Re: [PATCH] Fix division by zero (explain.c) - Mailing list pgsql-hackers

On Sat, May 09, 2020 at 06:48:59AM -0300, Ranier Vilela wrote:
>Em sáb., 9 de mai. de 2020 às 01:45, Tom Lane <tgl@sss.pgh.pa.us> escreveu:
>
>> James Coleman <jtc331@gmail.com> writes:
>> > There are always full sort groups before any prefix groups can happen,
>> > so we know (even though the tooling doesn't) that the 2nd test can
>> > never contradict the first.
>>
>> So maybe an assertion enforcing that would be appropriate?
>> Untested, but:
>>
>> -                       if (fullsortGroupInfo->groupCount == 0 &&
>> -                               prefixsortGroupInfo->groupCount == 0)
>> +                       if (fullsortGroupInfo->groupCount == 0)
>> +                       {
>> +                               Assert(prefixsortGroupInfo->groupCount ==
>> 0);
>>                                 continue;
>> +                       }
>>
>I agree, asserts always help.
>

That doesn't work, because the prefixSortGroupInfo is used before
assignment, producing compile-time warnings.

I've pushed a simpler fix without the assert. If we want to make this
check, perhaps doing it in incremental sort itself would be better than
doing it in explain.

regards

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