#ci

harald@hub.volse.no

When I finally figure out what was wrong, I always feel really stupid. It appears, as often is the case, the solution was really simple once I knew what the problem actually was.

Fighting a CI pipeline that didn't provide any useful feedback, and that I wasn't able to replicate locally made the whole experience rather frustrating.

But in the end — after probing it repeatedly with slightly different configurations, and sneaky ways to try to get any useful logs out of it — the problem was of course on my end.

Don't include the id column when preloading test data into the database.

Can't remember why I included the id column in the first place, but it caused me much pondering. Even stranger was that it worked perfectly fine in my local dev/test setup.

At least the CI setup has become a lot more robust, and I've added a lot more logging to the test runs, so it should hopefully be easier to track down next time something mysteriously doesn't work.

#mywork #developerlife #ci #testing #hubzilla-dev

harald@hub.volse.no

Wohoo! I have the Hubzilla test suite with a mysql database running in the framagit gitlab ci environment!

Screenshot_20240104_151009.png

Was a pain to debug, as the local command line ci runner is useless and unsupported.

Now, try to get the tests running with postgres too, then!

And get the coverage report working... One step at the time...

#mywork #hubzilla #hubzilla-dev #ci