| 8 Feb 2022 |
 tpw_rules | the nlink values in the initrd hydra and i both built | 23:50:03 |
| tpw_rules | (which are the same) | 23:50:07 |
 atemu12 | But why is /etc 3? | 23:51:20 |
| tpw_rules | i don't know. that's what the file says | 23:51:34 |
| atemu12 | All directories should be 2 | 23:51:39 |
| tpw_rules | this says at least 2: https://www.systutorials.com/docs/linux/man/5-cpio/ | 23:51:56 |
| tpw_rules | which means that the file where they are all 1 seems invalid to me | 23:52:04 |
| atemu12 | macOS' HFS has directory hardlinks, so I'd assume support for >1 is for that | 23:52:42 |
| atemu12 | * macOS' HFS has directory hardlinks, so I'd assume support for >2 is for that | 23:53:26 |
| tpw_rules | i mean i hope this is not using HFS+ on linux | 23:53:38 |
| atemu12 | Don't think it supports dir hardlinks on Linux | 23:54:21 |
| atemu12 | Could you find out how nlink is determined? | 23:55:14 |
| tpw_rules | i briefly glanced at the cpio source code and it seems it is set equal to 1 when performing crc calculations | 23:55:50 |
| atemu12 | Maybe try creating new cpios out of the files that are supposed to be packed up and selectively delete | 23:56:21 |
| tpw_rules | not sure what you mean | 23:57:10 |
| atemu12 | The idea is to find a minimal reproducer for creating directories with different nlink counts | 23:58:04 |
| atemu12 | Something must be causing that phenomenon | 23:58:16 |
| 9 Feb 2022 |
 @rnhmjoj:maxwell.ydns.eu | stupid question: does the fact that uname -v include a timestamp mean the kernel is not reproducibile? | 08:35:41 |
 raboof | In reply to @rnhmjoj:maxwell.ydns.eu
stupid question: does the fact that uname -v include a timestamp mean the kernel is not reproducibile? AFAIK the kernel is reproducible, so that suggests it's ok - presumably it takes the date from a 'stable' source like commit or filesystem timestamps instead of from the clock | 08:51:07 |
 j-k | https://reproducible-builds.org/docs/source-date-epoch/
git log -1 --pretty=%ct
or if you want something more human readable
git log --date=format:'%Y-%m-%dT%H:%M:%SZ' -1 --pretty=%ad
| 08:56:30 |
| j-k | not sure what the kernel builds actually use for date tho | 08:56:46 |
 tomberek | In reply to @tpw_rules:matrix.org
i mean i hope this is not using HFS+ on linux It’s on BTRFS. | 12:17:24 |
| tpw_rules | oh, that's the problem: https://linux-btrfs.vger.kernel.narkive.com/oAoDX89D/btrfs-st-nlink-for-directories#post3 | 17:41:19 |
| tpw_rules | btrfs always answers 1 for st_nlink, allegedly (have not yet tested personally) | 17:41:43 |
| tpw_rules | i feel like this is cpio's problem, because the file format states that st_nlink is always at least 2 | 17:52:23 |
| tpw_rules | should we use tar in cpio mode? | 17:57:24 |
| tomberek | hrm... i was going to shift my builder to ZFS. I guess it's a good thing to have diversity of testing for reproducibility | 18:14:06 |
| atemu12 | In reply to @tpw_rules:matrix.org
should we use tar in cpio mode? If that fixes the issue and outputs a valid cpio, sure | 18:21:26 |
| 12 Feb 2022 |
|  @marius851000:newsmatrix.pmdcollab.org left the room. | 15:06:16 |
| 14 Feb 2022 |
|  @test:boba.best changed their display name from Tseb to Tseb (Old). | 12:52:27 |
