Add more smoke-tests packages up to cardano-node

[michaelpj]

This extends the smoke-test package builder to build the newest versions of:

• plutus-ledger-api
• cardano-ledger-api
• ouroboros-network
• ouroboros-consensus-cardano
• cardano-api
• cardano-node

To make all this work I had to do a lot of revisions. I took the hard but hopefully thorough route of just keeping trying to build things and then revising whatever the problem with the particular package that failed to build was. That meant a lot of"add an upper bound to foo-X -> cabal picks foo-Y -> add an upper bound to foo-Y -> ..., but I got there eventually.

The revisions are obviously very hard to review at the moment, so I guess you'll just have to trust me. If anyone wants to spot-check, I tried to list in the commit message what the revision was.

There is only one bit of global config I added: allow-newer: ekg:json. cardano-node needs it, upstream seems dead, we should probably just patch it in CHaP so we can bin that.

Note that this means that in future the CI will block new releases of these packages unless they build cleanly!

[michaelpj]

(The build was fast because I've been experimenting by building on nixbuild.net, so it was already cached)

[lehins]

LGTM

[andreabedini]

There is only one bit of global config I added: allow-newer: ekg:json. cardano-node needs it, upstream seems dead, we should probably just patch it in CHaP so we can bin that.

Unfortunately the situation is complex, see my flawed attempt of making a revision for ekg. tl;dr: This is a consequence of us forking ekg-json. We fixed ekg-json but ekg still need relaxing its bounds. Maybe it makes sense to relax the bounds anyway? ekg does work with aeson>=2 as long as ekg-json works with aeson>=2. I might reopen that issue and check what the trustees think about this. Edit: I have opened haskell-infra/hackage-trustees#359

[andreabedini]

@michaelpj I am not sure whether you have already done this or not but I am going to suggest it anyway for the sake of institutional knowledge.

When we make a revision for pkg-a, we need to make sure that our repos where pkg-a is imported keep working. I.e. we could have accidentally put bounds that are too restrictive. This is what happened in #154 and the fix is still WIP (see IntersectMBO/plutus-apps#1023, the typo in freer-extras.cabal is still in CHaP 🤦).

[michaelpj]

When we make a revision for pkg-a, we need to make sure that our repos where pkg-a is imported keep working

Right. I guess what I'm hoping is that ultimately this CI check can substitute for that. It does make me realise that I should probably throw some of the plutus-apps packages into the smoke-test too. But yes I guess I'll try and build master of the various projects too...

[michaelpj]

@andreabedini one data point: I was able to build cardano-node master with this, including after deleting most of the constraints stanza. I'll try and check plutus-apps too.

[michaelpj]

Uhhhh. So someone merged something and now I need to update the timestamp on literally everything? 😭

[michaelpj]

I'm going to merge this as I think it's an improvement and I will iterate from here if there are problems.