both global lock in fragment mgr should only protect the map logic, could not use it to protect cancel method.
fragment ctx cancel method should be protected by a lock.
query ctx cancel --> pipelinex fragment cancel ---> query ctx cancel will dead lock.
1. And selected partition number indicator in PaimonScanNode in explain result.
2. Add raw file split and total split indicator in PaimonScanNode in explain result.
3. Opt the error msg of table valued function, print the error msg first to avoid the long msg be truncated.
Move the logic for version+1 from `unprotectedCommitTransaction `to`FinishTransaction`, as the write lock for the table was obtained in `FinishTransaction`
1. In cloud mode, get visible version is a rpc to metaservice, while
loads would get visible version for all partitions.
2. VunionNode should follow batch size.
The log file names are in the format fe.log.${date}-${index}, e.g.
```
fe.log.20240311-1
fe.log.20240311-2
fe.log.20240311-3
...
```
In the previous, fe.log/fe.audit.log will be renamed to fe.log.xxx-1/fe.audit.log.xxx-1
with the minimum index `1` when they reach rotation size. e.g.
```
fe.log.20240311-1 -> fe.log.20240311-2
fe.log.20240311-2 -> fe.log.20240311-3
fe.log.20240311-3 -> fe.log.20240311-4
fe.log -> fe.log.20240311-1
```
there are some drawbacks
1. log4j needs to rename all existing rotated log to spare the
index `1`
2. it's hard to deal with the duplicated names, because every time the
log rotates, it starts from index `1`
After this change, the log will rename to a file name with larger index
instead of `1`. e.g. when rotate
```
fe.log.20240311-1
fe.log.20240311-2
fe.log.20240311-3
fe.log -> fe.log.20240311-4
```
- fix syntax problems of only one table used in leading or mistake usage of brace
example: leading(t1),leading(t1 {t2})
- fix cte used in subquery of using leading
example: with cte as (select c1 from t1) select count(*) from t1 join (select /*+ leading(cte t2) */ c2 from t2 join cte on c2 = cte.c1) as alias on t1.c1 = alias.c2;
which cte used in subquery and subquery also have leading
- fix data mismatched with original plan cause of on predicate push to nullable side
example: select count(*) from t1 left join t2 on c1 > 500 and c2 > 500 can not change to select count(*) from t1 left join t2 on c2 > 500 where c1 > 500
Fix problem that if fe upgrade from a older version, it has error like:
```
MySQL [test]> show full tables;
ERROR 1105 (HY000): NullPointerException, msg: java.lang.NullPointerException: Cannot invoke "org.apache.doris.thrift.TInvertedIndexStorageFormat.toString()" because the return value of "org.apache.doris.catalog.OlapTable.getInvertedIndexStorageFormat()" is null
```
