 | NixOS JVM | 121 Members |
| 27 Servers |
| NixOS JVM | 121 Members |
| 27 Servers |
| Sender | Message | Time |
|---|---|---|
| 17 Jul 2025 | ||
 emily | we use Zulu because we don't have a source build for OpenJDK/OpenJFX on Darwin. that should be easy to fix, but even ignoring that, the value-adds of Zulu over Temurin binaries are:
| 16:44:43 |
| emily | I think
| 16:45:24 |
| emily | * I think
would be sensible | 16:45:27 |
 Alyssa Ross | In reply to @infinidoge:inx.moeWe have done overrides for this before. It's fine. | 16:45:36 |
| emily | (I planned to do this, I just ran out of steam) | 16:45:43 |
| emily | (we also don't need OpenJFX to be part of the OpenJDK bootstrap, it can just be used as a separate package rather than integrated into the OpenJDK build these days I think, I believe all of that logic is ~pointless) | 16:46:13 |
| emily | ah, sorry, I see the Darwin source build was raised. I didn't know there was a PR | 16:46:49 |
| emily | I don't think we need a "demotion to community-supported" status because there is no second-tier community to support it… the Zulu packages just accumulate CVEs even faster than the other ones, it's pretty droppable | 16:47:39 |
 msgilligan | So, my personal priorities ("itches I need to scratch" for other projects I'm working on) are:
And I also want to learn more about Nix and help the community in general, so that's why I want to make:
It's also relatively easy for a newcomer like me to help with updates on some of the binary distributions. And anything I can do to get any JDK 24 or JDK 24 on | 16:49:22 |
| msgilligan | * So, my personal priorities ("itches I need to scratch" for other projects I'm working on) are:
And I also want to learn more about Nix and help the community in general, so that's why I want to make:
It's also relatively easy for a newcomer like me to help with updates on some of the binary distributions. And anything I can do to get any JDK 24 or JDK 25 on | 16:50:08 |
| msgilligan | Another thing I would like to be able to do is develop native command-line tools that are built with GraalVM and contribute them to Nixpkgs. | 16:50:58 |
| msgilligan | I've also found that having an integrated OpenJFX sometimes solves build issues in various circumstances. (In theory this shouldn't be needed, but in practice it is.) | 16:52:50 |
| msgilligan | * So, my personal priorities ("itches I need to scratch" for other projects I'm working on) are:
And I also want to learn more about Nix and help the community in general, so that's why I want to make:
It's also relatively easy for a newcomer like me to help with updates on some of the binary distributions. And anything I can do to get improved JDK 24 or JDK 25 on | 16:53:16 |
 j4m3s | I seem to remember I encountered issues where the graalvm compiled binaries had some references to some jars in the graal derivation which meant it was pulled as a dependency. Kinda defeated the purpose of having small binaries :/ (that info might be outdated) | 16:53:37 |
| msgilligan | When I talk about GraalVM, I mean using native-image to generate binaries, but that is not all that GraalVM does. For the native-image case, if done properly, I"m pretty sure the binaries are stand-alone. But they are very large compared to a C binary. The advantage is they are standalone and start up quickly. | 16:56:08 |
| msgilligan | * When I talk about GraalVM, I mean using native-image to generate binaries, but that is not all that GraalVM does. For the native-image case, if done properly, the binaries are stand-alone. But they are very large compared to a C binary. The advantage is they are standalone and start up quickly. | 16:56:29 |
| msgilligan | Keeping the Zulu packages up-to-date should not be that hard if we get slightly better tooling in place. I'm willing to help with that -- at least until we have an OpenJDK build for Darwin. | 17:00:19 |
| msgilligan | And I want to make JDK 25 EA available -- at least as on overlay -- so I (and others) can begin testing things with JDK 25 and to help expedite support for JDK 25 GA when it is ready. | 17:03:06 |
| emily | I think just s/zulu/temurin/ for Darwin would be fine. I would be somewhat surprised if anyone cares much about OpenJFX on Darwin | 17:11:30 |
| j4m3s | In reply to @msgilligan:matrix.orgIirc (again, I could be wrong) nix looks for /nix/store references in binaries to infer runtime deps. So native-image compiled binaries are light, but since it brings the whole graalvm as a runtime deep, the whole thing is heavy. Even if compiled with --static, the issue is the string ref to graal | 17:11:42 |
| emily | (but if they do, then "make OpenJFX build on Darwin" should be a smaller task than "make OpenJDK build on Darwin") | 17:11:48 |
| emily | split outputs likely solve this | 17:12:05 |
| emily | well, or do you mean the JAR references are purely artificial and not used at runtime at all? | 17:12:19 |
| emily | in that case we can remove-references-to | 17:12:23 |
| j4m3s | Oh really there's a feature tu remove refs ? I skimmed through the manual and didnt find any ;_; | 17:15:15 |
| j4m3s | (Sorry I kinda hijacked the discussion 😅) | 17:15:39 |
 Infinidoge 🏳️⚧️ | I still don't understand $out/nix-support to be honest | 17:15:42 |
| Infinidoge 🏳️⚧️ | That's a directory of witchcraft I seek to understand | 17:15:53 |
| emily | it just replaces the paths with /nix/store/eeeeeeee…-abc/… | 17:16:48 |
| emily | so it's a pretty heavy hammer | 17:16:53 |