Re: Redesigning GHC’s build system

The blog GHC Mutterings doesn’t allow posting of comments without logging in. Since I have absolutely no interest at all in creating yet another online account I hope they allow pingbacks :-)

In this post they discuss the planned redesign of GHC’s build system, and here’s my comment:

The goal of moving away from make is probably a good one, but when failing to get Cabal to fit your needs you revert back to redesigning it using make again? There are numerous other build tools out there, many of them remarkably better than make. What other tools were considered and why were they discarded? I simply refuse to believe that the only build tool to make the shortlist was make. :-)

⟸ Keysafe and A first for me... ⟹


FWIW, make it is hard to beat in mindshare, ease of installation and, related, build dependencies. Ian and SimonM know make, and one big reason was to also make the build-system more accessible to new developers (ATM, a lot of the logic is implemented in Cabal itself or even indirectly through code generated by Cabal). Switching to a completely new system would therefore be a loss for all participants; so it was pretty much a no-brainer. Sure, make has its weaknesses, but so have all the other tools out there. At one point, GHC may be able to completely use Cabal and it’s current in-development make-framework, but we cannot expect this to become better than make anytime soon.

Leave a comment