default enableSeparateDataOutput to true

[TravisWhitaker]

No description provided.

[TravisWhitaker]

Fixes things like #899 (comment) and #1473 (comment). There's no downside to this I can think of.

[hamishmack]

bors try

[hamishmack]

I think this will change alex and happy and therefor rebuild all the GHCs. So I've taken it back to ifdLevel 0 to avoid doing it during hydra eval.

[iohk-bors]

try

Timed out.

[hamishmack]

bors try

[iohk-bors]

try

Timed out.

[hamishmack]

bors try

[iohk-bors]

try

Build succeeded:

• ci/hydra:Cardano:haskell-nix:required
• buildkite/haskell-dot-nix

[hamishmack]

bors try

[iohk-bors]

try

Timed out.

[TravisWhitaker]

@hamishmack simply to cure my own ignorance, what exactly does this do?

So I've taken it back to ifdLevel 0 to avoid doing it during hydra eval.

Or at the very least, where can I read about what the ifdLevel stuff in haskell.nix does?

[hamishmack]

It looks like the ifdLevel 3 jobset has now built on hydra so this should be safe to merge now.

Because of the recursive nature of Haskell.nix (GHC is used to build the tools used to build GHC and work out what cabal components exist), some changes will result in hydra rebuilding the world when it tries to work out what should be in jobset (eval time). ifdLevel is a way we can limit the jobset to just the jobs we know about with the help of materialised nix (pre-calculated output of the tools that cause bootstrap issues).

[hamishmack]

Hydra build results are here https://hydra.iohk.io/jobset/Cardano/haskell-nix-bors-trying

[michaelpj]

If #1477 (comment) is true, I'm surprised we didn't have any failures in the jobset 🤔 I guess it isn't a problem?

[michaelpj]

Ignore me, that comment is about stripping, not data outputs.