Upgrade GitHub Actions to latest versions #51
Conversation
Co-authored-by: cprecioso <[email protected]>
Copilot
AI
changed the title
|
@copilot remove the |
Co-authored-by: cprecioso <[email protected]>
Done! I've removed the |
|
@copilot remove manual ghc installation and uncomment the action to install haskell (last version is |
Co-authored-by: cprecioso <[email protected]>
Done! I've removed the manual GHC installation steps and uncommented the haskell-actions/setup action, updating it to use the latest v2 version. The workflow now uses the official Haskell setup action instead of manual ghcup installation, which should be more reliable and maintainable. (91c6897) |
|
what the heck |
we're in 2025, AI can update your dependencies now 😂 |
Maybe it can, but that doesn't mean it should :D. Ok so this is a real PR. Can you give me some background on it? How much was done by AI, how much by you? How confident are you in this PR? Do you understand every change that was done in it, and have you thought yourself through them? Have you thought through possible other solutions or what else could be done here? |
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
|
@Martinsos I wouldn't put up for review a PR that didn't pass the same muster as my own. This is just me testing tools out, but the PR is real and needed, as the Actions won't run anymore because the versions are so outdated (see the output here). As you can see if you scroll up, I've been giving the AI specific instructions for the kind of changes I want to see get done, plus you can also see that I've added commits made by myself in this PR, and the PR description has been edited by me to give a clear overview of the changes made. Also, I've tagged Filip as a reviewer on this PR so I made sure that one more pair of eyes are on this. Yep, I've reviewed, and understand all the changes done, as I've asked for them specifically and in the end even changed some things to make it as I would have done. The context on this is that I'm trying out this automatic Copilot PRs, and thought that it could make a decent starting point for upgrading the Haskell toolchain at Wasp (wasp-lang/wasp#2950), taking out the manual work of bumping version numbers and testing dependencies. When I showed that to Filip, he asked me if it could also do the same for this library here, and I tried it too (#50), but the actions are outdated so the CI was not running. For those two other PRs, I wouldn't think to send them as I am not expert enough in Haskell to understand all the changes. I provided them to a colleague with a clear indication that this is AI-generated, and with the intention of providing a starting point. But this specific PR here does touch things I am familiar with and I specifically asked for the kinds of changes I wanted to see. |
|
My bad, the only thing I noticed from your involvement was two messages you exchanged with Copilot, nothing else. I didn't see that you edited the initial comment, nor that you added additional commits. Those are easy to miss, I am afraid, because the initial comment says it was authored by Copilot and looks AI generated (and I think it would have helped if you gave some context on it in the initial comment or the additional comment after it, explaining how you did this PR, basically what you explained after I asked for it. AI is still a new thing and is changing fast, and I can't know how you or anybody else specifically treats it or how confident are you in its abilities, and how much are you willing to let it take over vs double-check or redo on your own. This also changes from model to model and AI to AI, the level of confidence you might have in it. When using AI, there is no question about us delegating a part of our thinking to it, because we will not have to think through some things we normally would (or we would not use it). The question is only how much is being delegated, to which degree, and that is what I as a reviewer need to understand for specific PR, to know how much I have to (re)think through myself. Therefore, it is great if you can provide additional assurance in situations like this, especially when the change looks predominantly AI-driven, to help the me as a review understand what I am dealing with. |
There was a problem hiding this comment.
Looks good in general!
One thing I am not sure about is how we install ghcup - there was a reason for it, and I am not sure that reason changed.
Also, I completely forgot we are still on stack here, these days I would probably look into using just cabal instead, and then we can avoid that problem with the stack action, and also possibly with the cache and delete that big comment above the cache (similar like we did for our main wasp CI). But ok, we can leave that for some other time.
| # We are setting up haskell via ghcup instead of using haskell/actions/setup | ||
| # because the mentioned gh action can be months late with the latest versions | ||
| # of Stack. |
There was a problem hiding this comment.
Is this reason not valid anymore? Of the action in question easily being late for months with the latest version of Stack?
|
|
||
| build: | ||
| name: Build StrongPath | ||
| runs-on: ${{ matrix.os }} | ||
| needs: code-formatter |
There was a problem hiding this comment.
How is it that we dropped this line? Was it never needed?
| matrix: | ||
| os: | ||
| - ubuntu-latest | ||
| - macos-latest | ||
| # macos-latest is only ARM, when we update to a newer haskell we can run on it |
There was a problem hiding this comment.
| # macos-latest is only ARM, when we update to a newer haskell we can run on it | |
| # We don't currently run on macos-latest since it is ARM and GHC version we use is too old for it. We can add it once we update GHC. |
mrkkrp/ormolu-actionwas renamed tohaskell-actions/run-ormolu@v17)concurrencygroup to cancel in-progress runs when a new one is queues (instead of the non-nativestyfle/cancel-workflow-actionaction)haskell-actions/setupinstead of the manualghcupcalls as they were workarounds that are not needed anymoremacos-latesttomacos-13, asmacos-latestis now an ARM image and our current Haskell version is not compatible (to be solved in Upgrade GHC to 9.2.8 #50)