Fix last bug in testsuite

Cabal/src/Distribution/Simple/GHC/Build.hs

@@ -49,6 +49,12 @@ rapidly reach this limit, in part, due to the long length of cabal v2
 prefixes. To minimize the likelihood, we use
 `makeRelativeToCurrentDirectory` to shorten the paths used in invocations
 (see da6321bb).
+
+However, in executables, we don't do this. It seems that we don't need to do it
+for executable-like components because the linking step, instead of passing as
+an argument the path to each module, it simply passes the module name, the sources dir, and --make.
+RM: I believe we can use --make + module names instead of paths-to-objects
+for linking libraries too (2024-01) (TODO)
 -}
 
 -- | The main build phase of building a component.
@@ -80,8 +86,11 @@ build numJobs pkg_descr = do
   liftIO do
     createDirectoryIfMissingVerbose verbosity True targetDir_absolute
     createDirectoryIfMissingVerbose verbosity True buildTargetDir_absolute
+
+  -- See Note [Build Target Dir vs Target Dir] as well
   targetDir <- makeRelativeToCurrentDirectory targetDir_absolute & liftIO
-  buildTargetDir <- makeRelativeToCurrentDirectory buildTargetDir_absolute & liftIO
+  buildTargetDir <- if isLib then makeRelativeToCurrentDirectory buildTargetDir_absolute & liftIO
+                             else return buildTargetDir_absolute
 
   (ghcProg, _) <- requireProgram verbosity ghcProgram (withPrograms lbi) & liftIO
 

Cabal/src/Distribution/Simple/GHC/Build/Link.hs

@@ -88,6 +88,7 @@ linkOrLoadComponent ghcProg pkg_descr extraSources (buildTargetDir, targetDir) (
             PD.ldOptions bi
               ++ [ "-static"
                  | withFullyStaticExe lbi -- ROMES:TODO: wb withStaticLib??
+                 -- ROMES:TODO: wouldn't this be best handled by re-using the build opt for the way being linked?
                  ]
               -- Pass extra `ld-options` given
               -- through to GHC's linker.