NixOS JVM | 133 Members | |
| 28 Servers |
| Sender | Message | Time |
|---|---|---|
| 14 May 2026 | ||
And once the above fix is merged, https://github.com/NixOS/nixpkgs/pull/519958 can be merged to remove jextract-21 which is obsolete and unmaintained. (cc: emily ) | 00:23:40 | |
| 07:38:38 | ||
| 16:32:49 | ||
| 21 May 2026 | ||
| In this comment on a PR to update
I've been doing some investigation and think I have a fix. | 18:22:58 | |
| It turns out that
This should stop a lot of rebuilds (especially in Python packages) based on | 18:26:19 | |
| For Nixpkgs, would it be best to make this one or two PRs? | 18:27:23 | |
Or one PR with two commits: one to remove FOP and the other to add the html target and some missing dependencies it actually needs? | 18:29:43 | |
| I think this is a change to igraphs doc build with a nice side effect. So could be one PR. | 19:13:22 | |
| Thanks a lot for looking into this! | 19:13:29 | |
| We should review other downstream builds that look suspicious and try to remove ourselfs as much as possible, esp from "foundational" packages like igraph. | 19:14:28 | |
| Here it is: https://github.com/NixOS/nixpkgs/pull/522762 | 19:14:44 | |
| It is two commits, and I have verified that the first one (removing fop) works independently, so it could be made into a separate PR if there is some objection to how the HTML docs are being generated, etc. The removal of fop should make the pending Maven update less breaking and might be worth expediting. | 19:21:55 | |
| 22 May 2026 | ||
| 08:42:54 | ||
| 19:18:42 | ||
The igraph PR has been merged. Can someone rebase maven: 3.9.12 -> 3.9.16? | 19:58:58 | |
| The
| 20:04:39 | |
| So to get an up-to-date Maven without breaking ~40 dependencies, for each dependency we need to either:
| 21:20:09 | |
| * So to get an up-to-date Maven without breaking ~40 dependencies, for each dependency we need to either:
| 21:21:09 | |
The advantage to the pom.xml approach is that it can be done independently (i.e. before) the Maven upgrade and won't break upon future Maven upgrades. | 21:23:17 | |
The mvnHash update must occur in the same commit as the Maven update because there is effectively a circular dependency. | 21:24:09 | |
| 23 May 2026 | ||
| Maybe we can convince some of the upstream maintainers to fix plugin versions, so we don't have to. | 14:42:37 | |
| https://issues.apache.org/jira/browse/FOP-3318 | 14:43:36 | |
| You're everywhere 😏 | 14:44:22 | |
| "The turtles in the rabbit hole go all the way down" or something like that. | 14:44:50 | |
| Since I made the PR to fix FOP, I figure I should at least open an issue upstream with a link to it. | 14:46:51 | |
| I think this can be helpful to identify more packages like igraph: https://github.com/craigmbooth/nix-visualize We should try to use it on Java packages and look for nodes that produce lots of downstream dependencies. So basically nodes that depend on something like Maven and then are depended upon by many others. | 14:47:34 | |
| Submitting the patch as a PR seems the logical next step. | 14:48:06 | |
| Yeah. I asked the Nixpkgs maintainer of FOP to open an issue and a PR (after he said it was a good idea) and he said he was too busy. | 14:48:52 | |
| I'be been thinking if there is a way to automate this. | 14:49:21 | |
The rough idea I have is that there should be a way to compose Maven repository entries for packages. I think the mechanism behind mvnHash might be doing something like this. Perhaps the maven package itself could have an output that is a mini-Maven-repo with its default plugin dependencies in it. Then a package like fop could access that through a symlink. | 14:53:02 | |