--semaphore is not compatible with build-type: Custom

[Bodigrim]

Describe the bug
Cabal HEAD fails to build any package with build-type: Custom when -j --semaphore is passed.

To Reproduce
Steps to reproduce the behavior:

$ cabal unpack ghc-paths-0.1.0.12 && cd ghc-paths-0.1.0.12 && cabal build -w ghc-9.8.0.20230809 -j --semaphore --allow-newer
Resolving dependencies...
Build profile: -w ghc-9.8.0.20230809 -O1
In order, the following will be built (use -v for more details):
 - ghc-paths-0.1.0.12 (lib:ghc-paths) (first run)
Created semaphore called cabal_semaphore_3 with 8 slots.
[1 of 2] Compiling Main             ( /private/tmp/ghc-paths-0.1.0.12/dist-newstyle/build/aarch64-osx/ghc-9.8.0.20230809/ghc-paths-0.1.0.12/setup/setup.hs, /private/tmp/ghc-paths-0.1.0.12/dist-newstyle/build/aarch64-osx/ghc-9.8.0.20230809/ghc-paths-0.1.0.12/setup/Main.o )
[2 of 2] Linking /private/tmp/ghc-paths-0.1.0.12/dist-newstyle/build/aarch64-osx/ghc-9.8.0.20230809/ghc-paths-0.1.0.12/setup/setup
Configuring ghc-paths-0.1.0.12...
unrecognized 'build' option `--semaphore=cabal_semaphore_3'
Error: cabal: Failed to build ghc-paths-0.1.0.12.

Expected behavior
The build should complete successfully. If I remove -j --semaphore, all works well.

System information

• Operating system: macOS
• cabal: HEAD (41225bb)
• ghc: 9.8.0.20230809

[Bodigrim]

If the full support is challenging in the run down towards Cabal 3.12 release, IMHO it would be fine for defaultMainWithHooks ignore any --semaphore=... arguments for now. Just do not fail please :)

CC @sheaf @mpickering @fgaz.

[mpickering]

I think the problem is just the version number on the guard in setupHsBuildFlags.

Perhaps that should be 3,11,0,0 rather than 3,9,0,0. Could a Cabal developer please comment about what the first version which --semaphore is supported is going to be?

[Bodigrim]

@Mikolaj @Kleidukos @ulysses4ever could someone please chime in?

[ulysses4ever]

3.12

[Mikolaj]

3.12

Which means that 3,11,0,0 (the version of master until we release 3.12) should be fine if @mpickering's diagnosis is correct. Any volunteers for a PR?

[mpickering]

#9221