Re: Inadequate executor locking of indexes - Mailing list pgsql-hackers

(this is not a reply to your full proposal, just something I thought to
point out)

On 2019/03/13 10:38, David Rowley wrote:
> i.e don't open the indexes for DELETEs.  I had ideas that maybe this
> could be changed to check the idxlockmode and open the indexes if it's
> above AccessSharedLock.  There didn't seem to be a very nice way to
> fetch the RangeTblEntry from the ResultRelInfo though,

Did you miss ri_RangeTableIndex?  It's the range table index of the result
relation for which a given ResultRelInfo is created.

Thanks,
Amit