fix(Cabal-syntax): Remove left-over file after #8980

[andreabedini]

In #8980 we started generating Lexer.hs with Cabal rather than as a external step. That PR forgets to remove the generatex file though which this PR rectifies.

Closes #9196

PS: now I wonder whether Cabal should complain/fail if an auto-generated file is already present in the source distribution. 🤔

[mpickering]

Thanks @andreabedini

[Mikolaj]

LGTM. Please feel free to set the merge_me label.

[Bodigrim]

It seems this change breaks GHC bootstrapping process: https://gitlab.haskell.org/ghc/ghc/-/jobs/1665720#L5755

Without Lexer.hs one cannot bootstrap GHC on a new platform, which does not have Alex available (because, well, it does not have GHC available yet).

Relevant part of https://gitlab.haskell.org/ghc/ghc/-/blob/master/.gitlab-ci.yml#L691:

    # Bootstrapping should not depend on HAPPY or ALEX so set them to false
    # so the build fails if they are invoked.
    - export HAPPY=/bin/false; export ALEX=/bin/false

Could Lexer.hs please be restored back?

[Bodigrim]

Just restoring Lexer.hs does not help, because now there are both Lexer.hs and Lexer.x in the same folder. I'm afraid we need to revert key parts of #8980, moving Lexer.x back to boot/ folder and restoring Makefile rule.

[mpickering]

@Bodigrim I will look into this tomorrow about whether we need to patch hadrian to resolve this.

[andreabedini]

I believe @hamishmack had run into this issue with Hadrian already and make a fix in Haskell.nix.
I'll look at this today too.

[hamishmack]

I only made a branch of GHC that bumped the Cabal submodule to a version with this PR. https://gitlab.haskell.org/hamishmack/ghc/-/tree/hkm/bump-Cabal

[andreabedini]

@Bodigrim @mpickering Is the quoted comment still applicable/relevant?

I admit that I don't have much experience with boostrapping or cross-compiling but I ask because just before that we bootstrap hadrian:

    - python3 ./hadrian/bootstrap/bootstrap.py -w $GHC --bootstrap-sources hadrian-bootstrap-sources-$GHC_VERSION.tar.gz
    - export HADRIAN_PATH="$PWD/_build/bin/hadrian"
    - .gitlab/ci.sh setup
    # Bootstrapping should not depend on HAPPY or ALEX so set them to false
    # so the build fails if they are invoked.

and, in this repository, we have tests to check we can bootstrap cabal-install in the same way (which also builds Cabal-syntax).

Cannot Cabal-syntax be compiled exactly as we compile hadrian? I am happy to help with figuring out build plans it that's of any help.

[mpickering]

@andreabedini I have pushed a commit which will hopefully fix hadrian (https://gitlab.haskell.org/ghc/ghc/-/merge_requests/11274/diffs?commit_id=326719c1d5e43013b660e138536c8042c7e3f197).

[Mikolaj]

Any progress here? This sounds mildly alarming, but I have no clue.

[mpickering]

This was fixed on the ghc side and we successfully bumped the Cabal submodule.

[Mikolaj]

Oh, so I assume there's nothing to do cabal-side. Great, so this PR is closed-closed.