Fix #8400 by marking output of cabal list-bin

[andreasabel]

Fix #8400 by marking output of cabal list-bin

The regular output of cabal list-bin should go to stdout always, but be marked for the sake of the testsuite.

• fixes cabal list-bin does not produce marked output #8400
• fixes cabal list-bin is broken #8664

If this PR is merged quickly, we do not need:

• Revert "add marked output to cabal list-bin" #8668

Replaces:

• add marked output to cabal list-bin #8622

[andreasabel]

@bacchanalia : I do not understand why the following assertOutputContains succeeds:

cabal/cabal-testsuite/PackageTests/ListBin/Script/cabal.test.hs

Lines 4 to 8 in bcfc79c

	res <- cabal' "list-bin" ["script.hs"]
	env <- getTestEnv
	cacheDir <- getScriptCacheDirectory $ testCurrentDir env </> "script.hs"
	assertOutputContains cacheDir res

The output is just this:

cabal/cabal-testsuite/PackageTests/ListBin/Script/cabal.out

Lines 2 to 5 in bcfc79c

	Resolving dependencies...
	Build profile: -w ghc-<GHCVER> -O1
	In order, the following will be built:
	- fake-package-0 (exe:cabal-script-script.hs) (first run)

I'd like to change the assertion into something that would fail if list-bin didn't print the path to the script, but first I'd like to understand why the current check does not fail.

[athas]

Looks correct to me.

[Mikolaj]

That's a puzzling CI failure. Looks like a random corruption, but I can't rule out it's somehow caused by this PR. @andreasabel: could you have a look?

[Mikolaj]

LGTM, except for the test result on Windows, which may be wrong due to Windows path separators that, for whatever reason, are not translated properly.

[andreasabel]

LGTM, except for the test result on Windows, which may be wrong due to Windows path separators that, for whatever reason, are not translated properly.

Thanks, good catch: https://github.com/haskell/cabal/actions/runs/3910085986/jobs/6681874680#step:18:138

dist\build\x86_64-windows\ghc-9.4.2\SelectedComponent-1.0.0\build\testexe\testexe.exe

The path is printed with backslashes under Windows, need to fix my acceptance criterion.

[Cmdv]

LGTM 👍

[ulysses4ever]

Good to know about withOutputMarker, and thank you for fixing the issue!

[Bodigrim]

Why not putStrLn?

[Mikolaj]

Ouch, all tests had EOF at the end, so this wasn't visible.

[andreasabel]

Yeah, putStrLn would be better in the normal operation. I guess I got pulled towards putStr because the withOutputMarker puts the EOL character there:

cabal/Cabal/src/Distribution/Simple/Utils.hs

Lines 622 to 625 in a78d77b

	withOutputMarker _ xs =
	"-----BEGIN CABAL OUTPUT-----\n" ++
	withTrailingNewline xs ++
	"-----END CABAL OUTPUT-----\n"

So maybe the correct way is:

  putStr $ withOutputMarker verbosity $ exe ++ "\n"