[
https://issues.apache.org/jira/browse/IGNITE-19794?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17740327#comment-17740327
]
Igor Sapego commented on IGNITE-19794:
--------------------------------------
Looks good to me.
> Thin 3.0: ClientMessagePacker packs binary incorrect tuples
> -----------------------------------------------------------
>
> Key: IGNITE-19794
> URL: https://issues.apache.org/jira/browse/IGNITE-19794
> Project: Ignite
> Issue Type: Bug
> Components: thin client
> Reporter: Konstantin Orlov
> Assignee: Pavel Tupitsyn
> Priority: Minor
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
> Attachments: corrupted_tuple_test.patch
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Take a look at the method:
> {code:java}
> // org.apache.ignite.internal.client.proto.ClientMessagePacker
> public void packBinaryTuple(BinaryTupleReader binaryTupleReader, int
> elementCount) {
> ByteBuffer buf = binaryTupleReader.byteBuffer();
> int len = buf.limit() - buf.position();
> if (elementCount > -1) {
> binaryTupleReader.seek(elementCount - 1);
> len = binaryTupleReader.end();
> buf.limit(len + buf.position());
> }
> packBinaryHeader(len);
> writePayload(buf);
> }
> {code}
> If {{elementCount}} is provided (> -1), tuple will be truncated up to the
> element with index {{elementCount - 1}}. A truncated tuple, although
> retaining logical equality, is not binary equal to the same tuple, but
> correctly assembled, since neither offset table nor null map are adjusted.
> Because {{ClientMessagePacker}} is a general purpose library to implement
> protocol between client and server, such behaviour is rather dangerous and
> should be fixed.
> In the attachment there is a test that highlights the problem.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
