Julian Reschke created OAK-7279:
-----------------------------------
Summary: segment-tar update from java 7 to java 8 may break
persisted names using invalid characters
Key: OAK-7279
URL: https://issues.apache.org/jira/browse/OAK-7279
Project: Jackrabbit Oak
Issue Type: Bug
Components: segment-tar
Reporter: Julian Reschke
segment-tar relies on {{String.getBytes()}} when persisting strings such as
item names.
The problem is that the behavior for this has been changed in Java 8 with
respect to invalid strings (here: null characters and unpaired surrogates).
In Java 7, these would roundtrip, as Java was using the so-called "modified
UTF-8" encoding (see
https://docs.oracle.com/javase/6/docs/api/java/io/DataInput.html#modified-utf-8).
This will produce byte sequence that are *not* valid UTF-8.
Java 7 will read them back, but Java 8 will map the non-conforming byte
sequences to the Unicode replacement character. Note that in particular,
multiple child entries might get identical names as a consequence.
I'm not sure about the severity of this, and whether something needs to be done
about it. AFAIC, this is another good reason to reject invalid strings as early
as possible in the stack.
cc [~mduerig]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
