Conversation

Notices

1. Embed this notice
   anna (navi@social.vlhl.dev)'s status on Wednesday, 18-Jun-2025 06:03:25 JST anna
   @sertonix uh
   
   i messed up some logic and used `&&` instead of `||`
   
   PR of shame: https://github.com/OpenRC/openrc/pull/884
   • Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 18-Jun-2025 06:04:51 JST Haelwenn /элвэн/ :triskell:

     in reply to
     @navi @sertonix Haha happens to the best of us.
     
     Although I tend to write bits of tests (and sometimes look at which branches are covered) to make sure I didn't get the logic wrong.
   • Embed this notice
     anna (navi@social.vlhl.dev)'s status on Wednesday, 18-Jun-2025 06:07:30 JST anna

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @lanodan @sertonix
     
     openrc at the moment has *zero* runtime tests, all CI does is check "does this build", and then run some crappy linters
     
     i'm actively working to make it better¹, but it's a lot since in multiple places openrc assumes being installed to /
     
     1: https://github.com/OpenRC/openrc/pull/837
     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 18-Jun-2025 06:08:26 JST Haelwenn /элвэн/ :triskell:

     in reply to
     @navi @sertonix Couldn't OpenRC actually get installed when it's in a container?
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 18-Jun-2025 06:17:12 JST Haelwenn /элвэн/ :triskell:

     in reply to
     @navi @sertonix I think for docker I always end up using it in anger.
     Much less frustration with lxc but haven't used it since lxc-templates got thrown away.
     Meanwhile bwrap is quite comfy, although I should also test unshare(1).
   • Embed this notice
     anna (navi@social.vlhl.dev)'s status on Wednesday, 18-Jun-2025 06:17:13 JST anna

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @lanodan @sertonix it could, and initially i proposed testing with podman/docker
     
     but it ends up really messy, and i suck at containers -- besides i think it's better practice to make the code more flexible wrt install locations anyway
     
     for chrooting, i couldn't find a decent way to build a chroot automatically and reliably, but i probably didn't look far enough
     
     i'll revisit containers and/or chroots when i get into integration tests tho, i still want unit tests to be able to simply run inside $build_dir
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 18-Jun-2025 06:24:55 JST Haelwenn /элвэн/ :triskell:

     in reply to

     @navi @sertonix bind mount of the original system + openrc install prefix with an overlay mount instead of pulling another rootfs?

     That said I think I'd also be fine with like creating an Alpine rootfs (as apk init is great for this and doesn't requires a lot of space unlike say gentoo stage3), better than no tests.

   • Embed this notice
     anna (navi@social.vlhl.dev)'s status on Wednesday, 18-Jun-2025 06:24:56 JST anna

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @lanodan @sertonix unshare would be my goto since it's less of a container engine and more of a simple namespace tool
     
     but that brings up the same issue as chrooting, which is, "how do i get a rootfs image in a way that is agnostic and automatic?"
     
     my goal with the tests is to make `meson test` work by itself without any prior user setup
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 18-Jun-2025 06:52:05 JST Haelwenn /элвэн/ :triskell:

     in reply to
     @navi @sertonix ro bind of / is something I do all the time in my sandbox scripts to like restrict writes to specific directories, like https://hacktivis.me/git/dotfiles/file/.local/bin/pkgdev.html
   • Embed this notice
     anna (navi@social.vlhl.dev)'s status on Wednesday, 18-Jun-2025 06:52:07 JST anna

     in reply to

     • Haelwenn /элвэн/ :triskell:
     @lanodan @sertonix
     
     oh... i didn't think of doing a ro bind mount of / -- that's kinda really smart, i'll try it
     
     only issue i can think of is a bit convoluted, but, it's if openrc is installed on /, and we remove a file from the version we want to test, but that file would be still there in the base image
     
     i think an alpine minimal rootfs may be simpler