Re: Dead code in _bt_split?

Tom Lane wrote:
> "Heikki Linnakangas" <heikki@enterprisedb.com> writes:
>> This is right after a for-loop, which exits when i = maxoff + 1. So the 
>> first if-statement could be written as "if (newitemoff > maxoff)". If 
>> that's true, newitemonleft shouldn't be true, because that would mean 
>> that we've split a page so that all items went to the left page, and the 
>> right page is empty.
> 
> No, it would mean that we split the page in such a way that only the new
> item is going to the right page.  Probably not hard to duplicate if you
> use near-maximal-sized keys.

In that case, newitemleft would be false, right?

I'm saying the piece marked with X> below is unreachable:

>     /* cope with possibility that newitem goes at the end */
>     if (i <= newitemoff)
>     {
>         if (newitemonleft)
>         {
X>             _bt_pgaddtup(rel, leftpage, newitemsz, newitem, leftoff,
X>                          "left sibling");
X>             itup_off = leftoff;
X>             itup_blkno = BufferGetBlockNumber(buf);
X>             leftoff = OffsetNumberNext(leftoff);
>         }
>         else
>         {
>             _bt_pgaddtup(rel, rightpage, newitemsz, newitem, rightoff,
>                          "right sibling");
>             itup_off = rightoff;
>             itup_blkno = BufferGetBlockNumber(rbuf);
>             rightoff = OffsetNumberNext(rightoff);
>         }
>     }

--   Heikki Linnakangas  EnterpriseDB   http://www.enterprisedb.com