cabal haddock imply enable-documentation

[ulysses4ever]

The problem described below was solved by nuking the store. All good now. Fixing #7462.

---

I hoped that implementing #7462 should be a breath. Alas, an obvious fix turns into a disfunctional cabal haddock: it keeps failing with

  67% (  2 /  3) in 'MyLib'
  Missing documentation for:
    Module header
haddock: internal error: /home/artem/.cabal/store/ghc-9.0.2/acme-cadre-0.1-084e219a2f22159b817e45921b58427b54e7a20f5f1736c7acdba987d88cc4e2/share/doc/html/doc-index.json: openBinaryFile: does not exist (No such file or directory)

For a simple project like this:

❯ cat t7462.cabal
cabal-version:      2.4
name:               t7462
version:            0.1.0.0
author:             Artem Pelenitsyn
maintainer:         [email protected]
extra-source-files: CHANGELOG.md

library
    exposed-modules:  MyLib
    build-depends:    base ^>=4.15.1.0
                    , acme-cadre
    hs-source-dirs:   src
    default-language: Haskell2010

❯ cat src/MyLib.hs 
module MyLib  where
import Acme.Cadre

-- | Some 'car' magic
someFunc :: IO ()
someFunc = putStrLn $ "someFunc" ++ (show $ car (1,2))

-- | Not very different from 'someFunc'
otherFunc :: IO ()
otherFunc = putStrLn "otherFunc"

Putting up this draft in case someone can spot the problem...

---

Please include the following checklist in your PR:

• Patches conform to the coding conventions.
• Any changes that could be relevant to users have been recorded in the changelog.
• The documentation has been updated, if necessary.
• A test added.

Please also shortly describe how you tested your change. Bonus points for added tests!

[andreasabel]

@ulysses4ever : Maybe good to report this internal error upstream...

[ulysses4ever]

@andreasabel fair, but my fear is that it's my change breaking cabal is to blame, not haddock itself. Haddock complains that it can't find something in the cabal store. So I assume the change broke some invariant, which, in turn, made cabal to call haddock with a non-existent input.

I'm very surprised that the CI is green. I wonder how good its testing of cabal haddock with external dependencies is. Because for me this breaks the simplest thing with an external dep possible (as explained in the OP).

[ulysses4ever]

Okay, after applying the nuclear option -- nuking the store -- I get it working fine. Time to think about a test, I guess.

[ulysses4ever]

Thanks to ffaf1, I managed a test. Should be good now.

Okay, I'm having trouble with a test.

Goal: have a package A with haddocks mentioning an entity from an external package with no HTML haddocks built. Calling cabal haddock on A would warn about missing references before this patch and otherwise (finish clean) with the patch.

The question is: how do I depend on an external package in a test? I only see examples of depending on boot packages, which all have HTML haddocks included, so not qualified for the goal. Depending on random Hackage package doesn't seem possible in a test (cabal-tests fails).

Any ideas?

[Mikolaj]

Looks great. How about a changelog file?

[ulysses4ever]

@Mikolaj forgot to stage it... fixed.

[Mikolaj]

Got it. Thank you.

[ulysses4ever]

Oh, I messed up git branches. Will take some time to recover, it seems. All good now.

[andreasabel]

Can it be that this unconditional setting leads to --disable-doc being ignored?

[ulysses4ever]

Yes, I think you're right.