Fix zenodraft

[egpbos]

Changes in this PR

Following up on #427, trying to fix the zenodraft upload process with some (potential) tweaks and try-outs:

• No more hardcoded date
• Title reverted to the one of the original Zenodo entry
• Remove .zenodo.json (Zenodo should work with CITATION.cff only)
• Fix the tag weirdly being added by zenodraft (or so it seems, maybe it's actually Zenodo making the tag? see Zenodo entries #427 (comment))
• Check if the Sandbox upload works after the tweaks
• Check if the release action works to the real Zenodo (just modify the action inside this PR to make it trigger; it should be harmless to try, because it only creates a draft deposition, not a final one)
  • Make sure the version is correct (uses tag created in GitHub release or does a sane autogenerated version otherwise)

[netlify]

✅ Deploy Preview for nlesc-guide-testing ready!

Name	Link
🔨 Latest commit	e7261a4
🔍 Latest deploy log	https://app.netlify.com/projects/nlesc-guide-testing/deploys/68878bfec72ae500087ee7d0
😎 Deploy Preview	https://deploy-preview-443--nlesc-guide-testing.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[egpbos]

Triggered PDF action: https://github.com/NLeSC/guide/actions/runs/15853408727

[egpbos]

That worked, but there's still an issue: the version now gets autogenerated to be "v1". Maybe this is only a sandbox issue, because we should generate a proper tag when doing a release. However, I'm not completely sure, we should check (added todo to top post). Also, some mysterious entity also again tagged the latest commit in this branch. In this case, the tag is just a shorthand of the commit hash: 04c7fff. I guess it's sane behavior to have a tag for a release, but preferably not for a draft sandbox one.

[egpbos]

Ah, it's indeed the Zenodraft Action that creates the tags on manual workflow_dispatch triggers, so fine.

[egpbos]

Triggered again, now without .zenodo.json: https://github.com/NLeSC/guide/actions/runs/15853945322

[egpbos]

Oops, I forgot to remove the metadata specification from the workflow. Again! https://github.com/NLeSC/guide/actions/runs/15854009546

[egpbos]

Ok, changed a bit too much in one go and of course it fails on that: I fixed the deposition ID, hoping that that would make zenodraft update the existing record: https://sandbox.zenodo.org/records/277498 However, it fails... with a bizarre error, because it seems to be using the wrong URL? It adds one to the ID...

[egpbos]

Seems to be a Zenodraft issue, so I reported it there: zenodraft/action#30 For now, we can work around it by simply subtracting 1 from our intended IDs 😄

[egpbos]

Triggered... https://github.com/NLeSC/guide/actions/runs/15870721913

[egpbos]

Hm, it ran without errors on GH Actions, but the result on Zenodo was a bit weird. Redoing it with verbose logging activated now... https://github.com/NLeSC/guide/actions/runs/15870872596

[egpbos]

So, this isn't working as I'd expect. Zenodraft apparently removes all metadata from the entry, at least that's what it says in the action log:

... and indeed it seems like Zenodo confirms, since this is what the draft deposition looks like:

[egpbos]

Digging through the zenodraft source, it seems like you need a .zenodo.json file after all. At least, there seems to be no CFF support in zenodraft itself. This is why it's not adding any metadata anymore now.

[egpbos]

Triggered: https://github.com/NLeSC/guide/actions/runs/15871923538

[egpbos]

Again with a fix (actually saving the json file...)... https://github.com/NLeSC/guide/actions/runs/15871998697

[egpbos]

The cffconvert action is picky, should work now... https://github.com/NLeSC/guide/actions/runs/15872132299

[egpbos]

Hm, again a metadata-less draft on Zenodo. This is extra surprising since the action itself errored after creating the .zenodo.json file:

[egpbos]

Grasping at straws, I put the +1 back in the ID... https://github.com/NLeSC/guide/actions/runs/15872239267

[egpbos]

Trying with a -1 again, but now the concept ID - 1: https://github.com/NLeSC/guide/actions/runs/15881558006

[egpbos]

Ok, I did some debugging with and hacking of Zenodraft to print the extra error information that is sent along with a response 400 (basically what I suggested here zenodraft/zenodraft#92)... and figured out that at least one issue is that there is an existing draft version deposition that may be blocking new version additions!

[egpbos]

• Removed the draft deposition.
• Reverted the -1 workaround
• Started a new workflow run... https://github.com/NLeSC/guide/actions/runs/16568068353

[egpbos]

Hurray, it works! New version drafted! The only thing I still have to manually change is the version, which I assume did not get filled out because I ran the workflow manually. We'll have to see whether it does work when we release in the normal way.

[egpbos]

Oh funky, zenodraft doesn't kid around, it also created a bunch of GitHub releases, apparently 😄

I find this a bit messy, so I'll just remove those entries, also remove the draft version deposition and just do a new release to test this workflow in production immediately.

[egpbos]

Just as a try-out, I modified the zenodraft action to also produce output and then generate a nice job summary with a link people should follow to actually publish. Triggered... https://github.com/NLeSC/guide/actions/runs/16569483262

[egpbos]

Eh, that run didn't work for mysterious reasons (there was a draft deposition in the Sandbox already, maybe that was clogging up the pipes?), trying again after removing the existing draft... https://github.com/NLeSC/guide/actions/runs/16569578803

[egpbos]

Another mysterious failure. Maybe the wait time was too short between deleting the draft and creating a new one? Also, I "fixed" the entries in the Sandbox by giving them actual versions, because there were two v2's, which I imagine might confuse the system. Maybe I need to create a new deposition altogether... But let's try one more time with this one: https://github.com/NLeSC/guide/actions/runs/16569708398

[egpbos]

Nope, still same error, something with version already existing, weird.

Let's try with a new concept_id: https://github.com/NLeSC/guide/actions/runs/16570062845

[egpbos]

Of course that doesn't work either, that concept doesn't exist and this workflow tries to make a new version of an existing concept. Reverted. Let's just publish whatever comes out of the broken workflow and see if that helps for subsequent versions, because there's definitely something screwed up in the order of the two "v2's" that are in the sandbox now.

[egpbos]

And, of course, weirdly, the workflow runs fine now... No more existing record stuff. Weird. Still, there is no version_id output! Or am I just referencing it wrong?

Ah, no I think I should have also "compiled" the zenodraft action code before use, did that, switched to the "production" version... Running... https://github.com/NLeSC/guide/actions/runs/16570636932

[egpbos]

Bam, that works!

Nice clickable link 🙂 And the clickable link is, indeed, clickable and correct!

Will have to wait for zenodraft/action#31 before we can use this though, don't want to depend on my crappy version.

[egpbos]

Against my own standards, I'd like to merge now without doing a rebase to clean up the history. Ideally, I'd also have liked to reuse the pdf.yml file so that all git history is preserved in one file, instead of now having the additional upload-pdf.yml file with its own history (although most of it is copy-pasted from pdf.yml), but this would require too much rebasing on old stuff, so let's just keep this all as is.