Re: [PATCH v2 0/4] MR585: Improve is_gecko_path() for handling non-ASCII characters.

[Jacek Caban (＠jacek)]

Jacek Caban (@jacek) commented about dlls/kernelbase/path.c:
> {
>              INT ih;
>              WCHAR buf[5] = L"0x";
> +
>              memcpy(buf + 2, src + 1, 2*sizeof(WCHAR));
>              buf[4] = 0;
>              StrToIntExW(buf, STIF_SUPPORT_HEX, &ih);
> -            next = (WCHAR) ih;
>              src += 2; /* Advance to end of escape */
> +
> +            if (flags & URL_UNESCAPE_AS_UTF8)
> +            {
> +                utf8_buf[utf8_len++] = ih;
> +                utf16_len = MultiByteToWideChar(CP_UTF8, MB_ERR_INVALID_CHARS, utf8_buf, utf8_len, NULL, 0);
> +                if (!utf16_len)
> +                    continue;

This doesn't seem reliable. For example, if there is non-escaped char between escaped multi-byte values, you will end up combining characters surrounding non-escaped one. See JSGlobal_decodeURI for an example how it can be handled.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/585#note_5938