Robotnix - Public Room Timeline

	Robotnix	244 Members
	Build Android (AOSP) using Nix \| https://github.com/danielfullmer/robotnix	75 Servers

Load older messages

Sender	Message	Time
5 Jun 2024
eyJhb	Do we have like.. three PRs that touch robotnix_common.py, trying to get LFS, and all fixing the same fault in the store path in their own unique way? :D	18:47:22
atemu12	That's what lack of maintenance will do to your project...	20:40:03
jaen	If it works, the currently proposed repo hack is quite an improvement from the previous one Do you mean what is currently in the PR or do you mean I should try the bundle suggestion? If it's the latter, then just to be clear that will only change what ends up in the store (git bundle vs. unpacked git repo), I don't think any of the existing steps to massage the repo can be easily avoided (I can try, but wouldn't hold my breath). Also let me know if you'd prefer for me to squash those commits into existing ones for cleaner history or not. all fixing the same fault in the store path in their own unique way Sorry or my part xD Unfortunately another painful thing I'm trying to get merged is having consistent formatting in the repo with treefmt, so it will probably conflict with all other PRs [*]	22:04:32
jaen	* If it works, the currently proposed repo hack is quite an improvement from the previous one Do you mean what is currently in the PR or do you mean I should try the bundle suggestion? If it's the latter, then just to be clear that will only change what ends up in the store (git bundle vs. unpacked git repo), I don't think any of the existing steps to massage the repo can be easily avoided (I can try, but wouldn't hold my breath). Also let me know if you'd prefer for me to squash those commits into existing ones for cleaner history or not. all fixing the same fault in the store path in their own unique way Sorry or my part xD Unfortunately another painful thing I'm trying to get merged is having consistent formatting in the repo with treefmt, so it will probably conflict with all other PRs [*]	22:04:50
atemu12	I meant the current state of your PR that I reviewed. Bundle wouldn't materially change the approach either IMV	22:08:31
6 Jun 2024
erahhal	Struggling to figure out why adding `git-lfs` to `nativeBuildInputs` here causes infinite recursion: https://github.com/nix-community/robotnix/blob/master/pkgs/fetchgit/default.nix#L66	07:30:17
jaen	Did you change the `fetchLFS` default to `true`? Because it seems `git-lfs` uses `fetchFromGitHub` for it's sources (https://github.com/NixOS/nixpkgs/blob/master/pkgs/applications/version-management/git-lfs/default.nix#L7-L12) which will then use `fetchgit` and since robotnix overlays it in nixpkgs, then you'll end up in the place you linked again and loop trying to evaluate `git-lfs` again. At least that would be my guess. But if you didn't change `fetchLFS` default to `true` I would've imagined laziness to take care of it. But maybe even then it's a good lead, you'd just have to start sprinkling `lib.traceVal` liberally to see where it goes wrong.	08:14:03
eyJhb	In reply to @erahhal:matrix.org Struggling to figure out why adding `git-lfs` to `nativeBuildInputs` here causes infinite recursion: https://github.com/nix-community/robotnix/blob/master/pkgs/fetchgit/default.nix#L66 https://github.com/nix-community/robotnix/pull/228/files couldn't you just use the work I made here for LFS?	11:10:33
eyJhb	Or is there something missing?	11:10:46
eyJhb	It basically just updates the prefetch-script to match the one in nixpkgs, with the added patches, and passes on fetchLFS	11:11:04
erahhal	No, I haven't tried your work, I will try it today!	13:50:48
eyJhb	In reply to @erahhal:matrix.org No, I haven't tried your work, I will try it today! * I make no guarantees however. But I just remember getting such errors as you were getting :p It might need to be modified.	14:02:14
erahhal	Made it to the next step with your change!	20:35:18
erahhal	Now hitting: `ccache: error: Failed to create directory /var/cache/ccache/tmp: Not a directory` Seems relatively straightfoward to address. digging into it now	20:35:39
erahhal	Does the build happen in in a chroot jail or something? I created this folder and it's still complaining	21:29:35
samueldr	Nix builds happen in the Nix sandbox	21:33:40
samueldr	so poking holes through requires some additional configuration	21:33:48
erahhal	Ah ok. seems like it would be best to create that directory in the sandbox rather than poking holes. i'll see if I can do that	21:34:26
erahhal	I'll try adding a `mkdir` command right before running make	21:35:06
erahhal	though	21:35:15
erahhal	there is no point to ccache if it's disappearing after the build right?	21:35:23
erahhal	How did this option work before? Seems that it must have had access to some persistent host folder	21:36:35
erahhal	Disabled ccache and now 1200 files into the build, looking pretty good so far	21:47:06
atemu12	As Samuel said, ccache requires configuration of the daemon	22:03:42
erahhal	oh, I didn't understand that, got it	22:04:35
jaen	In reply to @erahhal:matrix.org How did this option work before? Seems that it must have had access to some persistent host folder It's `extra-sandbox-paths` IIRC, if you want to know what to google for. Incidentally, since we're on the topic of ccache — does it need anything else other than a directory (say, `programs.ccache.enable = true` or something)?	22:04:37
atemu12	Enabling it without persistence is counter-productive as ccache makes cache misses a little slower actually	22:04:55
atemu12	Just the dir	22:05:24
erahhal	When I first tried robotnix a while back with a supported phone and OS version it worked with ccache enabled and no daemon settings	22:05:36
atemu12	Ccache manages that don its own	22:05:41

Show newer messages

Back to Room ListRoom Version: 6