| 23 Aug 2021 |
baloo | but this is a different issue | 18:50:07 |
baloo | https://github.com/NixOS/nixpkgs/pull/135464 | 18:56:53 |
baloo | andi-: I can't reproduce the issue anymore with this applied | 18:59:24 |
baloo | is this the same you were hitting? | 18:59:37 |
andi- | I wasn't hitting any of those errors. I was just confirming that hydra.nixos.org had it | 22:57:29 |
andi- | I personally never had that issue | 22:57:33 |
baloo | running with resolver 127.0.0.1? | 22:59:02 |
baloo | Besides the ones I fixed in the last PR, I can't find any other issue. | 23:00:45 |
andi- | I'm running a recursive resolver on almost all of my machines so that might be a difference | 23:02:55 |
baloo | yeah same here | 23:03:04 |
| 24 Aug 2021 |
baloo | got my mirrors PR merged, but r13y rebuilt before that. | 16:54:20 |
baloo | have to wait 24h :( | 16:54:36 |
[0x4A6F] | r13y.com mentioned in DebConf21 talk: https://debconf21.debconf.org/talks/86-reproducible-buster-bullseye-bookworm-where-we-come-from-and-where-we-are-going/ | 21:11:54 |
| 25 Aug 2021 |
j-k | I guess the recordings will show up in a month or two? | 10:06:48 |
[0x4A6F] | I think they reused https://media.ccc.de/v/470-reproducible-buster-bullseye-bookworm-where-we-come-from-and-where-we-are-going#t=432 | 11:07:06 |
Foxboron | They did,
https://twitter.com/ReproBuilds/status/1430449940736663554 | 12:52:45 |
j-k | thanks both | 14:28:19 |
baloo | got to be the nscd issue then. Still not fixed with the latest build | 15:11:08 |
baloo | https://github.com/NixOS/nix/issues/5089#issuecomment-905193921 | 15:11:18 |
baloo | if anyone has an idea how to reproduce this issue | 15:11:51 |
| 28 Aug 2021 |
| @grahamc:nixos.org invited NixOS Moderation Bot. | 02:32:38 |
| NixOS Moderation Bot joined the room. | 02:32:39 |
| NixOS Moderation Bot banned matthewcroughan (<no reason supplied>). | 02:35:47 |
| hexagonk joined the room. | 12:19:23 |
| 30 Aug 2021 |
baloo | nrdxp: https://blog.superbaloo.net/posts/emmett-vdso/ got some more details here | 02:06:26 |
cleverca22 | while building some rpi disk images, i discovered ext4 images generated by nixos/lib/make-ext4-fs.nix not being reproducible
using debugfs on the images, and running show_super_stats, i found the following difference:
Directory Hash Seed: 153a13f8-fd63-4dec-9295-e702caea9421
Directory Hash Seed: 7ab5f8a0-ae47-4beb-a549-b32270a9ee1d
2021-08-30 11:15:54 < slyfox> e2fsprogs' tests force UUID to a constant to get stable output: tests/t_change_uuid_mcsum/script:$MKE2FS -O metadata_csum,^metadata_csum_seed -U 6b33f586-a183-4383-921d-30da3fef2e5c -F $TMPFILE > /dev/null 2>&1
but when applying that to nixpkgs:
diff --git a/nixos/lib/make-ext4-fs.nix b/nixos/lib/make-ext4-fs.nix
index 416beeb32f2..179ab5998e3 100644
--- a/nixos/lib/make-ext4-fs.nix
+++ b/nixos/lib/make-ext4-fs.nix
@@ -64,7 +64,7 @@ pkgs.stdenv.mkDerivation {
truncate -s $bytes $img
- faketime -f "1970-01-01 00:00:01" fakeroot mkfs.ext4 -L ${volumeLabel} -U ${uuid} -d ./rootImage $img
+ faketime -f "1970-01-01 00:00:01" fakeroot mkfs.ext4 -O metadata_csum,^metadata_csum_seed -L ${volumeLabel} -U ${uuid} -d ./rootImage $img
export EXT2FS_NO_MTAB_OK=yes
# I have ended up with corrupted images sometimes, I suspect that happens when the build machine's disk gets full during the build.
the build fails with:
ext4-fs.img> Creating journal (8192 blocks): done
ext4-fs.img> Copying files into the device: ext2fs_mkdir: Directory block checksum does not match directory block while creating directory "lib"
ext4-fs.img> __populate_fs: Directory block checksum does not match directory block while making dir "lib"
ext4-fs.img> mkfs.ext4: Directory block checksum does not match directory block while populating file system
ext4-fs.img> builder for '/nix/store/i6y239h646p67l6pa9qyzw2v50cyf432-ext4-fs.img.drv' failed with exit code 1
| 14:37:56 |
baloo | the alternative could be to use tune2fs -O '^metadata_csum_seed' to disable it afterwards | 16:45:06 |
| 1 Sep 2021 |
| @orpheus1:matrix.org joined the room. | 18:09:40 |
| @orpheus1:matrix.org left the room. | 20:10:22 |
| 3 Sep 2021 |
| Zhaofeng Li joined the room. | 05:54:31 |