[libc++] regex uncaught exception using libc++, but libstdc++ not

[Zhenhang1213]

code:

#include <iostream>
#include <regex>

int main() {
        const char* pattern = "\\$\\_se";
        std::regex regexObj(pattern);
        const char* testString = "$_se";
        if (std::regex_match(testString, regexObj)) {
        std::cout << "匹配成功!" << std::endl;
    } else {
        std::cout << "匹配失败!" << std::endl;
    }
        return 0;
}

error:

libc++abi: terminating due to uncaught exception of type std::__1::regex_error: The expression contained an invalid escaped character, or a trailing escape.
Program terminated with signal: SIGSEGV

demo:
https://godbolt.org/z/szGjjr5bs

[hstk30-hw]

CC @ldionne

[frederick-vs-ja]

The exception is thrown from here (in basic_regex<_CharT, _Traits>::__parse_character_escape):

llvm-project/libcxx/include/regex

Lines 3957 to 3965 in 1594fa8

	default:
	if (*__first != '_' && !__traits_.isctype(*__first, ctype_base::alnum)) {
	if (__str)
	*__str = *__first;
	else
	__push_char(*__first);
	++__first;
	} else
	std::__throw_regex_error<regex_constants::error_escape>();

Looks like that the input sequence \_ is treated as an atom with an invalid atom escape.

[frederick-vs-ja]

I guess the condition should be changed from *__first != '_' && !__traits_.isctype(*__first, ctype_base::alnum) to !__traits_.isctype(*__first, ctype_base::alnum). It seems that the underscore character (_) shouldn't be special when parsing character escape.

[Zhenhang1213]

I guess the condition should be changed from *__first != '_' && !__traits_.isctype(*__first, ctype_base::alnum) to !__traits_.isctype(*__first, ctype_base::alnum). It seems that the underscore character (_) shouldn't be special when parsing character escape.

Yes, I don't think this underscore character (_) should be captured by exceptions, it's unreasonable here.