replication:index out of range error about go-mysql

[siddontang]

为什么会多出一个字节？正常情况这个多出来的字节应该是个bug了吧，我现在还不清楚。

[iampastor]

那为什么python的要加1呢？

[siddontang]

不知道，所以我就不清楚啥时候会多出来一个字节了。
这个多出来的字节是协议里面带上的，正常就应该被处理，你那边能确定是什么语句可以导致这个问题？

[iampastor]

从binlog信息来看就是普通的Update，并没有特别之处。
不能确定SQL语句是怎样的，因为数据不是我自己提供的，我再看一看能否手动触发这个问题。

[gdey]

If I'm understanding the original inquiry correctly (Or I should say if google translate is translating it correctly.) Then the problem most likely is in this part of the code:
https://github.com/siddontang/go-mysql/blob/master/replication/row_event.go#L266-L270
We don't check here to see that pos is still within the bound of the len(data) but we do access data using the updated pos.

Now why the python code does a +1 on the byte count read; does not make sense to me either.

[siddontang]

Aha, thanks google translate, @gdey .

So how to fix it? use for pos + 1 < len(data) ?
But I still think if we decode data correctly, the updated pos may be the len(data), so only for pos < len(data) is ok.

I think we should construct a case for this.

[gdey]

I agree with you that pos < len(data) is okay. I don't know why the python code is doing the pos+1. What I was pointing out, was that the index out of range could be happening in those lines. This is because we increment pos, but then don't check to see if it's still less than len(data) before using it. If it happens that the pos is greater than len(data) that would cause and index out of range panic.

[siddontang]

@gdey I new an issue in python-mysql-replication for this, maybe @noplay can help us. :-)

[siddontang]

@iampastor
what's your binlog row image, minimal, noblob or full?
https://dev.mysql.com/doc/internals/en/binlog-row-image.html