| 16 Jan 2022 |
|  kraem set a profile picture. | 09:54:32 |
|  @mrckndt:matrix.org joined the room. | 19:48:03 |
| 19 Jan 2022 |
|  jake :: Stateless -> Farm joined the room. | 14:39:37 |
| 20 Jan 2022 |
|  andi- left the room. | 00:14:26 |
| 21 Jan 2022 |
| @mrckndt:matrix.org left the room. | 17:12:00 |
 tpw_rules | is there anything remaining to merge https://github.com/NixOS/nixpkgs/pull/153409 ? | 17:31:57 |
| 24 Jan 2022 |
| tpw_rules | has anybody investigated the reproducibility issue with the initrd? | 04:49:54 |
 baloo | which issue? | 23:53:17 |
| 25 Jan 2022 |
 tomberek | (i suspect referring to the diff at: https://tomberek.info/r13y.com/ , right tpw_rules ?) Graham's hasn't updated in a while. | 04:18:00 |
| tpw_rules | yes. i was seeing something like that when cross compiling a bunch for aarch64 | 04:28:05 |
|  @rnhmjoj:maxwell.ydns.eu joined the room. | 13:16:47 |
| @rnhmjoj:maxwell.ydns.eu | why hasn't the fix for unreproducible man pages been backported to nix 2.4? nix being the only unreproducible package in the minimal image is pretty bad | 13:21:37 |
 toonn | What's preventing that Nix from being updated wholesale. Considering 2.6.0 has just been released? | 13:24:29 |
| @rnhmjoj:maxwell.ydns.eu | nixos-unstable nix seems to be already at 2.5, maybe the image contains nix_2_4 explicitely? | 13:32:11 |
| tomberek | There was a bit of concern about 2.3->2.4 during 21.11 release. So that and some behavior changes/bugs with the newer releases is holding things back a bit. | 16:39:33 |
| tomberek | I'd like to fix the bugs and get everyone up to date and avoid fragmentation. My goal for 22.05 is to have a full new Nix by default. Maybe 3.0 if it is ready? | 16:40:48 |
| 27 Jan 2022 |
| tpw_rules | looks like the initrd is not reproducible because cpio packs hard links | 02:58:18 |
| tpw_rules | with this particular nonstandard configuration, it's 144325 blocks, then i do nix-store --optimise, then it's 144122 blocks and no longer matches. not sure if there's some flag to fix that. i will try to investigate later | 02:58:57 |
| tpw_rules | * looks like the initrd is not reproducible because cpio seems to pack hard links | 03:03:26 |
| tpw_rules | does nar know about hard links? | 03:08:17 |
| tpw_rules | i wouldn't think so? | 03:08:22 |
| tpw_rules | looks like it may be as simple as s/prd/prP/ here: https://github.com/NixOS/nixpkgs/blob/master/pkgs/build-support/kernel/make-initrd.sh#L34 | 03:40:42 |
| tpw_rules | if i am reading the cp manpage right | 03:40:45 |
| tomberek | Looking at Figure 5.2 from thesis the supported types are regular(executable and normal), Symlink, Directory. I believe this is still true: https://github.com/NixOS/nix/blob/master/src/libutil/archive.cc#L202-L219 (btw: considering how long ago it was, the NAR format is still on the original version 1 ! A good testament to keeping it simple) | 03:41:31 |
| tpw_rules | i'm just wondering if it's a problem if we potentially remove a hardlink a derivation has made itself. i am not sure they are distinguishable from ones made by --optimise | 03:42:32 |
| tpw_rules | and i guess not, if nar does not convey that information | 03:43:00 |
| tpw_rules | i will do some testing and submit a PR tomorrow night | 03:43:19 |
| tomberek | Sounds reasonable | 03:47:20 |
 afontain | So the point is to have two copies instead of one, until it gets optimized? | 10:01:22 |
 atemu12 | It's to always have two copies in the CPIO, no matter the stateful optimisation. | 12:44:16 |
