| 5 Nov 2025 |
 522 it/its ⛯ΘΔ |
Lix is pronounced like licks — i.e. what you might do if someone handed you a delicious ice cream cone; which may or may not be stacked up with lambdas.
| 15:51:15 |
 Cyborus | ooh should've looked there. thanks! | 15:51:37 |
 Unawarewolf | OOOOooooooh that explains the icecream logo!!!!!!!! | 18:55:48 |
| Unawarewolf | that just clicked now!!!! | 18:55:55 |
| Unawarewolf | (having a username checks out moment I think) | 18:56:51 |
| 6 Nov 2025 |
|  nhnn changed their profile picture. | 15:12:33 |
| nhnn changed their profile picture. | 15:18:29 |
| 7 Nov 2025 |
 MangoIV | there seems to be issues with downgrading flake locks resulting into flake locks which follow inputs that are newer than themselves. | 11:34:21 |
| MangoIV | i.e. flakelock with input i1 is has been updated at 10, now it is downgraded to 1, but its inputs, i.e. i1.i2 i1.i3 still are at 7,8, etc. deleting the flake.lock and recreating it fixes it. | 11:35:21 |
| MangoIV | this obviously leads to really problematic incosistencies. | 11:35:41 |
| MangoIV | and it's of course a direct consequence of flake.locks being flat | 11:36:02 |
| MangoIV | It really puzzled me because it made the upstream flake not reproduce anymore. | 11:38:38 |
 kuruczgy | What's the standard procedure when nix store gc breaks with a full disk (is it intended to be broken in this case?
error (ignored): error: aborting transaction: SQL logic error, cannot rollback - no transaction is active (in '/nix/var/nix/db/db.sqlite')
0 store paths deleted, 0.00 MiB freed
error: committing transaction: database or disk is full, database or disk is full (in '/nix/var/nix/db/db.sqlite')
Also after running nix-store --gc a second time my nix daemon crashed, I assume this is not intended (but it's Lix 2.91.3, so maybe this has been fixed since)
systemd-coredump[1048338]: [🡕] Process 1048334 (nix-daemon) of user 0 dumped core.
| 17:37:34 |
| kuruczgy | * What's the standard procedure when nix store gc breaks with a full disk (is it intended to be broken in this case?)
error (ignored): error: aborting transaction: SQL logic error, cannot rollback - no transaction is active (in '/nix/var/nix/db/db.sqlite')
0 store paths deleted, 0.00 MiB freed
error: committing transaction: database or disk is full, database or disk is full (in '/nix/var/nix/db/db.sqlite')
Also after running nix-store --gc a second time my nix daemon crashed, I assume this is not intended (but it's Lix 2.91.3, so maybe this has been fixed since)
systemd-coredump[1048338]: [🡕] Process 1048334 (nix-daemon) of user 0 dumped core.
| 17:39:10 |
 hexa | are you on a cow filesystem? | 18:25:27 |
|  @awwpotato:envs.net changed their display name from awwpotato (she/her) to -> @da157:catgirl.cloud. | 18:30:23 |
| kuruczgy | Yes, btrfs | 18:33:31 |
|  0xda157 joined the room. | 18:37:06 |
| @awwpotato:envs.net left the room. | 18:41:54 |
 Sofie 🏳️⚧️ (she/her) | When trying to overlay the monorepoSrc on llvmPackages.override I get
> mismatch in the minor version! we have version 21.1.2 and expected the minor version to be '1'; the source has '0' instead
| 21:39:57 |
| Sofie 🏳️⚧️ (she/her) | ah | 21:49:43 |
| Sofie 🏳️⚧️ (she/her) | it was just | 21:49:45 |
| Sofie 🏳️⚧️ (she/her) | version? | 21:49:46 |
| 8 Nov 2025 |
 aloisw | The "standard" procedure is deleting /nix/var/nix/db/reserved, but given that it's a relatively small file of zeroes it may not be enough. | 06:57:15 |
 Emma [it/its] | TIL that file exists | 08:12:45 |
| Emma [it/its] | personally i'd go with rm -rfI /tmp/* | 08:13:34 |
| aloisw | If your /tmp is on the same partition as /nix, that should work too of course. But on many systems it isn't. | 08:34:20 |
| aloisw | I wonder if we should just make the reserved file larger (maybe ~200 MiB, which is a bit above the 160 MiB required for the WAL) and fill it with random data. | 08:35:08 |
 tc424 (Steve D) | Feels like something that could usefully be configurable, given the behaviour of different filesystems etc is likely to have an impact | 09:49:04 |
| aloisw | Hm, what are you thinking about in concrete terms? I only know the behaviours "deleting the file frees the space it occupies" (which is the full size with random data), "deleting the file is useless" (don't hardlink or reflink that file then) or "deleting the file fails with ENOSPC`. In none of these cases a configurable size helps, except for maybe turning off the reserve file entirely if you expect it to be useless anyway. | 09:52:46 |
