| 14 Jul 2023 |
|  cole-h changed their display name from cole-h to cole-h (back 1 Aug). | 23:51:17 |
| 15 Jul 2023 |
 piegames | Hi, I'm currently implementing RFC 127 in https://github.com/NixOS/nixpkgs/pull/177272. There is this comment in check-meta.nix, "!!! reason strings are hardcoded into OfBorg, make sure to keep them in sync". So I'd like to coordinate this change with the OfBorg maintainers | 19:22:09 |
| 16 Jul 2023 |
| piegames | So I did some keyword search through the OfBorg code base, the only relevant thing I found is https://github.com/NixOS/ofborg/blob/released/ofborg/src/outpaths.nix#L28-L34, which does not look like it needs updating | 08:02:16 |
| piegames | On the other hand, the implementation pull request causes OfBorg internal errors (according to the label it added), so clearly there is something weird going on here | 08:02:48 |
| piegames | Where do I find the OfBorg error logs? | 08:04:56 |
 Lily Foster | In reply to @piegames:matrix.org
Where do I find the OfBorg error logs? I'm not sure where the ofborg logs are, but if it did fail in outpath caculation, you could try to run that manually with the instructions at https://github.com/NixOS/ofborg#running-meta-checks-locally (I just hope you have a system with 60+ GiB of RAM though...) | 10:55:04 |
| piegames | Unfortunately that is not an option for me then | 11:00:21 |
| Lily Foster | In reply to @piegames:matrix.org
Unfortunately that is not an option for me then So I'm running some stuff manually for you and it looks like ofborg may be failing in the stdenv check, which runs with a merged stderr+stdout to attempt to get the stdenv output path (but this PR throws a bunch of maintainless warnings to stderr) | 11:18:20 |
| piegames | Thanks. | 11:19:26 |
| Lily Foster | (I feel like ofborg should probably not be merging stderr and stdout when keep_stdout is passed to nix::Nix::run but I do not know much on why that is like that to begin with) | 11:19:59 |
| piegames | I simply forgot to set the maintainerless default to "ignore" again, but actually that's a good thing because there definitely will be warn-by-default things in the log in the future | 11:20:00 |
| Lily Foster | (or at least the stdenv check and possibly others should be run with nix::Nix::run_stderr_stdout which differentiates stderr/stdout) | 11:21:14 |
| Lily Foster | You can replicate this one particular issue with nix-env --query --available --no-name --attr-path --out-path --option extra-experimental-features no-url-literals -f . -A stdenv yourself though (no obscenely large amount of RAM required)
I am admittedly just guessing though since running ofborg one-off is still not currently supported and I'm just manually running the nix commands that it does and inspecting code
| 11:22:36 |
| Lily Foster | (cole just seems to be out for the next few weeks, so I figured I'd help a bit since I'm not sure who else would be able to get ofborg logs) | 11:23:00 |
| piegames | Does ofborg automatically remove the "internal error" label or should I do it manually? | 14:09:53 |
 Mindavi | It does remove it automatically, if it stays then the ofborg maintainer will sometimes have a look at the related logs | 14:12:25 |
 Artturin | In reply to @rick:matrix.ciphernetics.nl
It does remove it automatically, if it stays then the ofborg maintainer will sometimes have a look at the related logs It's not removed automatically | 14:34:01 |
| Artturin | Never seen it happen | 14:34:49 |
| Mindavi | Hmm | 15:02:49 |
| Mindavi | At lease Cole said he'd like it if they are kept, not sure about autoremoval otherwise | 15:03:14 |
 7c6f434c | I have assumed the label is treated as «an open OfBorg issue» with removal triggered manually | 15:06:12 |
 Sandro 🐧 | So, when do we turn failed checks red instead of grey? | 22:15:37 |
| 17 Jul 2023 |
| Artturin | Maybe we could try it on non-staging PRs and see what issues come up | 01:24:03 |
|  trofi left the room. | 06:09:53 |
| 18 Jul 2023 |
| 7c6f434c | Is there any code deployed that tracks timeout/transient failures (network, ENOSPC)/dep failures/direct failures? Maybe its output should be the last line of failure log? It would be useful on its own when build logs are numerous enough for confusion, and without such code red checks are a glaringly obvious bad idea | 08:06:00 |
| Sandro 🐧 | It would probably already be enough if the logs wouldn't be combined | 08:53:20 |
| 7c6f434c | Properly choosing best last ten lines for a lots-of-small-stuff build is hard, a summary of which kinds of failures were present is useful, and it's a clear prerequisite for red checks anyway | 08:55:00 |
| Sandro 🐧 | Last x lines are not always useful, eg pytest shows the failure summary there which is usually not enough to provide anything other than that failed | 08:56:54 |
| 7c6f434c | For a single build at least you know it is a pytest failure… | 08:58:09 |
| Mindavi | In reply to @7c6f434c:nitro.chat
Is there any code deployed that tracks timeout/transient failures (network, ENOSPC)/dep failures/direct failures? Maybe its output should be the last line of failure log? It would be useful on its own when build logs are numerous enough for confusion, and without such code red checks are a glaringly obvious bad idea I don't think so | 17:50:55 |
