Re: Assert in heapgettup_pagemode() fails due to underlying buffer change

Hello Noah,

06.06.2024 22:07, Noah Misch wrote:
>
>> I don't know, but if the locks are really missing now, I feel like the
>> first question is "which commit got rid of them?". It's a little hard
>> to believe that they've never been there and somehow nobody has
>> noticed.
>>
>> Then again, maybe we have; see Noah's thread about in-place updates
>> breaking stuff and some of the surprising discoveries there. But it
>> seems worth investigating.
> $SUBJECT looks more like a duplicate of
> postgr.es/m/flat/20240512171658.7e.nmisch@google.com (Hot standby queries see
> transient all-zeros pages).

Thank you for the reference! Yes, it looks very similar. Though I can't
say the sleep you proposed helps the failure reproduction (I've tried
026_overwrite_contrecord.pl and saw no more frequent failures or so).

My bisect run ended with:
210622c60e1a9db2e2730140b8106ab57d259d15 is the first bad commit

Author: Thomas Munro <tmunro@postgresql.org>
Date:   Wed Apr 3 00:03:08 2024 +1300

     Provide vectored variant of ReadBuffer().

Other buildfarm failures with this Assert I could find kind of confirm this:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=calliphoridae&dt=2024-04-03%2003%3A32%3A18
(presumably a first failure of this sort)
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=tamandua&dt=2024-04-04%2015%3A38%3A16
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=margay&dt=2024-05-07%2004%3A00%3A08

Best regards,
Alexander