| 30 Jul 2021 |
 David Arnold (blaggacao) | * So I look at the backtraces and place a few traces in nixpkgs to understand what's really going on and then use that modified version of nixpkgs as an input. | 16:41:36 |
| David Arnold (blaggacao) | * ... at least it guarantees that you are eventually able to find the root cause without resorting to rolling the dices. | 16:42:31 |
| David Arnold (blaggacao) | Gytis Ivaskevicius has made a PR to nix to eventually improve that situation. | 16:43:20 |
 @yusdacra:nixos.dev | In reply to @blaggacao:matrix.org
So I look at the backtraces and place a few traces in nixpkgs to understand what's really going on and then use that modified version of nixpkgs as an input. Ooooh | 16:44:34 |
| David Arnold (blaggacao) | https://github.com/NixOS/nix/pull/4914 | 16:44:34 |
| @yusdacra:nixos.dev | That's useful | 16:46:01 |
 @timdeh:matrix.org | I find some common patterns that reliably cause infinite recursion:
- careless use of
rec in overlays
- computation directly in an import
Both of these can be easily solved by using a let binding instead. Whenever I find infinite recursion and I can't trace it back to my local changes, i.e. tons of nixpkgs traces, I just start commenting stuff out until it works, and then work backwards from there 😅
| 16:46:35 |
| David Arnold (blaggacao) | ... buts somtimes chaos monkey dice rolling is quicker 😃🦧 | 16:46:46 |
| @yusdacra:nixos.dev | I would love to add debugging traces to a nix library of mine, but using builtins.trace would print way too much stuff | 16:46:48 |
| David Arnold (blaggacao) | In reply to @timdeh:matrix.org
I find some common patterns that reliably cause infinite recursion:
- careless use of
rec in overlays
- computation directly in an import
Both of these can be easily solved by using a let binding instead. Whenever I find infinite recursion and I can't trace it back to my local changes, i.e. tons of nixpkgs traces, I just start commenting stuff out until it works, and then work backwards from there 😅
I can complement:
- Similar to
rec, it can also happen with overlays and "recless" crossing of prev and final dependency.
- Using flake
self outside of an output value seems to cause that too.
| 16:48:54 |
| @timdeh:matrix.org | Maybe we change to channel to DevOS/Nix Debugging 🤣 | 16:49:43 |
| @timdeh:matrix.org | * Maybe we change the channel name to DevOS/Nix Debugging 🤣 | 16:49:53 |
| @timdeh:matrix.org | What you refer to as DevOS is in fact Nix Debugging/DevOS 🤓 | 16:50:35 |
| David Arnold (blaggacao) | If we are a nixos nursery, we need to definitly empower users as much as possible. 😎 | 16:50:44 |
| David Arnold (blaggacao) | Which makes me think of adding a section to the docs for debugging... | 16:51:33 |
| David Arnold (blaggacao) | * Which makes me think of adding a section to the docs for (teaching) debugging... | 16:51:44 |
| David Arnold (blaggacao) | I'd rather be an enabler than a fixer. | 16:52:32 |
 @gytis-ivaskevicius:matrix.org | In reply to @blaggacao:matrix.org
Gytis Ivaskevicius has made a PR to nix to eventually improve that situation. if you are talking about builtins.debug - dont expect it to get merged :D | 16:53:20 |
| @timdeh:matrix.org | d4hines: hopefully our little Nix debugging venting session is useful for your future endeavors 😛 | 16:57:56 |
 @d4hines:matrix.org | Yeah, this looks like helpful info but I don't have the context to place it yet | 16:58:18 |
| David Arnold (blaggacao) | I think another journyman status standard move is rg 'someThingThatINeedToUnderstand =' <path to nixpkgs> | 16:59:45 |
| David Arnold (blaggacao) | * I think another journeyman status standard move is rg 'someThingThatINeedToUnderstand =' <path to nixpkgs> | 17:00:08 |
| David Arnold (blaggacao) | * I think another journeyman status standard move is rg 'someThingThatINeedToUnderstand =' <path to nixpkgs> (but there might be an improved version of that move?) | 17:00:34 |
