Regression on master: building hashable-1.3.0.0 fails with `Duplicate modules in benchmark`

[andreasabel]

Current master (e74a53b) of cabal-install fails to build hashable-1.3.0.0:

$ cabal install hashable-1.3.0.0
Resolving dependencies...
Build profile: -w ghc-8.6.5 -O1
In order, the following will be built (use -v for more details):
 - hashable-1.3.0.0 (lib) (requires build)
Configuring library for hashable-1.3.0.0..
Error: cabal: Duplicate modules in benchmark: Data.Hashable.RandomSource

Error: cabal: Failed to build hashable-1.3.0.0. The failure occurred during
the configure step.

$ cabal --version
cabal-install version 3.7
compiled using version 3.7.0.0 of the Cabal library 

Build succeeds with version 3.6:

$ cabal-3.6 install -w ghc-8.10.7 hashable-1.3.0.0
...
Installing library in /Users/abel/.cabal/store/ghc-8.10.7/incoming/new-49369/Users/abel/.cabal/store/ghc-8.10.7/hshbl-1.3.0.0-02ec191e/lib

$ cabal-3.6 --version
cabal-install version 3.6.0.0
compiled using version 3.6.1.0 of the Cabal library 

Build also succeeds with Setup.hs:

$ cabal get hashable-1.3.0.0
$ cd hashable-1.3.0.0
$ runhaskell Setup.hs configure
$ runhaskell Setup.hs build

(However, runhaskell Setup.hs configure --enable-benchmarks does not succeed to construct a build plan.)

This is the content of the cabal.project.local file:

$ cat cabal.project.local 
ignore-project: False
benchmarks: True

[Mikolaj]

But this error is right, there is a duplicated other-modules: Data.Hashable.RandomSource.

I think this might have been introduced in a PR that fixed the problem of duplicate module errors in cabals with if-then-else.

[phadej]

That's actually also a mistake in hashable-1.3.0.0 file, if I have the same module in top-level exposed-modules:

% cabal check              
Warning: The package will not build sanely due to these errors:
Warning: Duplicate modules in library: Data.Hashable
Warning: Hackage would reject this package.

but cabal check doesn't catch that when the duplicate is "hidden" in the conditional.

However, something weird happens in cabal anyway as the conditional is (in https://hackage.haskell.org/package/hashable-1.3.0.0/hashable.cabal)

  if impl(ghc >= 6.8)
    Ghc-options: -fwarn-tabs
  else
    c-sources:         cbits/getRandomBytes.c
    other-modules:     Data.Hashable.RandomSource
    if os(windows)
      extra-libraries: advapi32

the alternative branch shouldn't be picked up!

---

But I cannot reproduce it on my machine:

/code/shared-haskell/cabal/dist-newstyle/build/x86_64-linux/ghc-8.6.5/cabal-install-3.7.0.0/x/cabal/build/cabal/cabal install hashable-1.3.0.0
Warning: /cabal/config: Unrecognized field world-file on line 28
Warning: /cabal/config: Unrecognized field world-file on line 28
Warning: both implicit and explicit configuration is present.When using configuration(s) from /codetmp/cabal.project.local, the following errors occurred:
The package location glob './*.cabal' does not match any files or directories.

and without cabal.project.local:

/code/shared-haskell/cabal/dist-newstyle/build/x86_64-linux/ghc-8.6.5/cabal-install-3.7.0.0/x/cabal/build/cabal/cabal install hashable-1.3.0.0
Warning: /cabal/config: Unrecognized field world-file on line 28
Warning: /cabal/config: Unrecognized field world-file on line 28
Resolving dependencies...
Up to date
Warning:
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@ WARNING: Installation might not be completed as desired! @
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
The command "cabal install [TARGETS]" doesn't expose libraries.
* You might have wanted to add them as dependencies to your package. In this
case add "hashable" to the build-depends field(s) of your package's .cabal
file.
* You might have wanted to add them to a GHC environment. In this case use
"cabal install --lib hashable". The "--lib" flag is provisional: see
https://github.com/haskell/cabal/issues/6481 for more information.

[phadej]

But this error is right, there is a duplicated other-modules: Data.Hashable.RandomSource.

I think this might have been introduced in a PR that fixed the problem of duplicate module errors in cabals with if-then-else.

Nice: I see:

hashable-1.3.0.0 % /code/shared-haskell/cabal/dist-newstyle/build/x86_64-linux/ghc-8.6.5/cabal-install-3.7.0.0/x/cabal/build/cabal/cabal check                   
Warning: The package will not build sanely due to these errors:
Warning: Duplicate modules in benchmark: Data.Hashable.RandomSource
Warning: These warnings may cause trouble when distributing the package:
Warning: 'ghc-options: -O2' is rarely needed. Check that it is giving a real
benefit and not just imposing longer compile times on your users.
Warning: Hackage would reject this package.

this is great improvement!

[andreasabel]

Ok, good, then this issue should be fixed on hackage for hashable-1.3.0.0.
(I suppose I need not open an issue for this, @phadej.)

Btw, I came by this issue when trying to build phadej/trustee. It picked up hashable-1.3.0.0 via some other dependency (hashable is not mentioned in trustee.cabal). Since trustee is using a certain index state of hackage, fixing this on hackage might not help for building trustee.

[phadej]

Ok, good, then this issue should be fixed on hackage for

I don't think we can fix it. Revisions don't allow editing module lists at the moment.

Arguably we should be able to remove duplicates, I guess that for any flag assignment / static conditions (os, impl, etc.) we still shouldn't be able to remove any modules, except a duplicate? It's tricky. (the every flag assignment / static condition bit).