fix(parser): add destructor directives to prevent memory leaks

[fzipi]

what

The Bison parser was missing %destructor directives for all semantic types using unique_ptr. When parsing errors occurred (via YYERROR), the parser would discard semantic values from the stack without properly destroying them, causing memory leaks detected by ASAN.

why

Added %destructor directives for:

• std::unique_ptractions::Action
• std::unique_ptr
• std::unique_ptr<std::vector<std::unique_ptractions::Action>>
• std::unique_ptr
• std::unique_ptr
• std::unique_ptr<std::vector<std::unique_ptr>>

The empty braces {} in the destructor directives are intentional. With C++ variant semantic values (api.value.type variant), Bison automatically calls the destructors of objects in the variant when discarding semantic values during error recovery.

This fixes memory leaks that occurred in scenarios such as:

1. When ACTION_INIT macro fails and calls YYERROR (lines 885, 892)
2. When rule creation fails and calls YYERROR (line 1130)
3. Any other parsing error that causes the parser to discard semantic values during error recovery

references

• Fixes Sanitizer reports (ASAN, UBSAN) #3448

[Copilot]

Pull request overview

Adds Bison %destructor directives for C++ variant semantic values that hold std::unique_ptr, aiming to ensure semantic values are properly cleaned up when the parser discards symbols during error recovery (e.g., via YYERROR), addressing ASAN-reported leaks (Fixes #3448).

Changes:

• Introduces %destructor directives for several std::unique_ptr semantic types in the Seclang Bison grammar.
• Adds inline comments documenting the intent and rationale for empty destructor bodies under api.value.type variant.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 4 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[fzipi]

The #line directives referencing "seclang-parser.tab.cc" are cosmetic and fine. Here's why:

What #line directives do:

• They tell the compiler what source file and line number to report in error messages
• They're automatically generated by Bison during parser generation
• .tab.cc is Bison's default intermediate naming convention

Why it says .tab.cc even though the file is .cc:

• Bison internally uses .tab.cc as the basename during generation
• Even when using -o seclang-parser.cc, the #line directives keep the .tab name
• This is standard Bison behavior and affects many projects

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
20 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[Copilot]

Pull request overview

Copilot reviewed 3 out of 4 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (3)

src/parser/seclang-scanner.cc:8694

• num_to_read was changed to yy_size_t, but it's computed from an int expression (yy_buf_size - number_to_move - 1). If that expression is negative, it will wrap to a huge unsigned value and the while (num_to_read <= 0) growth loop will never run, risking an out-of-bounds write in YY_INPUT. Keep this value in a signed type (e.g., int/ptrdiff_t) or restructure the calculation so the “no space left” condition is checked before converting to an unsigned type.

			yy_size_t num_to_read =
			YY_CURRENT_BUFFER_LVALUE->yy_buf_size - number_to_move - 1;

		while ( num_to_read <= 0 )
			{ /* Not enough room in the buffer - grow it. */

src/parser/seclang-scanner.cc:8709

• The overflow/resize guard here no longer works as intended after switching new_size to yy_size_t: if (new_size <= 0) is effectively dead code for a size type, so the fallback growth path is unreachable. Consider keeping new_size signed (or computing in yy_size_t and checking for overflow explicitly) so buffer growth remains correct for large yy_buf_size values.

				yy_size_t new_size = b->yy_buf_size * 2;

				if ( new_size <= 0 )
					b->yy_buf_size += b->yy_buf_size / 8;
				else
					b->yy_buf_size *= 2;

src/parser/seclang-scanner.cc:9384

• This yyless redefinition was updated to use yy_size_t, but the earlier #define yyless(n) (near line ~215) still uses int yyless_macro_arg = (n);. With yyleng now being yy_size_t, calls like yyless(yyleng) can still truncate via the first macro definition. Update the first yyless macro to use yy_size_t as well to keep behavior consistent.

#undef yyless
#define yyless(n) \
	do \
		{ \
		/* Undo effects of setting up yytext. */ \
        yy_size_t yyless_macro_arg = (n); \
        YY_LESS_LINENO(yyless_macro_arg);\
		yytext[yyleng] = (yy_hold_char); \
		(yy_c_buf_p) = yytext + yyless_macro_arg; \
		(yy_hold_char) = *(yy_c_buf_p); \
		*(yy_c_buf_p) = '\0'; \
		yyleng = yyless_macro_arg; \

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[airween]

LGTM

[airween]

Thank you @fzipi!