| 1 Jul 2021 |
 fzakaria | --check is kind of frustrating | 16:02:04 |
| fzakaria | because it has to already be in the store | 16:02:10 |
| fzakaria | I think there's opportunity in the newer CLI to make it a better experience here. | 16:02:23 |
| fzakaria | maybe integrate with diffoscope directly. | 16:02:30 |
 Synthetica | Yeah, I want a "no-really-build-this-even-if-you-think-you-already-have-it" flag | 16:45:57 |
| Synthetica | maybe a bit less verbose | 16:46:02 |
 siraben | Boss level: https://github.com/ligurio/unreliablefs | 17:23:24 |
 raboof | nix-build '<nixpkgs>' -A mypackage && nix-build '<nixpkgs>' -A mypackage --check? ;) | 17:31:42 |
| siraben | --check is nice but barely adds variation to the build setup | 17:45:22 |
| siraben | I want to be able to automatically test on different kernels and at different times somehow | 17:45:42 |
 afontain | In reply to @siraben:matrix.org
Boss level: https://github.com/ligurio/unreliablefs that's like disorderfs, but much more evil | 17:48:19 |
| afontain | I'd expect many software not to be able to survive this | 17:48:35 |
| siraben | radiation hardening time | 17:51:09 |
| siraben | level Cthulhu: survive the van Allen belt | 17:52:16 |
| fzakaria | I posted that I like git's model of allowing subcommands so people can create their own.
It lets you have very ergonomic naming commands | 17:54:42 |
| fzakaria | (git searches for any tool on $PATH that is prefixd with git-) | 17:54:50 |
| fzakaria | something like:
nix verify-reproducible <PACKAGE> [--repeat 0] sounds a lot better | 17:55:15 |
| fzakaria | When it does repeat, losing all the previous attempts was frustrating; but i'm being nitpicky | 17:56:18 |
| fzakaria | --check was pretty good | 17:56:23 |
| 2 Jul 2021 |
|  Irenes joined the room. | 09:21:34 |
|  immae changed their display name from immae (he/him) to immae. | 17:46:04 |
| immae changed their profile picture. | 17:46:14 |
| immae changed their profile picture. | 17:47:53 |
 εΉΈη« (πππΎπ/πππΎπ) | In reply to @siraben:matrix.org
It should be easier for people to check reproducibility of their own packages as well Debian's gitlab (salsa) CI has a default reprotest job. something like that would be nice. | 18:38:10 |
|  immae joined the room. | 18:42:42 |
| immae left the room. | 22:02:59 |
| 4 Jul 2021 |
| siraben | Does the reproducibility patches lead to slower binary performance? | 11:23:48 |
 Rick (Mindavi) | Some do, like the python patches that disable optimized builds and disabling gcc profile guided optimization | 11:49:27 |
| Rick (Mindavi) | I think there are some wins to be had with making those reproducible as well, but that's quite involved | 11:50:01 |
| Rick (Mindavi) | And requires intricate knowledge of the code and tooling, so that requires effort upstream | 11:50:54 |
