Re: Should buffer of initialization fork have a BM_PERMANENT flag

On 10.03.2017 04:00, Michael Paquier wrote:
> On Thu, Mar 9, 2017 at 10:25 PM, Artur Zakirov <a(dot)zakirov(at)postgrespro(dot)ru> wrote:
>> I think this is good fixes. I've checked them. And in my opinion they are
>> correct.
>>
>> The code also is good.
>
> Having something with conflicts is not nice, so attached is a rebased version.

Thank you!

I've rerun regression and TAP tests. They all passed.

Also maybe it will be good to fix comments.

In buf_internals.h:
> #define BM_PERMANENT (1U << 31) /* permanent relation (not
> * unlogged) */

And in FlushBuffer():
> /*
> * Force XLOG flush up to buffer's LSN. This implements the basic WAL
> * rule that log updates must hit disk before any of the data-file changes
> * they describe do.
> *
> * However, this rule does not apply to unlogged relations, which will be
> * lost after a crash anyway. Most unlogged relation pages do not bear

Because BM_PERMANENT is used for init forks of unlogged indexes now.

--
Artur Zakirov
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company