var, parent, bb, 1, prefix_op); + read_lvalue(&lvalue, var, parent, bb, true, prefix_op); /* is it an indirect call with function pointer? */ if (lex_peek(T_open_bracket, NULL)) { @@ -1234,13 +1234,13 @@ void read_lvalue(lvalue_t *lvalue, var_t *var, block_t *parent, basic_block_t **bb, - int eval, + bool eval, opcode_t prefix_op) { ph1_ir_t *ph1_ir; var_t *vd; - int is_address_got = 0; - int is_member = 0; + bool is_address_got = false; + bool is_member = false; /* already peeked and have the variable */ lex_expect(T_identifier); @@ -1260,6 +1260,23 @@ void read_lvalue(lvalue_t *lvalue, while (lex_peek(T_open_square, NULL) || lex_peek(T_arrow, NULL) || lex_peek(T_dot, NULL)) { if (lex_accept(T_open_square)) { + /* if subscripted member's is not yet resolved, dereference to + * resolve base address. + * e.g., dereference of "->" in "data->raw[0]" would be performed + * here. + */ + if (lvalue->is_reference && lvalue->is_ptr && is_member) { + ph1_ir = add_ph1_ir(OP_read); + ph1_ir->src0 = opstack_pop(); + vd = require_var(parent); + strcpy(vd->var_name, gen_name()); + ph1_ir->dest = vd; + opstack_push(vd); + ph1_ir->size = 4; + add_insn(parent, *bb, OP_read, ph1_ir->dest, ph1_ir->src0, NULL, + ph1_ir->size, NULL); + } + /* var must be either a pointer or an array of some type */ if (var->is_ptr == 0 && var->array_size == 0) error("Cannot apply square operator to non-pointer"); @@ -1303,8 +1320,8 @@ void read_lvalue(lvalue_t *lvalue, ph1_ir->src1, 0, NULL); lex_expect(T_close_square); - is_address_got = 1; - is_member = 1; + is_address_got = true; + is_member = true; lvalue->is_reference = true; } else { char token[MAX_ID_LEN]; @@ -1313,7 +1330,7 @@ void read_lvalue(lvalue_t *lvalue, /* resolve where the pointer points at from the calculated * address in a structure. */ - if (is_member == 1) { + if (is_member) { ph1_ir = add_ph1_ir(OP_read); ph1_ir->src0 = opstack_pop(); vd = require_var(parent); @@ -1327,7 +1344,7 @@ void read_lvalue(lvalue_t *lvalue, } else { lex_expect(T_dot); - if (is_address_got == 0) { + if (!is_address_got) { ph1_ir = add_ph1_ir(OP_address_of); ph1_ir->src0 = opstack_pop(); vd = require_var(parent); @@ -1337,7 +1354,7 @@ void read_lvalue(lvalue_t *lvalue, add_insn(parent, *bb, OP_address_of, ph1_ir->dest, ph1_ir->src0, NULL, 0, NULL); - is_address_got = 1; + is_address_got = true; } } @@ -1376,8 +1393,8 @@ void read_lvalue(lvalue_t *lvalue, add_insn(parent, *bb, OP_add, ph1_ir->dest, ph1_ir->src0, ph1_ir->src1, 0, NULL); - is_address_got = 1; - is_member = 1; + is_address_got = true; + is_member = true; } } @@ -1860,7 +1877,7 @@ bool read_body_assignment(char *token, int size = 0; /* has memory address that we want to set */ - read_lvalue(&lvalue, var, parent, bb, 0, OP_generic); + read_lvalue(&lvalue, var, parent, bb, false, OP_generic); size = lvalue.size; if (lex_accept(T_increment)) { diff --git a/tests/driver.sh b/tests/driver.sh index c80b5005..d0cdd2a2 100755 --- a/tests/driver.sh +++ b/tests/driver.sh @@ -433,6 +433,49 @@ int main() { } EOF +# Mixed subscript and arrow / dot operators, +# excerpted and modified from issue #165 +try_output 0 "DDDDDDMMMEEE1" << EOF +#include +#include + +char a[100]; + +typedef struct { + char *raw; +} data_t; + +int main() { + strcpy(a, "DATA"); + data_t *data = malloc(sizeof(data_t)); + data->raw = a; + data_t data2; + data2.raw = a; + char *raw = data->raw; + char *raw2 = data2.raw; + /* mixed arrow / dot with subscript operators dereference */ + printf("%c", a[0]); + printf("%c", raw[0]); + printf("%c", data->raw[0]); + printf("%c", a[0]); + printf("%c", raw2[0]); + printf("%c", data2.raw[0]); + /* mixed arrow / dot with subscript operators assignment */ + data2.raw[0] = 'M'; + data->raw[1] = 'E'; + printf("%c", a[0]); + printf("%c", raw[0]); + printf("%c", data->raw[0]); + printf("%c", a[1]); + printf("%c", raw2[1]); + printf("%c", data2.raw[1]); + /* their addresses should be same */ + printf("%d", &data2.raw[0] == &data->raw[0]); + free(data); + return 0; +} +EOF + # global initialization try_ 20 << EOF int a = 5 * 2;