| 20 Sep 2023 |
 colemickens | yes | 20:12:45 |
 Artturin | https://github.com/nix-community/flake-firefox-nightly/pull/23 | 20:16:30 |
 K900 ⚡️ | So this is a problem | 20:32:32 |
| K900 ⚡️ | Based on the conversation on the issue, Firefox basically hardcodes an offset into the binary now | 20:32:55 |
| K900 ⚡️ | Which means patchelf can't really be expected to work | 20:33:14 |
| K900 ⚡️ | Unless | 20:33:44 |
| K900 ⚡️ | Unless we do even more horrible things | 20:34:18 |
 hexa | in all fairness, they do call it a hack | 20:36:23 |
| K900 ⚡️ | I think I have a horrible idea but I'll have to look at patchelf code to see if it can work | 20:39:26 |
| K900 ⚡️ | Basically, what if instead of moving all the sections around, we just unlist the old dynamic section and add our new one at the end | 20:42:53 |
| K900 ⚡️ | Actually I wonder what the linker will do if you just have more than one .dynamic section | 20:44:03 |
| colemickens | In reply to @artturin:matrix.org
https://github.com/nix-community/flake-firefox-nightly/pull/23 ready? I can merge | 20:48:41 |
| colemickens | oops, just saw its still a draft. no rush | 20:48:59 |
| Artturin | ready | 20:50:23 |
| colemickens | In reply to @artturin:matrix.org
ready done | 20:57:19 |
| K900 ⚡️ | For posterity, last one wins: https://elixir.bootlin.com/glibc/latest/source/elf/rtld.c#L1152 | 20:58:36 |
| K900 ⚡️ | But also this is definitely a glibcism and the manual says to not do that | 20:58:55 |
| K900 ⚡️ | Also the person on the issue actually read patchelf source code and it turns out we already basically do the bad cursed thing | 21:00:39 |
| K900 ⚡️ | Except we also clobber the old data explicitly | 21:00:48 |
| K900 ⚡️ | Which I guess we can just not do | 21:00:53 |
| colemickens | Artturin: don't want to rush you, but trying to be cognizant of the nix-community "namespace" (ugh) minutes available: https://github.com/nix-community/flake-firefox-nightly/pull/24 let me know what you think | 21:46:06 |
| Artturin | roberth approved https://github.com/NixOS/nixpkgs/pull/192593 | 21:46:52 |
| Artturin | merged it | 21:47:50 |
| colemickens | oh, that's small enough I feel comfort looking and ... | 21:47:54 |
| colemickens | and ope its gone | 21:47:56 |
| Artturin | we can limit the test time because the building shouldn't take longer than 10 min right | 21:48:51 |
| colemickens | yeah, that's what the PR says :D | 21:49:02 |
| Artturin | and then limit the nixos test to some seconds or minutes | 21:49:03 |
| colemickens | for now I just big-hammered-it and limited the whole workflow run because it was easy | 21:49:26 |
| colemickens | regarding your merged nixos-test change: on the one hand we/you could add master as an input to use it, but also, I'm not sure how long it's going to take to fix this properly anyway | 21:49:37 |
