Open stevemcleod opened 6 years ago
This is perhaps a non-problem, at least as far as Spark is concerned.
Jetty throws these exceptions in response to invalid URIs. I think what I'm seeing is Jetty refusing to accept invalid URIs.
An example bad URI is example.com/?%%0
We see a lot of these exceptions in our logs. It could be that these bad URIs are in the arsenal of "script kiddy" tools.
I think this could be closed.
We get occasional errors in production as follows:
I think the problem lies in two different implementations of NotUtf8Exception being used:
spark.utils.urldecoding.Utf8Appendable:91
throws an instance of org.eclipse.jetty.util.Utf8Appendable.NotUtf8ExceptionBut
spark.utils.urldecoding.UrlDecode:91
catches spark.utils.urldecoding.Utf8Appendable.NotUtf8ExceptionAs far as I can tell, spark.utils.urldecoding.Utf8Appendable.NotUtf8Exception is never instantiated at all in the Spark code base.
As a result, the fallback
decodeISO88591Path
method inspark.utils.urldecoding.UrlDecode
for handling ISO-8859-1 paths is never used.I'm happy to submit a pull request for this myself. But it would be nice if someone with a better understanding of the relevant code can confirm whether my analysis is correct.