Re: BUG #2379: Duplicate pkeys in table

Philip Warner wrote:

>  Item   7 -- Length:  168  Offset: 3920 (0x0f50)  Flags: USED
>   XMIN: 32902771  CMIN: 20  XMAX: 0  CMAX|XVAC: 32902872
>   Block Id: 0  linp Index: 7   Attributes: 34   Size: 36
>   infomask: 0x2913
> (HASNULL|HASVARWIDTH|HASOID|XMIN_COMMITTED|XMAX_INVALID|UPDATED)
>   t_bits: [0]: 0x9f [1]: 0x80 [2]: 0x7e [3]: 0x84
>           [4]: 0x00

Hmm, shouldn't we see the MOVED_OFF bit set also if the cmax/xvac field
is actually xvac?

>  Item  27 -- Length:  168  Offset: 8024 (0x1f58)  Flags: USED
>   XMIN: 32902771  CMIN: 20  XMAX: 33048159  CMAX|XVAC: 20
>   Block Id: 318  linp Index: 6   Attributes: 34   Size: 36
>   infomask: 0x2913
> (HASNULL|HASVARWIDTH|HASOID|XMIN_COMMITTED|XMAX_INVALID|UPDATED)
>   t_bits: [0]: 0x9f [1]: 0x80 [2]: 0x7e [3]: 0x84
>           [4]: 0x00

I'm confused -- the original report showed this tuple with ctid (72,27),
but this seems to be in a different block?

What's the explanation for this tuple to have cmin=cmax?  Is this
normal?

Sorry I have only questions :-(

--
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.