[Blue1977]

Description

This PR fixes an issue in scripts/update_lit_checks.py where the lit check updater incorrectly counts unbalanced parentheses inside strings (such as "hello(" or "hello)") as part of the WebAssembly syntax structure. This bug causes CHECK assertions to map incorrectly to subsequent functions inside data segments.

Solution

Modified find_end to handle string boundaries and escaped characters by implementing in_string and escaped state tracking. Parentheses inside string literals are now successfully skipped, ensuring accurate structure parsing.

Validation Results (WSL Ubuntu)

Local regression test suite executed using test/unit/test_update_lit_checks_strings.py:

• before log status: 5 expected subcase failures detected (string='hello(', hello)', hello((), hello\"(, hello\\().
• after log status: OK (Ran 1 test containing all subcases successfully).

No issues or PR numbers are linked as this stems from an internal code FIXME.

[gemini-code-assist]

Code Review

This pull request updates scripts/update_lit_checks.py to correctly handle parentheses nested inside string literals, including escaped quotes and backslashes, and adds a comprehensive unit test suite to verify this behavior. A review comment identifies an off-by-one bug in the loop termination logic of find_end that could truncate the closing parenthesis if it is the last character in the string, and provides a suggestion to fix it.

[gemini-code-assist]

There is an off-by-one bug in the loop termination logic. If the matching closing parenthesis is the very last character of the module string, the loop terminates naturally without executing the next iteration where depth == 0 is checked. As a result, end is returned as len(module) - 1 instead of len(module), which incorrectly truncates the closing parenthesis when slicing.

We can fix this and simplify the logic by checking if depth == 0 immediately after decrementing it, and returning end + 1 directly from within the loop.

Suggested change

	for end in range(start + 1, len(module)):
	if depth == 0:
	break
	if in_string:
	if escaped:
	escaped = False
	elif module[end] == '\\':
	escaped = True
	elif module[end] == '"':
	in_string = False
	elif module[end] == '"':
	in_string = True
	elif module[end] == '(':
	depth += 1
	elif module[end] == ')':
	depth -= 1
	return end
	for end in range(start + 1, len(module)):
	if in_string:
	if escaped:
	escaped = False
	elif module[end] == '\\':
	escaped = True
	elif module[end] == '"':
	in_string = False
	elif module[end] == '"':
	in_string = True
	elif module[end] == '(':
	depth += 1
	elif module[end] == ')':
	depth -= 1
	if depth == 0:
	return end + 1
	return len(module)

[tlively]

Sounds like it's worth fixing this at the same time.

[tlively]

Thanks for the PR!

[tlively]

Rather than a Python unit test, let's just have a small .wast lit test that contains parens inside strings and make sure that the auto-update script can properly generate output for it.

[tlively]

Sounds like it's worth fixing this at the same time.