| 28 Mar 2024 |
 eyJhb | Might have to rethink it. Not sure I can actually do that part myself, which is annoying. My "NAS" doesn't have much storage | 06:43:55 |
 atemu12 | :/ | 06:44:07 |
| atemu12 | btw do note that repo and device dirs are separate | 06:44:45 |
| eyJhb | In reply to @atemu12:matrix.org
:/ Nix ASOP mirror when? :D | 06:45:25 |
| 29 Mar 2024 |
|  SebTM joined the room. | 03:39:13 |
|  CyberShadow joined the room. | 09:15:49 |
| eyJhb |
I suspect that to be the case too. Considering that nixpkgs scripts are executed without errexit or any other bash strict mode flags, I am not at all surprised about such bugs.
I was quite sure that it was required for all phases to exit with code 0. Isn't that the case?
| 09:16:52 |
| CyberShadow | In reply to @eyjhb:eyjhb.dk
I suspect that to be the case too. Considering that nixpkgs scripts are executed without errexit or any other bash strict mode flags, I am not at all surprised about such bugs.
I was quite sure that it was required for all phases to exit with code 0. Isn't that the case?
Hi!
Yes, but without errexit, you can have a failing command followed by a successful command, and the failing command won't stop the script, and the successful command will still run and cause the script to exit with success (status of last command)
| 09:18:50 |
| eyJhb | Ahhhhhhhhhhhhhhhhhhhh, yeah that makes sense | 09:19:15 |
| eyJhb | I wonder how much of Nix would break with errexit. Most likely, a lot. | 09:19:30 |
| CyberShadow | Yes, same is true for most shell scripts written without it in mind in general | 09:20:35 |
| CyberShadow | There are gotchas like grep returning failure if it doesn't find at least one match, so a lot of grep invocations need to be pacified with || true OSLT | 09:21:17 |
| eyJhb | Yeah true. But I think we have many hidden failures as well. Bash might not have been the best way to package things :p | 09:22:29 |
| CyberShadow | Oh, another annoying thing is that the Android build scripts (that you have to source into your shell) don't work with errexit either. | 09:24:18 |
| eyJhb | But would it make sense to have a separate PR that bumped nixpkgs to 23.11, and then when that is merged, rebase the changes you've made? - Just to keep it simple (maybe). | 09:24:47 |
| eyJhb | In reply to @cybershadow:cy.md
Oh, another annoying thing is that the Android build scripts (that you have to source into your shell) don't work with errexit either. I think Android build tools are made of duct tape + hope and dreams | 09:25:00 |
| CyberShadow | Redacted or Malformed Event | 09:27:51 |
| CyberShadow | Redacted or Malformed Event | 09:28:03 |
| CyberShadow | Redacted or Malformed Event | 09:28:28 |
| CyberShadow | Redacted or Malformed Event | 09:28:58 |
| CyberShadow | Redacted or Malformed Event | 09:29:56 |
| eyJhb | Why the deletion of messages, did you discover something? :D | 09:38:30 |
| CyberShadow | Yes, discovered I was wrong, haha | 09:38:42 |
| eyJhb | AT some point I'll try to bump nixpkgs to 23.11, and see if I can make it build for FP4 | 09:49:10 |
| eyJhb | Just to have some sort of "benchmark" | 09:49:18 |
| eyJhb | Or... A pass | 09:50:09 |
| CyberShadow | I think this is it: https://github.com/nix-community/robotnix/pull/230 | 09:56:42 |
| CyberShadow | It's confusing because nixpkgs' setup.sh (which we source) starts with set -eu. But it restores the flags at the bottom. | 09:58:01 |
| eyJhb | Niiiiiiiiiiiiice, good job tracking that down!! | 10:17:14 |
| eyJhb | I can test it the OOM way again if you need that :) | 10:17:27 |
