| 30 Jan 2024 |
hexa | why does https://reproducible.nixos.org/nixos-iso-gnome-runtime/ not list firefox-unwrapped? | 22:36:41 |
hexa | I would be surprised if it would be reproducible | 22:37:03 |
hexa | * I would be surprised if it was reproducible | 22:37:09 |
raboof | hmm, that's weird indeed - I'm excluding a few slow builds there (noto-fonts-emoji,xgcc,clang,llvm,openssl,openssh,webkitgtk,qtwebengine,ghc-binary,firefox), but I'd expect them to show up in 'unchecked paths'. Probably I made a logic mistake somewhere :) | 22:38:17 |
| 31 Jan 2024 |
hexa | https://github.com/NixOS/nixpkgs/pull/285159 | 03:31:30 |
hexa | alrighty | 03:31:33 |
hexa | that'll make the non-pgo consumers (firefox/thunderbird, rest untested) reproducible | 03:31:52 |
hexa | * that'll make the non-pgo flavors (firefox/thunderbird, rest untested) reproducible | 03:32:03 |
| @federicodschonborn:matrix.org changed their profile picture. | 03:36:22 |
| @federicodschonborn:matrix.org changed their profile picture. | 06:21:56 |
Pol | Thank you hexa ! | 07:55:06 |
| @asymmetric:matrix.dapp.org.uk left the room. | 10:47:20 |
| @grossmap:in.tum.de joined the room. | 21:40:57 |
| 2 Feb 2024 |
| Specx joined the room. | 07:11:05 |
| 3 Feb 2024 |
| raboof changed their display name from raboof to raboof @FOSDEM. | 07:38:16 |
| Martin Schwaighofer changed their display name from Martin Schwaighofer to Martin Schwaighofer@FOSDEM. | 13:13:48 |
| Martin Schwaighofer set a profile picture. | 13:14:37 |
| 4 Feb 2024 |
| Bryan Honof changed their profile picture. | 11:31:08 |
| Meliketoast changed their display name from Emil NordgÄrd to Meliketoast. | 14:04:30 |
| raboof changed their display name from raboof @FOSDEM to raboof. | 17:40:57 |
| 5 Feb 2024 |
| 98765abc set a profile picture. | 04:45:10 |
| 98765abc changed their profile picture. | 04:48:27 |
| bbenno joined the room. | 13:06:21 |
| 6 Feb 2024 |
| haraldh joined the room. | 07:36:29 |
| Moritz Sanft set a profile picture. | 17:17:57 |
Toma | So, some of you might remember that I started working on making more java apps reproducible in https://github.com/NixOS/nixpkgs/issues/278518
and currently one of the roadblocks is that the hook that's used (canonicalize-jars-hook) is a little bit broken.
I had opened a PR which tidied it up a bit (https://github.com/NixOS/nixpkgs/pull/278322/files), but there were several weird issues and special cases that came up.
One of the issues was that unzipping and rezipping jars in jars (e.g. dependencies packaged by springboot) caused it to not work anymore. (Probably due to this method not preserving the original compression-type)
But I found that just using zip-traversal can mitigate this issue. Example: https://github.com/NixOS/nixpkgs/pull/283305/files
I was pretty happy with the solution other than it would require java to run if I wanted to package it standalone.
So inspired by this, I recently made a tool in rust, which could replace canonicalize-jar as it is today.
https://github.com/TomaSajt/stripzip-rs
I was wondering if packaging this could be an acceptable solution. | 18:33:00 |
Pol | In reply to @tomasajt:matrix.org So, some of you might remember that I started working on making more java apps reproducible in https://github.com/NixOS/nixpkgs/issues/278518 and currently one of the roadblocks is that the hook that's used (canonicalize-jars-hook) is a little bit broken.
I had opened a PR which tidied it up a bit (https://github.com/NixOS/nixpkgs/pull/278322/files), but there were several weird issues and special cases that came up. One of the issues was that unzipping and rezipping jars in jars (e.g. dependencies packaged by springboot) caused it to not work anymore. (Probably due to this method not preserving the original compression-type) But I found that just using zip-traversal can mitigate this issue. Example: https://github.com/NixOS/nixpkgs/pull/283305/files I was pretty happy with the solution other than it would require java to run if I wanted to package it standalone.
So inspired by this, I recently made a tool in rust, which could replace canonicalize-jar as it is today. https://github.com/TomaSajt/stripzip-rs I was wondering if packaging this could be an acceptable solution. Wouldn't it be better to do it in Java to avoid mixing different programming languages? | 18:41:13 |
Toma | In reply to @drupol:matrix.org Wouldn't it be better to do it in Java to avoid mixing different programming languages? It could be better, but we couldn't hard-code the jdk version in the hook, as we wouldn't want to use two different jdk versions for building. An option could be to just expect jdk to be in the build environment, another would be to always do canonicalize-jars-hook.override { inherit jdk; } or something similar | 18:45:33 |
Toma | Opened the PR for it https://github.com/NixOS/nixpkgs/pull/286805 | 21:04:06 |
Toma | * In reply to @drupol:matrix.org
Wouldn't it be better to do it in Java to avoid mixing different programming languages?
It could be better, but we wouldn't be able to use a hard-coded the jdk version in the hook, as we wouldn't want to use two different jdk versions for building.
An option could be to just expect jdk to be in the build environment, another would be to always do canonicalize-jars-hook.override { inherit jdk; } or something similar | 22:27:09 |