| 7 Nov 2024 |
 Tomodachi94 (they/them) | In reply to@vengmark2:matrix.org
So if I install an NPM package globally, that won't be available in a project Nix shell? That depends on if the shell is --pure or not. It also depends on the shell configuration (doesn't NPM install executables into a separate directory?) | 19:32:22 |
| Tomodachi94 (they/them) | That depends on if the shell is --pure or not. It also depends on the user's shell configuration if the shell is impure (doesn't NPM install executables into a separate directory?) | 19:51:38 |
 @vengmark2:matrix.org | In reply to @tomodachi94:matrix.org
That depends on if the shell is --pure or not. It also depends on the user's shell configuration if the shell is impure (doesn't NPM install executables into a separate directory?) Ah, that clarifies it, and shows a real use case: Doing pkgs.python310.withPackages (ps: [ps.numpy]) in a Nix shell in a Python project ensures that I get a specific Python interpreter version with a specific package set, with no need for having to develop in a pure shell. Being able to do the same thing with Node.js would be great: pkgs.nodejs_X.withPackages (ps: [ps.foo]) would result in having a specific Node.js interpreter version and a specific package set, with no worries about interference from globally installed packages. | 19:53:19 |
| Tomodachi94 (they/them) | Iirc impure shells should prepend to the PATH, so explicitly specifying the packages should work if they're packaged | 23:48:41 |
| Tomodachi94 (they/them) | * Iirc impure shells should prepend to the PATH, so explicitly specifying the packages should work if they're packaged in Nix | 23:56:16 |
| 8 Nov 2024 |
|  moved to @amadaluzia:tchncs.de changed their display name from (artur 'manuel) to @amadaluzia:tchncs.de. | 14:28:48 |
| moved to @amadaluzia:tchncs.de changed their display name from @amadaluzia:tchncs.de to moved to @amadaluzia:tchncs.de. | 14:30:07 |
| 10 Nov 2024 |
|  loudgolem joined the room. | 04:44:48 |
| 16 Nov 2024 |
|  🐰 xiaoxiangmoe joined the room. | 06:22:03 |
| 18 Nov 2024 |
|  aktaboot changed their profile picture. | 08:03:12 |
 Tumble | Could I make a node js service call the nixos-rebuild command? | 12:14:00 |
 Cobalt | You can? It's just an executable (script). You might need elevated permission though. | 13:03:30 |
| Cobalt | * You can. It's just an executable (script). You might need elevated permission though. | 13:04:19 |
 @joepie91:pixie.town | Tumble: it's a bash script, so to call it 'safely' (insofar it can be called safely) you'd call it as childProcess.spawn("bash", [ "/path/to/nixos-rebuild", "switch" ]) or something along those lines | 13:16:06 |
| @joepie91:pixie.town | taking into account that something that controls what gets rebuilt has essentially root access | 13:16:19 |
| 28 Nov 2024 |
|  @drupol:matrix.org joined the room. | 13:47:24 |
| @drupol:matrix.org | Hey everybody, I have a friend running into this issue on his linux machine: https://gist.github.com/drupol/58bab21103636f622bb3a623850ca0b3 | 13:47:46 |
| @drupol:matrix.org | Is there something I can try to help him? | 13:50:19 |
 ghpzin | In reply to@drupol:matrix.org
Is there something I can try to help him? If you still need help, it just means they OOMd while running it:
3: 0xe7d2a4 v8::internal::V8::FatalProcessOutOfMemory
It should show up better on the in logs before that with something like:
FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory | 15:44:32 |
| ghpzin | If you still need help, it just means they OOMd while running it:
3: 0xe7d2a4 v8::internal::V8::FatalProcessOutOfMemory
It should show up better in logs before that with something like:
FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory | 15:44:51 |
| ghpzin | <del>If you still need help, it just means they OOMd while running it:
3: 0xe7d2a4 v8::internal::V8::FatalProcessOutOfMemory
It should show up better in logs before that with something like:
FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory</del>
Run it myself and node needs more than default memory for whatever open-webui is doing.
With env.NODE_OPTIONS = "--max-old-space-size=8192"; it seems to build.
| 16:14:05 |
| ghpzin | ~If you still need help, it just means they OOMd while running it:
3: 0xe7d2a4 v8::internal::V8::FatalProcessOutOfMemory
It should show up better in logs before that with something like:
FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory~
Run it myself and node needs more than default memory for whatever open-webui is doing.
With env.NODE_OPTIONS = "--max-old-space-size=8192"; it seems to build. | 16:14:34 |
| ghpzin | If you still need help, it just means they OOMd while running it:
> 3: 0xe7d2a4 v8::internal::V8::FatalProcessOutOfMemory
It should show up better in logs before that with something like:
`FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory`
Run it myself and node needs more than default memory for whatever open-webui is doing.
With env.NODE_OPTIONS = "--max-old-space-size=8192"; it seems to build. | 16:14:54 |
| @drupol:matrix.org | OK | 16:22:39 |
| @drupol:matrix.org | I will let him know | 16:22:44 |
| @drupol:matrix.org | Thanks for the help :) I'll keep you posted | 16:22:53 |
| @drupol:matrix.org | ghpzin: Do you think I should submit a PR to fix this upstream? | 16:23:33 |
| ghpzin | As in nixpkgs or open-webui ?
open-webui preferably would look into why it happens, but afaik for most of these problems people just tell to add --max-old-space-size with some high number and if it works - it works. | 16:27:40 |
| ghpzin | Pretty sure in nixpkgs there are a few of those too. | 16:28:15 |
| ghpzin | Pretty sure in nixpkgs there are a few of those too, that just have --max-old-space-size with some higher number.
If you want to just temp override your flake, patch:
https://gist.github.com/ghpzin/2c953016de3f744648ad55fcb1cc7ad6
(thought it couples to frontend with version/src, because it is not exposed from nixpkgs package) | 16:29:47 |
