[mysql-cdc] Fix table metadata field type 0 when parsing binlog event #2785
Conversation
|
Thanks @EchoLee5 for the contribution, the analysis looks good to me, could you fix the CI failure firstly ? |
.../com/github/shyiko/mysql/binlog/event/deserialization/TableMapEventMetadataDeserializer.java
Outdated
Show resolved
Hide resolved
I've solved it, thank you. |
|
You need add a suppression[1] when override a class to let check style pass |
I have added unit tests. |
Yes, I revised it again, I hope it passes |
| * packed. | ||
| * | ||
| * <p>Line 123: The length of the {@link TableMapEventMetadataDeserializer#readBooleanList} is | ||
| * determined by the length * of the packet header. |
There was a problem hiding this comment.
Please correct this description. There is a useless * in it.
| TableMapEventMetadata tableMapEventMetadata = | ||
| metadataDeserializer.deserialize(new ByteArrayInputStream(data), 10, 8); | ||
|
|
||
| assertThat(tableMapEventMetadata.getSignedness().get(1)); |
There was a problem hiding this comment.
The result is not checked here.
There was a problem hiding this comment.
Could you please add some ITCase to reproduce the problem?
There was a problem hiding this comment.
@ruanhang1993 thank you for your review, I've modified the description and unit tests, the bytes used in my test case can be reproduced before modifying the code. I also mentioned the conditions for reproduction in #2784
|
Duplicate to #2682 |
|
@EchoLee5 Thanks for your contribution. |
As described by #2784, in specific scenarios will appear table metadata field type 0 when parsing binlog event
When reading the sign bit of a numerical type, use the packed to determine the number of bytes to read to avoid the possibility of misreading due to errors in calculations.