GitLab

So we don't try to recreate directories just because they are older
than mkdirhier, since mkdirhier won't touch them.

And add a comment explaining why these options are here

You can now have multiple versions of a package installed, and
gen_contents_index will do the right thing.

This means that we get the right program name in error messages etc.

See comments for details.

This means we can remove some conditional stuff from the Makefiles,
and means the testsuite doesn't have to work out whether or not it's
on Windows.

There is now less makefile code duplication.
Also, stage2 packages are registered after GHC, which is important as
they might depend on the ghc package (as dph now does).

We now use an msys/mingw perl tarball

This is the case on Windows (as we have no shell wrappers), and mingw's
shell doesn't like an empty list of things in a "for" command.

Now that we make bindists during validate, we can't assume that ISCC
is available.

To quiet those warnings from make about not being able to delete
directories.

There are now 104 calls to mkdirhier, down from 1201, when validating.

This should catch make going into an infinite loop.

We now do all the hard work in a bindistprep subdirectory, and just move
the result to the root directory. This way we can delete anything in
bindistprep/ without worrying about deleting anything important.

Haddock uses TcRnDriver.tcRnGetInfo, which is only available if
GHCI is built. Set HADDOC_DOCS to NO if GhcWithInterpreter is NO,
and disable the haddock build if HADDOC_DOCS = NO.

This fixes at least one problem: libHSffi-ghc<version>.so wasn't being
cleaned, because $(dyn_libsuf) wasn't defined while cleaning.