On 6/17/21 11:13 AM, Andrew Dunstan wrote:
> On 6/17/21 11:05 AM, Tom Lane wrote:
>> Andrew Dunstan <andrew@dunslane.net> writes:
>>> Here's a patch I propose to apply to fix this bug (See
>>>
<https://www.postgresql.org/message-id/flat/759e997e-e1ca-91cd-84db-f4ae963fada1%40dunslane.net#b1cf11c3eb1f450bed97c79ad473909f>)
>> If I'm reading the code correctly, your change in RelationBuildTupleDesc
>> is scribbling directly on the disk buffer, which is surely not okay.
>> I don't understand why you need that at all given the other defenses
>> you added ... but if you need it, you have to modify the tuple AFTER
>> copying it.
>
> OK, will fix. I think we do need it (See Andres' comment in the bug
> thread). It should be a fairly simple fix.
>
>
revised patch attached.
cheers
andrew
--
Andrew Dunstan
EDB: https://www.enterprisedb.com