Github Actions (CI) - Mailing list pgsql-hackers

From Thomas Munro
Subject Github Actions (CI)
Date
Msg-id CA+hUKG+y_SHVQcU3CPokmJxuHp1niebCjq4XzZizf8SR9ZdQRQ@mail.gmail.com
Whole thread Raw
Responses Re: Github Actions (CI)
Re: Github Actions (CI)
List pgsql-hackers
Hi hackers,

I'm looking for more horsepower for testing commitfest entries
automatically, and today I tried out $SUBJECT.  The attached is a
rudimentary first attempt, for show-and-tell.  If you have a Github
account, you just have to push it to a branch there and look at the
Actions tab on the web page for the results.  Does anyone else have
.github files and want to share, to see if we can combine efforts
here?

The reason for creating three separate "workflows" for Linux, Windows
and macOS rather than three separate "jobs" inside one workflow is so
that cfbot.cputube.org could potentially get separate pass/fail
results for each OS out of the API rather than one combined result.  I
rather like that feature of cfbot's results.  (I could be wrong about
needing to do that, this is the first time I've ever looked at this
stuff.)

The Windows test actually fails right now, exactly as reported by
Ranier[1].  It is a release build on a recent MSVC, so I guess that is
expected and off-topic for this thread.  But generally,
.github/workflows/ci-windows.yml is the weakest part of this.  It'd be
great to get a debug/assertion build, show backtraces when it crashes,
run more of the tests, etc etc, but I don't know nearly enough about
Windows to do that myself.  Another thing is that it uses Choco for
flex and bison; it'd be better to find those on the image, if
possible.  Also, for all 3 OSes, it's not currently attempting to
cache build results or anything like that.

I'm a bit sad that GH doesn't have FreeBSD build runners.  Those are
now popping up on other CIs, but I'm not sure if their free/open
source tiers have enough resources for cfbot.

[1] https://www.postgresql.org/message-id/flat/CAEudQArhn8bH836OB%2B3SboiaeEcgOtrJS58Bki4%3D5yeVqToxgw%40mail.gmail.com

Attachment

pgsql-hackers by date:

Previous
From: Ashutosh Bapat
Date:
Subject: Re: Printing LSN made easy
Next
From: Masahiko Sawada
Date:
Subject: Re: [HACKERS] logical decoding of two-phase transactions