Re: Use of uninitialized variables in ExecFindPartition() for parent partition without leaves (HEAD only)

On 2017/12/01 11:48, Michael Paquier wrote:
> On Thu, Nov 30, 2017 at 10:17 AM, Amit Langote
> <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote:
>> Oops, I messed up taking the diff and mistakenly added noise to the patch.
>
> Which is that bit:
> - * BuildSlotPartitionKeyDescription
> + * ExecBuildSlotPartitionKeyDescription

Yep.

>> Fixed in the attached.
>
> For information, it is easy enough to run into rather nasty behaviors
> here. Taking for example the test case in the last patch, but without
> the actual fix:
> -- no partitions, so fail
> insert into range_parted values ('a', 11);
> ! ERROR: invalid memory alloc request size 2139062147

Yikes.

> So the test case you are adding is a good thing to have, and I am fine
> with the comment.

Thanks.

> I have added a CF entry for this thread by the way
> (https://commitfest.postgresql.org/16/1392/), and marked the thing as
> ready for committer as we agree about the fix. Let's track properly
> this issue until it gets committed.

Yeah, thanks.

Regards,
Amit