[RKOTLIN-1089] Guard decoding issues when core logs invalid utf-8 messages #1792
Conversation
| for (std::string::size_type i = 0; i < str.size(); ++i) | ||
| ret << " 0x" << std::hex << std::setfill('0') << std::setw(2) << (int)s[i]; | ||
| ret << "; "; | ||
| ret << "in_begin = " << in_begin << "; "; | ||
| ret << "in_end = " << in_end << "; "; | ||
| ret << "in_begin = " << (void*) in_begin << "; "; |
There was a problem hiding this comment.
why casting to opaque pointer?
There was a problem hiding this comment.
If treating in_begin and in_end as char * they will be considered as 0-terminated strings and printed to the stream ... in which case we end up including the invalid UTF-8 bytes in the message - Just as for ret << "StringData.data = " << str << "; ";. Want to avoid having invalid unicode in the exception message as spelled out in #1792 (comment). Just including the pointers to give some insights if we were to try to understand and reproduce an issue based on one of these exceptions.
| @@ -82,13 +82,12 @@ static std::string string_to_hex(const std::string& message, realm::StringData& | |||
| ret << "error_code = " << error_code << "; "; | |||
| ret << "retcode = " << retcode << "; "; | |||
| ret << "StringData.size = " << str.size() << "; "; | |||
| ret << "StringData.data = " << str << "; "; | |||
There was a problem hiding this comment.
We're not printing data?
There was a problem hiding this comment.
This string_to_hex is only used to make a hex-dump of the data when we fail to decode it from UTF-8. So don't think it makes sense to include the non-decodable string itself. The result of this method is used as the message of an exception and we cannot pass that on to the JVM if the message itself is not valid unicode.
| jstring j_message = NULL; | ||
| try { | ||
| j_message = to_jstring(jenv, message); | ||
| } catch (RuntimeError exception) { |
There was a problem hiding this comment.
Did you observe this on Android & JVM?
There was a problem hiding this comment.
It is on both platforms
rorbech
force-pushed
the
cr/fix-logger-encoding-issues
branch
from
db2e957
to
7af0a91
Compare
We can actually derive some context from it. The |
rorbech
changed the title
Closes #1760