| 5 Jun 2021 |
 cdepillabout | In reply to @sternenseemann:systemli.org
I can do darwin testing and fixing, but I'm noticing that I'm doing it out of a feeling of obligation and it is frustrating me I really want to urge you not to feel obligation here. I think in open source, if someone wants something working, the burden is really on them to get it working. You really shouldn't feel an obligation getting something working that you don't use, especially if it could possibly burn you out. | 01:42:24 |
| cdepillabout | I guess I see our role more as providing a platform and framework that maintainers can use to get their Haskell packages building, rather than the three of us actually being responsible for keeping all the top-level Haskell packages building. | 01:45:25 |
| cdepillabout | We should probably keep pandoc building though... | 01:46:52 |
|  thomasjm joined the room. | 08:46:03 |
 sterni (he/him) | cdepillabout: well that's what I am afraid of, too, but I just can't help myself very often | 10:21:34 |
| sterni (he/him) | I think we should make the broken list limited to the default GHC somehow | 10:23:55 |
| sterni (he/him) | it may be very annoying to deal with marked-as broken packages on GHC 9 which have just been marked as broken because they require the newer base or time library | 10:24:29 |
| sterni (he/him) | although it is probably an extra eval time cost to manage this separately | 10:24:46 |
 maralorn | I am not sure about this. | 10:25:51 |
| maralorn | The broken state between different ghc versions is probably highly correlated if you compare it across the whole package set. | 10:27:13 |
 ahdyt | is that broken state also cause hls to crash? | 10:45:11 |
| maralorn | ahdyt: I don‘t understand. | 10:45:39 |
| sterni (he/him) | maralorn: but it doesn't matter since we don't build the other sets anyways | 11:05:16 |
| sterni (he/him) | it is let fan service | 11:05:19 |
| sterni (he/him) | * it is les fan service | 11:05:25 |
| sterni (he/him) | * it is less fan service | 11:05:29 |
| sterni (he/him) | but the non default package sets are always gonna be broken in suprising ways… | 11:05:44 |
| maralorn | True | 11:05:54 |
| maralorn | Yeah, it probably wouldn‘t be bad in principle. But I am not sure how to do it. | 11:06:43 |
| maralorn | I think using the broken flags in hackage-packages.nix seems reasonable. | 11:07:05 |
| maralorn | Maybe we insert a step, when evaluating a non standard ghc that strips all broken flags right after loading hackage-packages but before applying any other overrides? | 11:08:11 |
| maralorn | That way we pay no additional cost on the standard ghc. | 11:08:24 |
| maralorn | And can keep the config as is. | 11:08:51 |
| sterni (he/him) | the way I'd do it is to pass an argument defaultCompilerBroken to hackage-packages.nix | 11:32:03 |
| sterni (he/him) | and always emit broken = defaultCompilerBroken; | 11:32:12 |
| maralorn | hm, that will be a lot more characters. And I don‘t understand it exactly. That variable would be a bool we set to true when we are evaluating for the default compiler? | 12:33:06 |
| maralorn | If so I imagine a broken = forDefaultCompiler would work better? | 12:33:34 |
| maralorn | Regarding to stackage-nightly. I am curious how much worse it would be to just not update stackage at all for a few weeks (or maybe even months?) compared to staying on LTS. | 12:37:01 |
| maralorn | otoh, when switching to LTS would mean nearly no package downgrades we could just switch. | 12:38:14 |
| maralorn | Or do you think the community will grow unsatisfied if we hop in between?^^ | 12:39:03 |
