feat: issue #319 update release github actions for supporting install via ghcup by rajcspsg · Pull Request #353 · well-typed/haskell-debugger

rajcspsg · 2026-08-09T06:05:52Z

@alt-romes @hasufell
I updated the release actions -

Adds a build-hdb-binary job to release.yaml producing .tar.xz archives for x86_64-linux, aarch64-linux, x86_64-apple-darwin, aarch64-apple-darwin at GHC 9.14.1, attached to the GitHub Release.
Updates cabal-version: 3.14 → 3.16 on docs-haskell-debugger (required for GHC 9.14.1's bundled boot Cabal package with Custom-build-type deps like entropy).

Tested the jobs in my main branch here. I tried adding windows support, but seems like there is existing issue so I removed it.
Please review and let me know if you have any feedback.

alt-romes

Thanks for contributing this! It is very worthwhile doing. See my inline comment though.

alt-romes · 2026-08-12T13:59:07Z

+
+      - name: Build hdb
+        shell: bash
+        run: cabal build exe:hdb --enable-executable-stripping


I don't expect that this would work, since hdb must be built with a handful of flags like --enable-executable-dynamic?

More broadly, what is the difference between the existing build-haskell-debugger and build-hdb-binary that you added?

It seems the important bit is to make sure the binary artifacts are uploaded (they currently aren't; just built.)

Also: the previous commit doesn't stand very well on its own; Might as well just squash both if you could.

Thanks!

@alt-romes thanks for your review. I initially verified releases page and couldn't find the built hdb in the release page. Also I didn't find build for different OS/architecture matrix in the build-haskell-debugger job debugger.yaml so I added duplicate one with support for multiple os/architecture.
I think the cleaner approach is update the existing build-haskell-debugger to build for all supported os/arch and upload the artifacts to releases.
I'll do the changes and let you know.

@alt-romes are you fine with this approach?

I think the cleaner approach is update the existing build-haskell-debugger to build for all supported os/arch and upload the artifacts to releases.

Yup!

alt-romes · 2026-08-12T14:00:25Z

After the binaries are uploaded to the Github Release, how does the GHCup support come into play ?

rajcspsg · 2026-08-18T06:35:05Z

After the binaries are uploaded to the Github Release, how does the GHCup support come into play ?

@alt-romes once the releases contains the binary, we need to update ghcup-metadata similar to agda/dhall/purescript.

hasufell · 2026-09-01T05:47:12Z

I've been swamped with life and other work. What's the status here?

hasufell · 2026-09-01T05:52:12Z

What's the status?

rajcspsg · 2026-09-01T05:52:29Z

@hasufell Thanks for getting back. I closed the PR by mistake.

@alt-romes I've updated the github actions to build for linux/mac x86_64 and arm architectures and include the built binaries in the releases. could you please take a look and let me know if this looks good to you?

rajcspsg · 2026-09-01T06:02:24Z

What's the status?

@hasufell I've initially created the pull request adding duplicate github actions and build the hdb executable. I got feedback from @alt-romes and decided to reuse the existing build-haskell-debugger action and upload the extuable to releases page.

I just updated the PR for the review comment. appreciate your feedback.

alt-romes · 2026-09-01T09:40:54Z

Looks good now. Could you squash the commits please and mention in the commit message this enables the GHCup distribution? Also add to the message "Closes #319"

@hasufell if you're available, could you take a quick look too?

alt-romes · 2026-09-01T14:33:37Z

Note: the macOS-13 x64 macOS job takes a LONG time to get picked up.

rajcspsg · 2026-09-01T19:20:50Z

Note: the macOS-13 x64 macOS job takes a LONG time to get picked up.

I think i used old mac github job runner. I updated the runner to macos-15-intel.

rajcspsg · 2026-09-01T23:46:15Z

@alt-romes any idea why the tests in macos is failing?

alt-romes · 2026-09-02T13:54:34Z

@alt-romes any idea why the tests in macos is failing?

I think there's a bug with process spawning on macOS Intel. Re-running solved the CI error. Fortunately, that only seems to be an issue with the non-default option of internal interpreter with runInTerminal proxy. It means if someone runs into this non-transiently, they have a way to work around it (e.g use the default --external-interpreter, or run with the internal interpreter using the cli mode).

I do not want to go out of my way to fix macOS Intel-specific since it's been discontinued. I think it's fine if we build binaries for it though, since there are certainly still many users.

I'm happy to land this now, but there are still two things to address:

ubuntu-24.04-arm and macOS-intel jobs should only be run on the release pipeline (inputs.release is true)
the commits should all be squashed and the commit message should say "enables GHCup distribution" and "Closes Ship via ghcup #319"

Note: if there are issues with the CI bits which are only run on the release pipeline, we'll deal with them afterwards when we do the next release (which we hope to do right after this lands)

hasufell · 2026-09-02T14:35:56Z

Can you describe the architecture of the CI?

What are the platforms for which artifacts will be built?
What are the artifacts precisely?
Are we building for multiple GHC versions?

rajcspsg · 2026-09-03T18:13:39Z

Can you describe the architecture of the CI?

What are the platforms for which artifacts will be built?

What are the artifacts precisely?

Are we building for multiple GHC versions?

@hasufell

What are the platforms for which artifacts will be built? x86_64, aarch, mac-arm and mac-intel. Windows build is not working due to existing open issue.
What are the artifacts precisely? the artifacts are hdb binary compressed in tar.gz
Are we building for multiple GHC versions? Right now we are building only for ghc 9.14.1 as thats the only supported version.

Please let me know if you have any more questions.

hasufell · 2026-09-03T18:24:30Z

Is an hdb release always associated with a specific GHC version and only one?

alt-romes · 2026-09-03T20:22:42Z

Is an hdb release always associated with a specific GHC version and only one?

Yes, to use GHC 9.14.1 you need hdb compiled against 9.14.1 lib. and so on and so forth.

For the time being, the only released supported version is 9.14.1, but we’ll also support 9.14.2 and 10.0 soon

alt-romes · 2026-09-04T09:43:20Z

@hasufell will this setup work well with GHCup? what else should we do to have 3rd party GHCup support? Thanks for your help in advance.

hasufell · 2026-09-04T10:26:01Z

That doesn't quite answer my question.

If hdb 1.2.3 only ever supports GHC 9.14.1, then that's a simple situation. If hdb 1.2.4 suddenly supports GHC 9.14.1 and 10.0.0, then it gets much harder.

alt-romes · 2026-09-04T10:34:22Z

@hasufell sorry, I wasn't very clear:

The hdb executable has to be compiled against the GHC library. If you choose 9.14.1 to compile hdb 1.2.3, you get something like hdb-1.2.3-9.14.1

So, hdb-1.2.3-9.14.1 only ever supports GHC 9.14.1

But, the source of hdb version 1.2.3 can also be compiled with 9.14.2. If you do so, you get hdb-1.2.3-9.14.2.
This is the case because the same features are supported by hdb 1.2.3 regardless of whether you compiled it with 9.14.2 or 9.14.1.

hasufell · 2026-09-04T10:48:55Z

So, will you potentially compile hdb against multiple GHC versions in a single release?

alt-romes · 2026-09-04T10:50:08Z

So, will you potentially compile hdb against multiple GHC versions in a single release?

Yes (though the next release will only be compiled with 9.14.1 specifically)

hasufell · 2026-09-04T11:29:42Z

Then you essentially need an HLS style CI and you need to account for the different bindist mappings across GHC versions to reach an exhaustive configuration.

E.g. GHC 9.12.4 may have the same bindist for debian 10 and 12, but GHC 10.0.0 might be shipped with different ones. That means hdb needs to be built on both debian versions.

The other thing is detecting ABI incompatibility when the user mixes toolchains. HLS has a startup script that does that: https://github.com/haskell/haskell-language-server/blob/master/bindist/wrapper.in

…via ghcup PR #353 Squashed commits: update cabal version to 3.16 chore: issue-319 add github actions to build hdb for linux/mac and release use macos-15-intel runner for mac x86-64 architecture github: extra arch runners only for release.

rajcspsg · 2026-09-08T18:56:03Z

@hasufell thanks for your feedback. I have a question.

I searched how the bindist/wrapper.in is used to build hlshere, is see it have GHC_VERSION used in GNUmakefile.
I'm looking at hls releases, I see the format doesn't have GHC version(hls supporting multiple GHC) in the release artifacts.

May I know why the releases don't have GHC_VERSION?

Saizan · 2026-09-08T19:29:03Z

@rajcspsg btw hdb already checks the runtime version of ghc used matches the one it was compiled with, if that helps. See this function.

Looking at HLS, it gets compiled against ~13 or so distros, spanning several versions of debian, fedora, mint, and ubuntu, plus a generic linux.
I guess most of those are done in docker containers?

So it seems that there are two remaining issues:

Extend the list of OS/arch we prepare binary distros for, matching the binary distros of GHC, with a sensible way to include more in the future.
For each OS/arch, include a binary for each version of ghc supported.

At the moment we support only 9.14.1 though, so (2) is already satisfied.

@rajcspsg would it make sense to you for this PR to focus on (1)? i.e. building for more OSes, and introduce a more general solution for (2) in a later PR?

It would be a good way to contain the scope, especially because we would like to make a new release soon (issue #366) to take advantage of hie-bios-0.20+ and it would be great to have your work merged so the new version is available on ghcup.

rajcspsg · 2026-09-09T06:46:18Z

@Saizan Thanks for your suggestion. That makes this merge request more easier.
I'll get the merge request add support for various binary distributions similar to hls here

rajcspsg · 2026-09-10T04:06:13Z

@Saizan @hasufell @alt-romes I created bindist github actions similar to hls(mostly copied).

Some of the distributions build are failing -

cbits/stack.c: In function 'bco_frame_args_offset':
cbits/stack.c:254:13: error:
     error: a label can only be part of a statement and a declaration is not a statement
      254 |             const void * return_frame_info = frame_entry.return_info;
          |             ^~~~~
    |
254 |             const void * return_frame_info = frame_entry.return_info;
    |             ^

`gcc' failed in phase `C Compiler'. (Exit code: 1)
Error: [Cabal-7125]
Failed to build haskell-debugger-0.13.1.0 (which is required by exe:hdb from haskell-debugger-0.13.1.0).

Any idea about this error?

rajcspsg · 2026-09-10T06:37:44Z

@Saizan @alt-romes @hasufell I fixed the other issues as well. Now all the binary distributions are building fine.
Please review and let me know if you have any questions

Saizan

Looks great, thank you for your work, just some small changes I'd like.

After those, can you also squash the commits to only one or maybe a few if they are truly independent?

hasufell · 2026-09-10T10:32:53Z

@hasufell thanks for your feedback. I have a question.

I searched how the bindist/wrapper.in is used to build hlshere, is see it have GHC_VERSION used in GNUmakefile. I'm looking at hls releases, I see the format doesn't have GHC version(hls supporting multiple GHC) in the release artifacts.

May I know why the releases don't have GHC_VERSION?

I don't quite understand the question. The wrapper does not build HLS. It's probably easier to understand the CI in my fork, because HLS upstream has turned their CI into an autogenerated monstrosity:

Saizan

Great! Would you mind rebasing to squash all commits into one? Otherwise I can do a squash merge if you prefer.

…via ghcup update cabal version to 3.16 chore: issue-319 add github actions to build hdb for linux/mac and release use macos-15-intel runner for mac x86-64 architecture add support for bindist similar to hls add trace to debug in build.sh add use stable ghc update INSTALL and TOOLS to PKG_INSTALL_CMD and PKG_TOOLS respectively fix deb 11 ghc install issue fix cbits compilation issue fix the ubuntu-2004 build waiting due to interactive tz prompt fixed the inputs.is-release bug always log ghc version on GHC_BIN and GHC_LIBDIR

rajcspsg · 2026-09-15T18:37:39Z

Great! Would you mind rebasing to squash all commits into one? Otherwise I can do a squash merge if you prefer.

I squashed the commits into single one.

rajcspsg changed the title ~~feat: issue-319 update release github actions for supporting install …~~ feat: issue-319 update release github actions for supporting install via ghcup Aug 9, 2026

rajcspsg force-pushed the issue-319 branch from d5017f5 to 29302b4 Compare August 11, 2026 15:33

rajcspsg mentioned this pull request Aug 11, 2026

Ship via ghcup #319

Open

alt-romes reviewed Aug 12, 2026

View reviewed changes

rajcspsg closed this Sep 1, 2026

rajcspsg force-pushed the issue-319 branch from 0e6f32a to 10488e6 Compare September 1, 2026 05:43

rajcspsg reopened this Sep 1, 2026

alt-romes reviewed Sep 1, 2026

View reviewed changes

Comment thread .github/workflows/debugger.yaml Outdated

alt-romes reviewed Sep 1, 2026

View reviewed changes

Comment thread .github/workflows/debugger.yaml Outdated

rajcspsg force-pushed the issue-319 branch from 76c6cef to 39d0622 Compare September 1, 2026 22:20

Saizan self-requested a review September 4, 2026 09:48

Saizan mentioned this pull request Sep 7, 2026

Minimize jobs on non-release PRs #367

Merged

rajcspsg force-pushed the issue-319 branch from 5d885d5 to 5f3a940 Compare September 10, 2026 00:39

Saizan requested changes Sep 10, 2026

View reviewed changes

Comment thread README.md

Comment thread .github/workflows/debugger.yaml Outdated

Comment thread bindist/wrapper.in

Saizan approved these changes Sep 15, 2026

View reviewed changes

Saizan changed the title ~~feat: issue-319 update release github actions for supporting install via ghcup~~ feat: issue #319 update release github actions for supporting install via ghcup Sep 15, 2026

rajcspsg force-pushed the issue-319 branch from be62e61 to f48260b Compare September 15, 2026 18:37

Saizan merged commit 8c6e08a into well-typed:master Sep 16, 2026

Conversation

rajcspsg commented Aug 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alt-romes left a comment

Choose a reason for hiding this comment

Uh oh!

alt-romes Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

rajcspsg Aug 18, 2026

Choose a reason for hiding this comment

Uh oh!

rajcspsg Aug 20, 2026

Choose a reason for hiding this comment

Uh oh!

alt-romes Aug 20, 2026

Choose a reason for hiding this comment

Uh oh!

alt-romes commented Aug 12, 2026

Uh oh!

rajcspsg commented Aug 18, 2026

Uh oh!

hasufell commented Sep 1, 2026

Uh oh!

hasufell commented Sep 1, 2026

Uh oh!

rajcspsg commented Sep 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rajcspsg commented Sep 1, 2026

Uh oh!

alt-romes commented Sep 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

alt-romes commented Sep 1, 2026

Uh oh!

rajcspsg commented Sep 1, 2026

Uh oh!

Uh oh!

rajcspsg commented Sep 1, 2026

Uh oh!

alt-romes commented Sep 2, 2026

Uh oh!

hasufell commented Sep 2, 2026

Uh oh!

rajcspsg commented Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hasufell commented Sep 3, 2026

Uh oh!

alt-romes commented Sep 3, 2026

Uh oh!

alt-romes commented Sep 4, 2026

Uh oh!

hasufell commented Sep 4, 2026

Uh oh!

alt-romes commented Sep 4, 2026

Uh oh!

hasufell commented Sep 4, 2026

Uh oh!

alt-romes commented Sep 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hasufell commented Sep 4, 2026

Uh oh!

rajcspsg commented Sep 8, 2026

Uh oh!

Saizan commented Sep 8, 2026

Uh oh!

rajcspsg commented Sep 9, 2026

Uh oh!

rajcspsg commented Sep 10, 2026

Uh oh!

rajcspsg commented Sep 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Saizan left a comment

Choose a reason for hiding this comment

rajcspsg commented Aug 9, 2026 •

edited

Loading

rajcspsg commented Sep 1, 2026 •

edited

Loading

alt-romes commented Sep 1, 2026 •

edited

Loading

rajcspsg commented Sep 3, 2026 •

edited

Loading

alt-romes commented Sep 4, 2026 •

edited

Loading

rajcspsg commented Sep 10, 2026 •

edited

Loading