| 18 Mar 2026 |
 piegames | In reply to @blokyk:matrix.org
(i wasn't sure what zulip channel would be best for this ;-;) is there anyway to re-launch the CI for a CL? i'm pretty sure the devShell and clang-tidy failures are just flaky ci, especially given the error messages and the fact that other pipelines (incl. on the same platform) were completely fine.
the contributing wiki for the ci says "just login to restart the build" but i don't have the access level for it. You can get permissions if you ask, though I forgot where to ask 🙃 | 23:09:39 |
 raitobezarius | In reply to @xokdvium:matrix.org
Can’t you force-push an amended commit without any diff? That should do the trick I think Code changes are required for that | 23:11:32 |
| raitobezarius | The CI is smart (still dumb) :p | 23:11:42 |
 Sergei Zimmerman (xokdvium) | In reply to @raitobezarius:matrix.org
Code changes are required for that Change one character and change back works? | 23:12:04 |
| Sergei Zimmerman (xokdvium) | Or does the source hash get memoised? (Just curious) | 23:12:43 |
| raitobezarius | Gerrit understands the kinds of delta a patchset | 23:14:19 |
| raitobezarius | A NO_CODE_CHANGE won't trigger CI | 23:14:35 |
| raitobezarius | Changing a character will trigger CI, yes | 23:14:47 |
| piegames | In reply to @piegames:flausch.social
You can get permissions if you ask, though I forgot where to ask 🙃 I restate my position here that retriggering CI builds should not be a gated permission and instead be accessible to all logged-in users | 23:17:53 |
 zoë (she/her) | if you can't remember i can just change the release notes slightly, that should be enough x) | 23:18:19 |
| zoë (she/her) | * if you can't remember i can just change the release notes slightly, that should be enough x)
edit: aaand it's pushed & restarted, thx folks | 23:20:15 |
| raitobezarius | In reply to @piegames:flausch.social
I restate my position here that retriggering CI builds should not be a gated permission and instead be accessible to all logged-in users I did not buy the Buildkite company yet alas | 23:20:31 |
| raitobezarius | (and I am not sure that a bot would make things more discoverable but maybe, not sure) | 23:21:50 |
 delroth | CI failure UI on the gerrit side could include instructions on how to retrigger via bot | 23:22:39 |
| delroth | (and I presume besadii could listen for gerrit CL comments, for example?) | 23:23:03 |
| raitobezarius | Yup | 23:24:24 |
| 19 Mar 2026 |
 sky1e | When I encounter a spurious CI failure, what I do is wait until someone else merges a cl and rebase my change on top and push. Not always the fastest, but doesn't involve asking anyone for help, which I'm rather bad at 🙃 | 02:00:05 |
 Winter | raitobezarius doing a bit of maintenance to the lix package set in nixpkgs, i remember mention of some weird self ref in nix-direnv. when trying to repro that, it appears that… all packages in the set are currently infreccing? e.g.:
(import ./. {
overlays = [(f: p: { inherit (p.lixPackageSets.latest) nixpkgs-review; })];
}).nixpkgs-review
| 03:29:31 |
| Winter | is this known? | 03:29:35 |
 aloisw | Yes, it is known that overlaying like this does not work. The only things it works for are Lix itself and its nix-eval-jobs. | 05:24:19 |
| Winter | i was under the (i guess false!) impression that it was only for a handful of bad derivations (e.g. nix-direnv), but i guess it’s actually never worked at all? it should probably be removed from the website then if this is indeed true | 05:35:14 |
| Winter | there was a Lix discussion at the time, but https://git.lix.systems/lix-project/lix-website/commit/26676b416a334fb69a7f6a2ac3d8464f321f4cca is what i remember | 05:37:38 |
| raitobezarius | https://git.lix.systems/lix-project/lix/issues/980#issue-14093 it's known for the past 6 months | 09:08:57 |
| raitobezarius | with no resolution yet | 09:08:59 |
| raitobezarius | we are planning to delete these instructions | 09:09:04 |
 Lisanna | is there a process for cloning a cppnix issue into the lix issue tracker? or does it happen automatically | 20:17:19 |
| raitobezarius | there's no automation of the sort as we have diverged quite | 20:24:39 |
| raitobezarius | so we usually want people to open their own issues they encountered on lix | 20:24:53 |
| raitobezarius | (mentions of cppnix sharing the same issue are welcome but are not super useful these days imho except if it's certain by the reporter there's a shared subsystem there) | 20:25:10 |
| sky1e | I noticed when hacking on lix's JSON parser that it silently accepts duplicate json object keys, throwing away all but the last one. Is this known and desired behavior? I tried making it crash instead and it seemed to still pass all of the tests, so this behavior isn't being tested for.
nix eval --expr 'builtins.fromJSON '"''"'{"a":"a","a":"b"}'"''"
{ a = "b"; }
| 23:17:47 |
