Re: Possible dereference after null check (src/backend/executor/ExecUtils.c) - Mailing list pgsql-hackers

From Kyotaro Horiguchi
Subject Re: Possible dereference after null check (src/backend/executor/ExecUtils.c)
Date
Msg-id 20210212.152801.1280542156314557027.horikyota.ntt@gmail.com
Whole thread Raw
In response to Possible dereference after null check (src/backend/executor/ExecUtils.c)  (Ranier Vilela <ranier.vf@gmail.com>)
Responses Re: Possible dereference after null check (src/backend/executor/ExecUtils.c)
List pgsql-hackers
At Wed, 10 Feb 2021 19:54:46 -0300, Ranier Vilela <ranier.vf@gmail.com> wrote in 
> Hi,
> 
> Per Coverity.
> 
> The functions ExecGetInsertedCols and ExecGetUpdatedCols at ExecUtils.c
> only are safe to call if the variable "ri_RangeTableIndex" is  != 0.
> 
> Otherwise a possible Dereference after null check (FORWARD_NULL) can be
> raised.

As it turns out, it's a false positive. And perhaps we don't want to
take action just to satisfy the static code analyzer.


The coment in ExecGetInsertedCols says:

> /*
>  * The columns are stored in the range table entry. If this ResultRelInfo
>  * doesn't have an entry in the range table (i.e. if it represents a
>  * partition routing target), fetch the parent's RTE and map the columns
>  * to the order they are in the partition.
>  */
> if (relinfo->ri_RangeTableIndex != 0)
> {

This means that any one of the two is always usable here.  AFAICS,
actually, ri_RangeTableIndex is non-zero for partitioned (=leaf) and
non-partitoned relations and ri_RootResultRelInfo is non-null for
partitioned (parent or intermediate) relations (since they don't have
a coressponding range table entry).

The only cases where both are 0 and NULL are trigger-use, which is
unrelated to the code path.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center



pgsql-hackers by date:

Previous
From: Kyotaro Horiguchi
Date:
Subject: Re: Is Recovery actually paused?
Next
From: "tsunakawa.takay@fujitsu.com"
Date:
Subject: RE: Parallel INSERT (INTO ... SELECT ...)