14 Oct 2024 |
Tristan Ross | In reply to @alexfmpe:matrix.org 3.24 or 32.4 ? 3.24 because the minor version has always been 2 digits afaik | 21:11:02 |
alexfmpe | I guess it works fine until 99.99 | 21:12:16 |
Tristan Ross | Yeah lol | 21:14:04 |
Tristan Ross | If it works, it works | 21:14:12 |
sterni | In reply to @emilazy:matrix.org FWIW, pkg-name_ver_sion is the Nixpkgs standard well, those attributes predate any sort of established standard for anything in nixpkgs | 21:55:20 |
sterni | honestly the only realistic scenario for a collision has been introduced by our own maralorn since the ghc{MAJOR}{MINOR} could feasibly collide with ghc{MAJOR}{MINOR}{PATCH} as soon as MINOR reaches the double digits | 21:56:41 |
sterni | So maybe we should remove that or relegate them to the realm of aliases :p | 21:56:56 |
sterni | In reply to @hellwolf:matrix.org maybe that would be the right timing for cleaning up this technical debt? the “technical debt” is a public API of nixpkgs that production code depends on | 21:57:42 |
sterni | emily: I don't think we have the bandwidth at the moment to deal with the 8.10 question in time for 24.11, to be honest | 21:58:50 |
Tristan Ross | Yeah, I think we're a bit too far into 24.11 for certain things | 21:59:56 |
Tristan Ross | Maybe post release would be a good time? | 22:00:23 |
sterni | Yeah, seems like a decent idea. Then we can figure out what to nuke and whether to do it in 25.05 or 25.11 | 22:01:42 |
Tristan Ross | Yeah | 22:03:49 |
Tristan Ross | I was thinking we can at least get a survey out now and give it a few months to collect | 22:04:10 |
Tristan Ross | I'm not sure what the influx of responses would be. | 22:04:33 |
Tristan Ross | So maximizing the possible amount of responses would be good so we deal with the least amount of pain . | 22:05:17 |
emily | In reply to @sternenseemann:systemli.org emily: I don't think we have the bandwidth at the moment to deal with the 8.10 question in time for 24.11, to be honest I was hoping that a summary post / survey could be linked in the release notes before 24.11 release so that it reaches as many people as possible | 22:12:46 |
emily | not that the data would be collected / a decision made by then :) | 22:12:52 |
maralorn | In reply to @sternenseemann:systemli.org the “technical debt” is a public API of nixpkgs that production code depends on I think there is a way to phase them out which won’t actually need a migration by anyone, it will just take a while. | 23:01:28 |
maralorn | Step 1. Provide underscored aliases for everything. | 23:01:47 |
maralorn | Step 2. don’t add ununderscored versions for newer ghcs instead just give them an informative throw. | 23:02:30 |
maralorn | Step 3. after a while don’t add the informative throws for newer versions anymore. | 23:03:00 |
maralorn | This way no production code will be harmed. Only trouble will be when people use the old scheme without watching for the feedback. | 23:04:03 |
maralorn | In theory we could also do something with warnings. | 23:05:53 |
maralorn | Of course, this is not a solution if we will never remove ghc 8.10.7. 😛 | 23:07:12 |
maralorn | In reply to @sternenseemann:systemli.org honestly the only realistic scenario for a collision has been introduced by our own maralorn since the ghc{MAJOR}{MINOR} could feasibly collide with ghc{MAJOR}{MINOR}{PATCH} as soon as MINOR reaches the double digits I still believe that change was totally worth it. And I will remind you of this point when we reach ghc 81.0.7 😛 | 23:08:05 |
maralorn | sterni: Will you were bumping ghcHEAD why only bump it by 13 days if you could bump it by 6 months? | 23:34:33 |
maralorn | https://github.com/NixOS/nixpkgs/pull/346720/files/46ee759dd57d2bc8b010778d1e735ab334057b0c..c91c22341c8d025d02d03ab0c1aff268d4fc6ab9 | 23:35:11 |
sterni | https://github.com/NixOS/nixpkgs/pull/346720/commits/c91c22341c8d025d02d03ab0c1aff268d4fc6ab9 | 23:36:52 |
maralorn | Uff. | 23:39:17 |