| 10 May 2026 |
 Vladimír Čunát | In reply to @emilazy:matrix.org
right, okay. I don't know exactly why it happens but I know vcunat has mentioned it happening multiple times before. To me the behavior looked like insufficient deduplication of the situations where different hydra jobs built the same derivation, e.g. on different jobsets. It seemed easy to happen. We're helped by the ratio of job count to build times usually, as the duplication only happens concurrently. And for linux we get helped by the builder pool being small and split along the big-parallel axis, so often the duplicate got scheduled to the same machine and deduplicated locally by nix in there. | 06:27:16 |
|  Marcel joined the room. | 12:45:09 |
 John Ericson | I think we should disable fallback paths | 15:13:06 |
| John Ericson | It seems like this feature is not being helpful at all | 15:13:20 |
| John Ericson | And it would be better to fail or delete the outputs first | 15:13:38 |
 K900 | Problem with delete the outputs is they may be in use | 15:14:09 |
| K900 | Problem with fail is Hydra needs to be smart enough to handle this | 15:14:20 |
| K900 | But ideally yes we should fail | 15:14:25 |
 hexa | not sure who restarted ip2unix on nixos/unstable, but it's stuck on a test | 17:13:30 |
| John Ericson | there will be temp roots if they are in use | 17:13:29 |
| hexa | https://hydra.nixos.org/build/326816958/nixlog/5/tail | 17:13:31 |
| hexa | root 374239 0.0 0.0 638608 18368 ? Ssl May06 2:47 nix-daemon --daemon
root 3020110 0.0 0.0 2137748 27724 ? Ssl 14:32 0:01 \_ nix-daemon 3016813
nixbld13 3022134 0.0 0.0 4568 3376 ? Ss 14:32 0:00 | \_ bash -e /nix/store/l622p70vy8k5sh7y5wizi5f2mic6ynpg-source-stdenv.sh /nix/store/shkw4qm9qcw5sc5n1k5jznc83ny02r39-default-builder.sh
nixbld13 3038056 0.0 0.0 50640 40648 ? S 14:32 0:02 | \_ /nix/store/0r6k8xa2kgqyp3r4v2w7yrb80ma2iawm-python3-3.13.12/bin/python3.13 /nix/store/vn448hbsgbh0ahk1bjv2l16f5lhambiq-meson-1.10.2/bin/meson test --no-rebuild --print-errorlogs --max-lines=1000000 --timeout-multiplier=0
nixbld13 3038254 0.0 0.0 37052 32488 ? Ss 14:32 0:00 | \_ /nix/store/0r6k8xa2kgqyp3r4v2w7yrb80ma2iawm-python3-3.13.12/bin/python3.13 /nix/store/wavm4b318kvk7xqy7qfam86i0r8znz1n-python3.13-pytest-9.0.2/bin/.pytest-wrapped -p no:cacheprovider --ip2unix-path=/build/source/build/ip2unix --libip2unix-path=/build/source/build/libip2unix.so --helper-accept-no-peer-addr=/build/source/build/tests/helpers/helper_accept_no_peer_addr --timeout=60 --systemd-support --systemd-sa-path=/nix/store/9rpism89x6lyjcwzzkp6kana25rs03nn-systemd-260.1/bin/systemd-socket-activate
nixbld13 3038432 0.0 0.0 7436 3088 ? S 14:32 0:00 | \_ /build/source/build/tests/helpers/helper_accept_no_peer_addr
| 17:13:53 |
| John Ericson | reminds me I really want to run all test suites in separate derivations | 17:13:54 |
| John Ericson | though also, the ability to pipeline outputs separately with the new RPC thing can ameliorate actually needing to split | 17:14:48 |
| John Ericson | just have a "slow output" for tests | 17:15:09 |
| Vladimír Čunát | It's stuck repeatedly in the past week. Probably always. 10 hours and more, until I cancel. | 18:09:09 |
| Vladimír Čunát | nixos-unstable is blocked anyway by
https://hydra.nixos.org/build/328803284/nixlog/1/tail | 18:10:25 |
| Vladimír Čunát | (and restarting won't help) | 18:10:40 |
| hexa | "bReAkIng cHanGeS fReEz" | 18:10:58 |
| hexa | Redacted or Malformed Event | 18:11:00 |
| hexa | https://github.com/nixos/nixpkgs/commit/1d1e96211ed2a10012dbc8615bcaf028a0853d24 | 18:12:12 |
| hexa | https://github.com/nixos/nixpkgs/commit/a06a6caa6b11f7069ecee892e8ae0a2b8b2f5ff5 revert | 18:12:34 |
| hexa | https://github.com/nixos/nixpkgs/commit/86cd4e7bc84cda9fe8d341a1d9a82574827fb8f3 reapply | 18:12:46 |
| hexa | looking for a replacement for anubis that is less disruptive | 18:33:43 |
| hexa | and for my private hydra iocaine works pretty well to tell bots apart from actual users | 18:34:02 |
| hexa | it does however play out garbage texts to bots, which may or may not be acceptable for h.n.o | 18:34:20 |
| hexa | personally, I don't mind. this is payback. | 18:34:51 |
| hexa | iocaine consumes https://github.com/ai-robots-txt/ai.robots.txt/ which we can update with flake bumps | 18:35:05 |
| hexa | Opinions? | 18:57:08 |
 Lun | In reply to @hexa:lossy.network
looking for a replacement for anubis that is less disruptive disruptive in the sense of interstitial being annoying or false positives or? | 19:40:38 |
