On 20 July 2017 at 11:33, Craig Ringer <cr...@2ndquadrant.com> wrote:
> Hi
>
> Whenever I'm debugging some kind of corruption incident, possible
> visibility bug, etc, I always land up staring at integer infomasks or using
> a SQL helper function to decode them.
>
> That's silly, so here's a patch to teach pageinspect how to decode
> infomasks to a human readable array of flag names.
>
> Example:
>
> SELECT t_infomask, t_infomask2, flags
> FROM heap_page_items(get_raw_page('test1', 0)),
> LATERAL heap_infomask_flags(t_infomask, t_infomask2, true) m(flags);
> t_infomask | t_infomask2 | flags
>
> ------------+-------------+---------------------------------
> -------------------------------------------
> 2816 | 2 | {HEAP_XMIN_COMMITTED,HEAP_
> XMIN_INVALID,HEAP_XMAX_INVALID,HEAP_XMIN_FROZEN}
> (1 row)
>
>
> To decode individual mask integers you can just call it directly. It's
> strict, so pass 0 for the other mask if you don't have both, e.g.
>
> SELECT heap_infomask_flags(2816, 0);
>
> The patch backports easily to older pageinspect versions for when you're
> debugging something old.
>
> BTW, I used text[] not enums. That costs a fair bit of memory, but it
> doesn't seem worth worrying too much about in this context.
>
> For convenience it also tests and reports HEAP_LOCKED_UPGRADED and
> HEAP_XMAX_IS_LOCKED_ONLY as pseudo-flags.
>
> I decided not to filter out
> HEAP_XMIN_COMMITTED,HEAP_XMIN_INVALID,HEAP_XMAX_INVALID
> when HEAP_XMIN_FROZEN is set
>
Er, decided not to filter out HEAP_XMIN_COMMITTED,HEAP_XMIN_INVALID.
Obviously wouldn't filter out HEAP_XMAX_INVALID, that was a copy-paste'o.
I wonder if it's worth dropping the HEAP_ prefix. Meh, anyway, usable as-is.
--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
