If there is a ghc parse error, sometimes cabal doesn't return an error

I noticed this the other day at work, but I haven't had a chance to investigate it yet. This is a ticket to remind someone (or myself in a few weeks) to do so.

Basically, I had a script like this I was calling to save time:

```
cabal configure && cabal build && ./dist/some/path/foo args
```

This worked fine on type errors. `cabal configure` would pass, `cabal build` would fail and the old version of `foo` would never run. _However_, if I introduced a parse error into the source tree, I would get the errors from `cabal build` spewed to the terminal but `foo` would still run directly afterwards (`&&` wasn't short circuiting...)

The cause is most likely that cabal is not returning an error code to the shell when GHC has a parse error. Or something. Could it even be a ghc bug?

Reproduce the reported command chain, `cabal configure && cabal build && ./dist/some/path/foo args`, using both a type error and a parse error. Check the exit status from `cabal build` and determine whether Cabal or GHC is responsible; done means a parse error prevents the following command from running.

Written by the indexing model from the issue text.