Revert style2str change in b547ead58bf09bb838c13f02afb2f1042ad1bc7c

[phadej]

if elabLocalToProject says package is local, than it is. inplace packages are things which depend on local packages, but don't have local sources. (e.g. hackage-security in Cabal project). And global packages are the ones which can go to the store.

The original change didn't include changelog for this part of the change nor tests, so it's an easy revert.

[phadej]

Ok, so this doesn't work as simply either. I wonder what the failing tests does

       -- | Is this package one of the ones specified by location in the
       -- project file? (As opposed to a dependency, or a named package pulled
       -- in)
       elabLocalToProject         :: Bool,

       -- | Are we going to build and install this package to the store, or are
       -- we going to build it and register it locally.
       elabBuildStyle             :: BuildStyle,

field description doesn't seem to suggest these are coupled. There are only three valid combinations of these three though (arguably)

• local=False, buildStyle=BuildInplaceOnly: (inplace) e.g. packages pulled from Hackage, which depend on project packages (hackage-security)
• local=False, buildStyle=BuildAndInstall: (global)dependencies, e.g. async
• local=True, buildStyle=BuildInPlaceOnly: (local) project packages, package: Cabal.
• local=True, buildStyle=BuildAndInstall: AFAIU this shouldn't happen. We shouldn't install local packages to store. (We can sdist them, making them non-local, and then install).

inplace and local should have dist-dir as they are built inplace (BuildInplaceOnly) to the project. global packages are not built (they can be in store cache).

[phadej]

As this is not as easy fix as I hoped I close the PR. The issue is there for grabs