Re: BUG #15943: Valgrind-detected error in SlruPhysicalWritePage

Hi,

On August 8, 2019 2:41:42 PM EDT, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>Andres Freund <andres(at)anarazel(dot)de> writes:
>> Seems there's no reason to not zero initialize memory here? But
>perhaps just by initializing the stack variable?
>
>I think all we need here is another suppression in
>src/tools/valgrind.supp. We have not insisted on zeroing
>pad bytes that get sent to disk in the places already
>enumerated in valgrind.supp, so why would we apply a
>different rule to async.c?

Well, with a lot of the other case there's a lot of sources for such uninitialized data. Here there probably is just one? If it weren't such a game of whack a mole, I'd even consider properly initializing the other places. And initializing the stack data here ought to be cheap in this case?

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