Forbid \C in UTF-8 patterns

[arnaud-lb]

Possible fix for GH-21134.

\C results in undefined, potentially unsafe behavior in UTF-8 mode, so error out in case the escape sequence is used in UTF-8 mode.

[devnexen]

would need to document it somehow (UPGRADING ?) wdyt ?

[devnexen]

yes the default message is a bit vague in this context, makes sense.

[devnexen]

I know enough pcre2 to see it s LGTM but cc @ndossche just in case :)

[ndossche]

I think the strlcpy call is ugly, but I won't object. Why not smth like this (and possibly rename error to error_buf):

Suggested change

	if (errnumber == PCRE2_ERROR_BACKSLASH_C_CALLER_DISABLED) {
	strlcpy((char*)error, "using \\C is incompatible with the 'u' modifier", sizeof(error));
	} else {
	pcre2_get_error_message(errnumber, error, sizeof(error));
	}
	php_error_docref(NULL,E_WARNING, "Compilation failed: %s at offset %zu", error, erroffset);
	const char *err_msg = (const char *) error;
	if (errnumber == PCRE2_ERROR_BACKSLASH_C_CALLER_DISABLED) {
	err_msg = "using \\C is incompatible with the 'u' modifier";
	} else {
	pcre2_get_error_message(errnumber, error, sizeof(error));
	}
	php_error_docref(NULL,E_WARNING, "Compilation failed: %s at offset %zu", err_msg, erroffset);