Re: Regression tests failures on Windows Server 2019 - on master at commit # d816f366b

Hi,

On 2022-03-04 09:10:14 -0800, Andres Freund wrote:
> > I find it a little suspicious that startcreate is merrily starting
> > the postmaster on port 5432, without concern for the possibility
> > that that port is in use or blocked by a packet filter.
>
> It's a container that's just created for that run, so there shouldn't be
> anything else running. We also see that the server successfully starts:
>
> 2022-03-03 23:24:15.261 GMT [5504][postmaster] LOG: starting PostgreSQL 15devel, compiled by Visual C++ build 1929, 64-bit
> 2022-03-03 23:24:15.270 GMT [5504][postmaster] LOG: listening on IPv6 address "::1", port 5432
> 2022-03-03 23:24:15.270 GMT [5504][postmaster] LOG: listening on IPv4 address "127.0.0.1", port 5432
> 2022-03-03 23:24:15.321 GMT [5808][startup] LOG: database system was shut down at 2022-03-03 23:24:15 GMT
> 2022-03-03 23:24:15.341 GMT [5504][postmaster] LOG: database system is ready to accept connections

Oh, huh. I added a printout of the running tasks, and it sure looks like
postgres is not running anymore when plcheck is run, without anything ending
up in the server log...

It looks like the CI agent had some changes about stdin / stdout of the shells
used to start scripts. The version between the last successful run and the
first failing run changed.
https://github.com/cirruslabs/cirrus-ci-agent/commits/master

I wonder if we're missing some steps, at least on windows, to make pg_ctl
start independent of the starting shell?

Greetings,

Andres Freund