Re: Heads Up: cirrus-ci is shutting down June 1st

Hi,

I found a few more issues with the patches.

- Most of them were around windows, particularly around the package
installs. I saw a bunch of failures with chocolatey installs failing. Since
all we used it for is bison, flex and diff, it seems faster and more
reliable to just install those via msys/pacman. Takes like 1/10th of the
time, to boot.

We don't need to install diff and ninja, those are already installed.

- We were installing openssl 1.1 ourselves, but there's already an existing
openssl 3.6 on the github runners. So it seems to make sense to just use
that, instead of spending close to a minute installing it?

- There was also a bunch of error handling missing, because windows shell
don't have set -e like behavior.

- While tackling these I realized I had dropped meson-logs/ archiving on
failure.

- Also added a commit to add back a runningcheck test. I see that Peter also
just posted that, I'll compare with that after this.

One difference I did see is that I added it to linux meson 32bit, because
that's the quickest task rn...

- some other boring stuff, see patches for details

Example run: https://github.com/anarazel/postgres/actions/runs/26869491578

I think after merging Peter's and my approaches to runningcheck, we can commit
this. There will be a lot more to do, but this is much better than nothing.

If somebody is looking for a small project to help out with: Right now we run
cpan on every run, without caching. There occasionally are visible network
issues and even without that it takes 30s. So that'd be a good target for
caching. We did that in the past, for cirrus/macos, so that shouldn't be too
hard (c.f. 93d97349461).

Greetings,

Andres Freund