Re: pgsql: Allow vacuum command to process indexes in parallel.

Hi,

On March 31, 2020 12:43:47 AM PDT, Mahendra Singh Thalor <mahi6run(at)gmail(dot)com> wrote:
>On Mon, 30 Mar 2020 at 09:44, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
>wrote:
>>
>> On Mon, Mar 30, 2020 at 4:18 AM Andres Freund <andres(at)anarazel(dot)de>
>wrote:
>> >
>> > > 2076
>> > > 2077 if ((shared_balance >= VacuumCostLimit) &&
>> > > >>> CID ...: Incorrect expression
>(UNINTENDED_INTEGER_DIVISION)
>> > > >>> Dividing integer expressions "VacuumCostLimit" and
>"nworkers", and then converting the integer quotient to type "double".
>Any remainder, or fractional part of the quotient, is ignored.
>> > > 2078 (VacuumCostBalanceLocal > 0.5 *
>(VacuumCostLimit / nworkers)))
>> > > 2079 {
>> > > 2080 /* Compute sleep time based on the local
>cost balance */
>> > > 2081 msec = VacuumCostDelay *
>VacuumCostBalanceLocal / VacuumCostLimit;
>> > > 2082
>pg_atomic_sub_fetch_u32(VacuumSharedCostBalance,
>VacuumCostBalanceLocal);
>> > > 2083 VacuumCostBalanceLocal = 0;
>> >
>> > Which seems like a fair enough complaint?
>> >
>>
>> Yeah, how can we set up and test a fix for this? Where can I see
>these results?
>
>I am able to make coverity setup. I am verifying fix and will post my
>results in coming days.

That doesn't seem necessary - we should commit a fix. We'll know in a few days for sure. But it's not hard to just theoretically look at the issue in this case?

Andres
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.