Uses https://github.com/marketplace/actions/flatpak-builder

According to the link, it seems the artifact should be uploaded automatically without needing a specific upload step?

Keen for feedback, if we need a cleanup step etc. This is just based on the example given.

Ok well glu and rapidjson are building fine but now it's failing with:

 -- Installing: /app/share/doc/RapidJSON/html/classrapidjson_1_1_encoded_input_stream.html
Committing stage build-rapidjson to cache
========================================================================
Building module avogadro2 in /__w/avogadroapp/avogadroapp/.flatpak-builder/build/avogadro2-1
========================================================================
Error: module: avogadro2: Can't find CMakeLists.txt
Error: Build failed: Error: The process '/usr/bin/xvfb-run' failed with exit code 1

and I really don't know why. As far as I can tell openchemistry was successfully cloned into /__w/avogadroapp/avogadroapp i.e. the source root, and all the other repos were successfully cloned to their respective subdirectories.

You might want to use this to debug the failure. I can't get the raw log.

   - name: Setup tmate session
      if: ${{ failure() }}
      uses: mxschmitt/action-tmate@v3

Does it fail or does it just time out?

In the annotations it says

System.IO.IOException: No space left on device : '/home/runner/runners/2.320.0/_diag/Worker_20241111-000701-utc.log'

If it ran out of available storage, that would explain why it just stops without giving an error

I've turned off the .flatpak-builder cache to see if that helps, added tmate at the same time.

I noticed that build_cmake.yml includes a cleanup step, as does build_m1.yml, but build_qt6.yml doesn't. How come the discrepancy? Is the cleanup step preferred, and if so should I update all the workflows to match?

Ok well now there seem to be problems downloading the third-party dependencies (I get the same error when trying to build the flatpak locally) so I guess I'll just leave this for a few days and come back to it...

Ah, I didn't see that the versions of spglib and libarchive looked for were bumped in the main repo... Looks like it'll work now.

Adding GitHub Actions to the openchemistry repo as planned will help avoid a mismatch.

In particular, the spglib bump fixed cmake build problems on Windows

In its current state this now builds fine locally, so I don't think the manifest is the issue. I don't know what the problem is really. It just gets randomly stuck during the open babel build step.

That's where I'd definitely use the remote shell. Does Flatpak build in a Docker image or something? If so, maybe it's running out of space?

Ah yes, that'll be it. I just used the workflow suggested by the Action's author, and that indeed runs it in a container. I'll increase the resources allocated to it.

I did think it seemed like a low storage issue but I checked my local build and everything only comes to a few GB, which is way less than the runners have, so I ruled it out on that basis. Hadn't thought about containers being potentially limited even further.

Ok it's now set up to build on the runner directly but it now can't do the clone like it's meant to be able to :(

Don't know if running flatpak-builder as sudo would fix that but seems like a bad idea right?

Don't know if running flatpak-builder as sudo would fix that but seems like a bad idea right?

The runners are images, so you can sudo if needed. If you look at the other build actions, they all install a variety of packages.

In the last workflow run the actual build was successful, even if it did take an hour, so with 2c56246 I'm expecting it to pass and this should now be ready to merge @ghutchis :)

Okay, I'l take a look tomorrow morning Pittsburgh time.