ghc-9.4.3 haddock crashes in Stackage Nightly

[juhp]

Running Haddock on library for MissingH-1.5.0.1..
                                                                                                                                    
<no location info>: error:                                                                                                              panic! (the 'impossible' happened)                                                                                              
  GHC version 9.4.3:                                              
        funResultTy                                                                                                                 
  BinPackerError size_a5dp obj_a5dq   
  Call stack:                                                     
      CallStack (from HasCallStack):                              
        callStackDoc, called at compiler/GHC/Utils/Panic.hs:182:37 in ghc:GHC.Utils.Panic
        pprPanic, called at compiler/GHC/Core/Type.hs:1334:49 in ghc:GHC.Core.Type

Please report this as a GHC bug:  https://www.haskell.org/ghc/reportabug

[andreasabel]

cabal haddock works for me with GHC 9.4.3.
I can reproduce the problem with stack haddock and resolver: nightly-2022-11-19.

But not sure what I should do on my side here, @juhp.
Maybe you should indeed report it upstream to GHC (and/or stack / haddoc).

[andreasabel]

I cannot reproduce this with nightly-2023-01-14 (GHC 9.4.4), so maybe this issue is resolved, @juhp?

[juhp]

Thanks I agree 👍

[andreasabel]

@juhp Shouldn't you have updated Stackage's build-constraints.yaml accordingly when this issue got closed?
I randomly noticed there was some left-over from this issue:

• Remove MissingH from expected-haddock-failures commercialhaskell/stackage#7659

Anyway, how do all these extra sections (expected failures etc.) in build-constraints.yaml get cleaned up? Is there a procedure to ensure they do not accumulate outdated information?

[juhp]

@juhp Shouldn't you have updated Stackage's build-constraints.yaml accordingly when this issue got closed?

Probably/ideally, sorry, the best way currently to ensure this is still to open a PR yes.

I randomly noticed there was some left-over from this issue:

* [Remove MissingH from expected-haddock-failures commercialhaskell/stackage#7659](https://github.com/commercialhaskell/stackage/pull/7659)

Thank you

Anyway, how do all these extra sections (expected failures etc.) in build-constraints.yaml get cleaned up? Is there a procedure to ensure they do not accumulate outdated information?

Manually 😢 I wish we had more automation/CI

Stackage is not really a CI/QA service, though it sometimes feels like one...
But yeah I too wish things were better. Sometimes I'm really tempted to disable all the extras: tests/benchmarks it is a big overhead for us with the current tooling -- well I guess docs we can't really disable, so yeah maybe they are more important actually and much fewer issues (which is maybe why they are easily forgotten 🤷‍).

[andreasabel]

Thanks, @juhp!
I think it is already good that recently all (? or many?) additions to the special sections are equipped with comments linking to the upstream issues.
Ideally there would be a more systematic evidence collection with automated checks whether the evidence still exists.
E.g.

• if the evidence is some upstream issue, one could check automatically whether the issue is still open.
• if the "evidence" is a certain GHC version, the evidence would go cold if the GHC version was upgraded.
• same if the evidence was a certain version of the package (but I guess this part is already (somewhat) automated by the curator).