Re: CFBot has failures on 027_stream_regress for a number of patches

Hi,

On 2022-03-28 15:28:20 -0400, Greg Stark wrote:
> I notice a number of patches have a failure in the cfbot on the
> 027_stream_regress test. I think these are related to a bug in that
> test being discussed in a thread somewhere though I don't have it
> handy. Is that right?

It looks more like a bug in the general regression tests that are more likely
to be triggered by 027_stream_regress.

> I think it doesn't indicate anything wrong with the individual patches right?

Depends on the patch, I think. It certainly also has triggered for bugs in
patches...

> For an example the "Add checkpoint and redo LSN to LogCheckpointEnd
> log message" patch which is a fairly simple patch adding a few details
> to some log messages is failing with this which doesn't make much
> sense since it shouldn't be affecting the actual recovery at all:

027_stream_regress.pl runs the normal regression tests via streaming rep - the
failure here is on the primary (i.e. doesn't involve recovery itself). Due to
the smaller shared_buffers setting the test uses, some edge cases are
triggered more often.

> diff -U3 /tmp/cirrus-ci-build/src/test/regress/expected/cluster.out
> /tmp/cirrus-ci-build/src/test/recovery/tmp_check/results/cluster.out ---
> /tmp/cirrus-ci-build/src/test/regress/expected/cluster.out 2022-03-28
> 01:18:36.126774178 +0000 +++
> /tmp/cirrus-ci-build/src/test/recovery/tmp_check/results/cluster.out
> 2022-03-28 01:23:24.489517050 +0000 @@ -467,7 +467,8 @@ where row(hundred,
> thousand, tenthous) <= row(lhundred, lthousand, ltenthous); hundred |
> lhundred | thousand | lthousand | tenthous | ltenthous
> ---------+----------+----------+-----------+----------+----------- -(0 rows)
> + 0 | 99 | 0 | 999 | 0 | 9999 +(1 row)

This one is: https://postgr.es/m/CA%2BhUKGLV3wzmYFbNs%2BTZ1%2Bw0e%3Dhc61fcvrF3OmutuaPBuZMd0w%40mail.gmail.com

Greetings,

Andres Freund