Re: zheap: a new storage format for PostgreSQL

From: Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>
To: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
Cc: PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Daniel Westermann <daniel(dot)westermann(at)dbi-services(dot)com>
Subject: Re: zheap: a new storage format for PostgreSQL
Date: 2018-11-05 14:51:06
Message-ID: e9dedb45-353a-fa55-5c9b-93b1468a8547@2ndquadrant.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 11/5/18 4:00 AM, Amit Kapila wrote:
> On Sat, Nov 3, 2018 at 9:30 AM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>> On Fri, Nov 2, 2018 at 6:41 PM Tomas Vondra
>> <tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
>>> I'm sure
>>> it's not the only place where we do something like this, and the other
>>> places don't trigger the valgrind warning, so how do those places do
>>> this? heapam seems to call fetch_att in the end, which essentially calls
>>> Int32GetDatum/Int16GetDatum/CharGetDatum, so why not to use the same
>>> trick here?
>>>
>>
>> This is because, in zheap, we have omitted all alignment padding for
>> pass-by-value types. See the description in my previous email [1]. I
>> think here we need to initialize ret_datum at the beginning of the
>> function unless you have some better idea.
>>
>
> I have pushed a fix on the above lines in zheap-branch, but I am open
> to change it if you have better ideas for the same.
>

Thanks. Initializing the variable seems like the right fix here.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Dmitry Dolgov 2018-11-05 15:19:07 Re: Flexible permissions for REFRESH MATERIALIZED VIEW
Previous Message Jean-Christophe Arnu 2018-11-05 14:37:06 Re: wal_dump output on CREATE DATABASE