| 14 Mar 2026 |
lassulus | so I want to optimize that | 18:38:02 |
lassulus | but everytime I want to, I can never reproduce it | 18:38:10 |
matthewcroughan | Okay I'll make a thing | 18:38:32 |
matthewcroughan | Maybe it's only on ZFS because of ARC | 18:42:47 |
lassulus | hmm, maybe? | 18:43:05 |
lassulus | but we could disable that for the image builder maybe? | 18:43:19 |
matthewcroughan | possibly as a kernel param.. yeah | 18:43:28 |
lassulus | but not sure it's ARC, why would it be? you can copy more files than you have memory on ZFS usually? | 18:44:36 |
matthewcroughan | VMs | 18:44:48 |
matthewcroughan | weird stuff | 18:44:52 |
matthewcroughan | but I can't reproduce it with ext4 atm | 18:45:00 |
matthewcroughan | Hah, okay so I just changed everything to zfs in the make-disk-image test | 18:48:12 |
matthewcroughan | immediate kernel panics | 18:48:15 |
| amadaluzia[uorg] changed their display name from amadaluzia to amadaluzia[uorg]. | 18:50:34 |
matthewcroughan | I think it's probably ARC.. | 18:53:30 |
matthewcroughan | I have a toy big closure, and it didn't happen with only 1024 of memory | 18:53:51 |
matthewcroughan | but I bet if I get the size closer to 10G it will happen | 18:54:03 |
matthewcroughan | I can imagine what's happening tbh | 18:56:02 |
matthewcroughan | xargs -P 8 -I '{}' cp --recursive '{}' /mnt/nix/store | 18:56:15 |
matthewcroughan | maybe if you're lucky, your {} gets filled with store paths that are small enough to succee | 18:56:28 |
matthewcroughan | * maybe if you're lucky, your {} gets filled with store paths that are small enough to succeed | 18:56:30 |
matthewcroughan | maybe if you're unlucky, you get 8 paths in {} that are too big to copy in one go for your current invocation | 18:56:51 |
matthewcroughan | so I can't reproduce it because my toplevel is very different to what I can create in the test | 18:57:12 |
matthewcroughan | hehe, I know | 19:00:50 |
matthewcroughan | diffoscope | 19:00:51 |
matthewcroughan | ah fuck it all | 19:02:56 |