| 1 Dec 2021 |
 baloo | tomberek: your fix for nix only fixes the core count, correct? | 13:51:43 |
| baloo | https://r13y.com/diff/6879c6d957e3fe6481311e064e2f4ded2a6e8aa47dfa243ff774f3976e966514-9f9d71c6fa6e21997dc038ffad503f9c2c94695e43f89844776366bfc1fb0beb.html | 13:51:45 |
| baloo | this does not fix the extraneous x86_64-v3-linux correct? | 13:52:24 |
| baloo | (not actually sure where that comes from) | 13:52:37 |
| baloo | this has to do with cpu support | 13:58:52 |
| baloo | https://github.com/NixOS/nix/blob/master/src/libutil/compute-levels.cc#L58 | 13:59:35 |
 j-k | should we even be recording the core count of the system that built the docs within the docs in the first place? | 14:01:00 |
| j-k | 
Download image.png | 14:01:03 |
| j-k | (another diffoscope with x86_64-v3-linux https://reproducible.archlinux.org/api/v0/builds/193410/diffoscope ) | 14:01:53 |
| baloo | that's not the only problem | 14:17:13 |
| baloo | well, tomberek fixed the core one, but I count 3 problematic ones | 14:17:37 |
| baloo | cores, system features and platform | 14:18:06 |
| baloo | all 3 are dependent on builder's environment | 14:18:18 |
| baloo | I have a patch that selects the configuration settings where to filter out default values | 15:05:09 |
| baloo | (and I'm removing all 3 of those options) | 15:05:19 |
| baloo | there might be a better approach but I don't know | 15:05:48 |
| baloo | https://github.com/NixOS/nix/pull/5702/files?w=1 | 15:13:02 |
 tomberek | looking... | 16:03:03 |
| 4 Dec 2021 |
|  immae left the room. | 10:59:16 |
| 5 Dec 2021 |
|  svhe joined the room. | 03:47:04 |
| 7 Dec 2021 |
|  cypher joined the room. | 22:54:24 |
| 8 Dec 2021 |
|  arcz joined the room. | 16:22:53 |
| arcz | hey, I'm wondering if r13y is really verifying that NixOS image is reproducible. It seems that only the direct dependencies are rebuilt, not the whole tree down to some root bootstrap package, is that right? Is it possible to rebuild the whole tree, is it even worth it? | 16:33:55 |
 raboof | In reply to @arcz:matrix.org
hey, I'm wondering if r13y is really verifying that NixOS image is reproducible. It seems that only the direct dependencies are rebuilt, not the whole tree down to some root bootstrap package, is that right? Is it possible to rebuild the whole tree, is it even worth it? https://r13y.com/ doesn't check everything. One thing you could do is take an 'old' image, install Nix, download nixpkgs, disable cachix, and build the image. I have found a couple of issues like that in the past (e.g. around the way we used mksquashfs IIRC), but I'm not aware of any remaining. | 16:58:04 |
| arcz | raboof: not sure if I understood you correctly, so you tried to build everything from source and it reproduced, but it's not automated in r13y? | 17:09:50 |
| 9 Dec 2021 |
| raboof | In reply to @arcz:matrix.org
raboof: not sure if I understood you correctly, so you tried to build everything from source and it reproduced, but it's not automated in r13y? I don't think I ever got around to actually successfully reproducing end-to-end, but AFAIK we did fix all the problems I found when I tried it the last time. Indeed r13y does not automate everything (but quite a lot). | 08:18:35 |
| 10 Dec 2021 |
 siraben | In reply to @raboof:matrix.org
https://r13y.com/ doesn't check everything. One thing you could do is take an 'old' image, install Nix, download nixpkgs, disable cachix, and build the image. I have found a couple of issues like that in the past (e.g. around the way we used mksquashfs IIRC), but I'm not aware of any remaining. A similar situation I ran into was when I was testing out CA, several URLs were broken because our caching kept it longer than upstream did | 00:56:10 |
| 12 Dec 2021 |
|  srid changed their display name from Moved to @srid:one.ems.host to Sridhar Ratnakumar. | 23:54:09 |
| srid set a profile picture. | 23:54:16 |
| 14 Dec 2021 |
| baloo | https://github.com/NixOS/nix/pull/5702 merged \o/ now we just need the backport and a release | 11:15:47 |
