Stackage LTS and GHC 8.0

The release of GHC 8.0.1 has recently been announced. Hooray! People are already asking about when LTS Haskell will include the new GHC. While I’m also excited for this to happen as soon as possible, it’s worth taking a look at what happened last time in order to set realistic expectations.

Here’s the rough timeline as observed last year:

• March 27: GHC 7.10.1 release announced
• April 2: Stackage LTS 2.0 released using GHC 7.8.4 (again)
• May 5: Stackage nightly switched to GHC 7.10.1
• July 29: GHC 7.10.2 release announced
• Aug 6: Stackage nightly switched to GHC 7.10.2
• Aug 12: Stackage LTS 3.0 released using GHC 7.10.2

It took about 4.5 months from the release of GHC 7.10 until its inclusion in Stackage LTS.

I’d like to see this time span shortened, but I don’t expect it to be much shorter this time around. Here’s an optimistic estimate of how I expect it to go down. This is just my personal estimate; specific timeline decisions are still being discussed amongst the Stackage curators.

* May 21: GHC 8.0.1 release announced
* June: Stackage LTS 6.0 released using GHC 7.10.3 (again)
* June: Stackage nightly switched to GHC 8.0.1
* Sept: Stackage LTS 7.0 released with GHC 8.0.1

There are a few reasons we might delay the release of LTS 7.0 with GHC 8.0.

First and foremost is the obvious: the whole package ecosystem needs to be ready. I am optimistic that this can be accomplished by September. Perhaps sooner. I expect some nightly snapshots to be available in the next few weeks that will be quite useable, if not the full Stackage experience.

Another reason to delay an LTS is because each LTS lasts for a minimum of 3 months before we start the next one (except in the case of LTS 4 which we cut short due to issues with aeson-0.10). LTS 5 has been around for about 4 months, and we’re itching to make a new LTS with aeson-0.11. So we intend to release LTS 6 soon, and then it’ll be at least 3 months until LTS 7.

One more reason to delay an LTS is because each LTS is also pegged to a particular compiler version. If GHC 8.0.2 is close to being released, we’ll probably want to delay the next LTS until it is released, like we did last year with GHC 7.10.2.

I hope this sets some clear expectations and explains some of the reasons why LTS Haskell is probably going to take a couple months to adopt GHC 8.0. LTS Haskell isn’t on bleeding edge, and that’s the whole point. LTS Haskell lags a little behind the latest and greatest in order to deliver a stable and cohesive Haskell experience.