[
https://issues.apache.org/jira/browse/IGNITE-3140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15286028#comment-15286028
]
Denis Magda commented on IGNITE-3140:
-------------------------------------
Igor,
The new serialization algorithm on Java side serializes all symbols that are
bigger than 0x07FF in 3 bytes.
It means that if there is a valid surrogate pair in a String like this one
{{0xD801, 0xDC37}} then the new algorithm will use 6 bytes to code it while
basic UTF-8 coders/decoders will use only 4 bytes. C++ side won't be able to
properly deserialize {{0xD801, 0xDC37}} on its side because it will be encoded
in 6 bytes.
Try to serialize this String on C++ side. It should be encoded in 4 bytes while
the new Java algorithm encodes it in 6 bytes.
{noformat}
str = new String(new char[] {0xD801, 0xDC37});
{noformat}
> C++: UTF-16 surrogate symbols are not serialized properly
> ---------------------------------------------------------
>
> Key: IGNITE-3140
> URL: https://issues.apache.org/jira/browse/IGNITE-3140
> Project: Ignite
> Issue Type: Bug
> Components: platforms
> Affects Versions: 1.5.0.final
> Reporter: Denis Magda
> Assignee: Vladimir Ozerov
> Fix For: 1.6
>
>
> There is an issue with serialization of a surrogate symbol with
> {{BinaryMarshaller}}. On Java side String's serialization logic was improved
> to support all the cases. Refer to IGNITE-3098.
> C++ serialization logic has to be updated as well. Please refer to the
> algorithm located in ignite-3098 branch in the following places:
> - {{BinaryUtils.utf8BytesToStr}} - serialization
> - {{BinaryUtils.strToUtf8Bytes}} - deserialization
> -
> {{IgniteSystemProperties.IGNITE_BINARY_MARSHALLER_USE_STRING_SERIALIZATION_VER_2}}
> controls which version of serialization logic to use (old or new).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
