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

Hi,

On 2022-03-05 13:21:26 +1300, Thomas Munro wrote:
> On Sat, Mar 5, 2022 at 10:56 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
> > I suspect one also needs the console detach thing.
>
> I tried adding DETACHED_PROCESS (which should be like calling
> FreeConsole() in child process) and then I tried CREATE_NEW_CONSOLE
> instead, on top of CREATE_NEW_PROCESS_GROUP. Neither helped, though I
> lost the postmaster's output.

I think the issue with the process group is real, but independent of the
failing test.Locally just specifying CREATE_NEW_PROCESS_GROUP fixes the
problem of a pg_ctl start'ed database being stopped after ctrl-c.

I think CI failing is due to cirrus bug, assuming a bit too much similarity
between unix and windows....

https://github.com/cirruslabs/cirrus-ci-agent/issues/218#issuecomment-1059657781

As indicated in the message above it, there's a workaround. Not sure if worth
committing, if they were to fix it in a few days? cfbot could be repaired by
just adding a repo environment variable of CIRRUS_AGENT_VERSION 1.73.2...
OTOH, committing and reverting a single line + comment is cheap.

Greetings,

Andres Freund