slice bounds out of range

[iakisey]

Please answer these questions before submitting your issue. Thanks!

1. What did you do?
   programer

2. What did you expect to see?
   

3. What did you see instead?
   
   

4. What version of TiDB are you using (tidb-server -V or run select tidb_version(); on TiDB)?
   Release Version: v2.1.0-rc.1-7-g38c939f

[zhexuany]

Thanks for your reporting. We will investigate this. @zz-jason PTAL when you have time.

[mccxj]

mini testcase:

create table t(c enum('a'));
insert into t values('a');
select c,count(1) from t group by c;

dumpTextRow use column info to decode chunk.Row, excepted type is Enum.
But firstrow agg function write it use type String, min(),max() has the same problem.
see AggFuncDesc#typeInfer4MaxMin

release-2.0 works fine. Maybe #6852 affected it.

[iakisey]

测过了 是这个版本的问题吧，之前安装过 v2.0.4-1-g06a0bf5 版本 就没问题

[tiancaiamao]

Thanks for you kind reply. @mccxj
The bug was actually introduced in #7268, and I'm now reviewing that PR to find what's going on.

[tiancaiamao]

It's around here:
https://github.com/pingcap/tidb/pull/7268/files#diff-2ed8f699513f49aaaf408561a56aa2e7L477

The old code evaluate using type.Datum, and enum type info would be handle properly.
After the change, we'll use chunk.Chunk instead of types.Datum, the eval type is always using string.

When the chunk execution finish, the column data stored in the chunk is string (it should be converted to enum somewhere), but the schema expect the column data to be a enum, so the panic happen.

PTAL @zz-jason

[zz-jason]

In the type infer phase, the result type for function first_row(enum) is inferred to string, and the column type in the schema should also be changed to string.

In a word, the type of the column in the schema should be reset after creating an aggregate function descriptor through the function aggregation.NewAggFuncDesc()

[mccxj]

@zz-jason duplicate with #7307

[zz-jason]

@mccxj Got it! Thanks!