Cabal: Cabal release for GHC 8.4

[bgamari]

GHC 8.4 is quickly approaching. What is the status here?

[phadej]

I need to fix a bug in https://github.com/haskell/cabal/issues/5055, will do that tomorrow.

[23Skidoo]

I'll create a branch once that is merged.

[angerman]

We need #4874 for hadrian (reloc).

[angerman]

There should be no release before #5053 is merged.

[phadej]

• #5053 is cabal-install change, so it's not a blocker for Cabal branch.
• #4874 has Cabal changes, so we have to get that in too.

[phadej]

@bgamari is https://ghc.haskell.org/trac/ghc/wiki/Commentary/Libraries/VersionHistory correct, time will stay at 1.8.0.2. @hvr when parsec-3.1.12 will be released? (Do we want to have >=3.1.12 lower bound?)

[Ericson2314]

Hoping to land https://github.com/haskell/cabal/pull/4383, so we will only have 2.0 allowing ambiguous libraries

[phadej]

AFAICS everything is in, can we cut the 2.2 branch?

[angerman]

4874

[phadej]

@angerman hopefully it will be merged by the time @23Skidoo notices my message! :)

[angerman]

Yea, I'm running some final hadrian tests, and will fix #5111 for inclusion (which unbreaks #4892, which broken when trying to get the flags to hide... for some reason my cross validation ran against some stale code :( -- it's now only blocked by the use of LambdaCase, I believe; will fix.)

[Ericson2314]

Ooo I hope I make it. https://github.com/haskell/cabal/pull/5100 also should be finished as we've got spurious warnings right now without it.

[phadej]

No please, no #5100. Or any other non-trivial patches.

• Fixing bugs :+1:
• Writing docs :+1:
• Improving error messages: :+1:
• Breaking API :-1:

The branch had to be cut for stabilization already 20 days ago (Jan 15th). The feature creep was enormous, let's not make it any worse.

ping @23Skidoo

[23Skidoo]

+1 on not including #5100 at this stage. I'll cut the branch later today or tomorrow, the remaining fixes will be cherry-picked.

[Ericson2314]

The branch had to be cut for stabilization already 20 days ago

A cabal branch or GHC branch?

[23Skidoo]

Cabal branch.

[Ericson2314]

Understood now, talking to @phadej on IRC. The plan was for Jan 15, but it has already slipped. Never mind me then.

[23Skidoo]

Yep, there was an announcement on cabal-devel: https://mail.haskell.org/pipermail/cabal-devel/2017-December/010410.html.

[angerman]

4874 is now green as well (after we gave travis a chance to turn green again ;-))

[23Skidoo]

Created the 2.2 branch.

[bgamari]

As documented in #14862, the HEAD of Cabal 2.2 breaks the ghc-8.4. @angerman has put quite a bit of time into trying to identify the culprit but there is currently no solution. As the release is supposed to go out tomorrow, I think we need to start looking for alternatives.

The last known working commit is d2cf3f13da. I recommend that this commit be designated 2.2.0.0 so we can release GHC in the same state that the release candidate was cut in. It's hard to tell precisely what fixes we will miss out on by doing this, but a quick glance suggests that it's nothing particularly awful.

[23Skidoo]

If we're sure that the basedir patch is the source of the problems, I can just revert it.

[hvr]

I'd like to look into the windows issue; this must be something simple... it's really just about a backslash getting eaten somewhere (the trouble was afaik that @angerman had no windows build env to try/debug it interactively)

[quasicomputational]

Cabal 2.2 has been released, the linked bugs have been resolved, and there aren't howls of pain about brokenness, so I guess this issue's served its purpose.