| 5 Jun 2021 |
cdepillabout | toonn: I'm not sure what you're trying to compare when you're talking about stackage vs nixpkgs. Nixpkgs generally has about 6000 haskell packages building, while stackage only has about 3000. There is really no comparison. | 13:01:20 |
toonn | cdepillabout: That's all well and good until you try to depend on more than one at the same time. | 13:01:55 |
toonn | Jailbreaking isn't a good option. | 13:02:11 |
cdepillabout | toonn: Ah, you mean depend on more than one version of GHC? | 13:02:24 |
toonn | When I used the infra I had to jailbreak and disable tests constantly. Felt like maintaining my own tiny stackage. | 13:02:55 |
toonn | No, just packages. | 13:03:02 |
toonn | Especially if you need a newer version of anything. | 13:03:28 |
cdepillabout | That's definitely not how I feel at all. | 13:03:39 |
toonn | You lose all the support instantly. | 13:03:43 |
cdepillabout | I mean, really you should have been upstreaming stuff. | 13:03:52 |
sterni (he/him) | part of the pain has been lessened by stackage nightly | 13:04:00 |
sterni (he/him) | and yeah a big problem is I think ppl having an overlay in their dev environment and not contributing fixes back | 13:04:18 |
cdepillabout | * I mean, really you should have been upstreaming those jailbreaks and dontChecks. | 13:04:21 |
toonn | No, I don't think package updates should be upstreamed, especially when, as in my case, they require bumping a ton of deps from the stackage versions. | 13:04:51 |
toonn | It'd unnecessarily break things for others. That's exactly why depending on Stackage is nice. | 13:05:14 |
cdepillabout | No, you really should upstream fixes to Nixpkgs. | 13:05:27 |
cdepillabout | You could at least unbreak the foobar_x_y_z version of the packages. | 13:05:55 |
maralorn | Well, if it includes manually bumping stackage packages, I am not sure. | 13:05:58 |
toonn | I think you're not hearing what I'm saying. The "fix" is just needing a newer version. | 13:06:02 |
toonn | And that newer version needs newer versions of other packages. | 13:06:12 |
cdepillabout | We have the newest versions of all Haskell packages in Nixpkgs. | 13:06:25 |
toonn | Which aren't necessarily compatible with the rest of Stackage. | 13:06:26 |
cdepillabout | Yeah, and it still great when people send fixes for them. | 13:06:43 |
toonn | I think maybe you're talking about the current situation? | 13:06:58 |
toonn | Back when I used the infra Nixpkgs definitely did not have the latest version of everything. | 13:07:15 |
toonn | I didn't know Stackage Nightly even used the latest versions of everything? | 13:07:54 |
toonn | That sounds like you'd be completely giving up the Stackage guarantee? | 13:08:05 |
toonn | What's even the point of using Stackage then? | 13:08:16 |
maralorn | I think toonn is right, when it’s about wanting a newer version of a stackage package. But then otoh that would be worse when using stackage. When we are talking about a package outside of stackage. We nearly always have the newest and we only pin it to an older version because of stackage in very rare cases. | 13:08:21 |
cdepillabout | Huh, maybe that was before my time. As far as I've been using it, Nixpkgs has always had the latest versions of all Haskell packages. Some of them don't work, like you're saying, but we still like getting fixes for them. | 13:08:21 |