[10:37:02] FYI, I'll test a new cookbook to disable puppet merges now [10:37:18] ack [10:39:07] and all back to normal [13:15:14] does the `wmf-certificates` deb pkg have a repo? it's tough to find in codesearch since we use it everywhere ;) [13:20:15] https://gerrit.wikimedia.org/g/operations/debs/wmf-certificates ? [14:07:53] volans ACK, thanks for the link! [14:22:14] Another somewhat-related question. I'm working on a deb pkg for LVMd, which is a golang app. Upstream provides a binary, is there a preference between build or binary? If we go with the binary, I would be verifying the checksums [14:31:16] The context is T429310 - My recommendation would be to build from source, especially as this will be a daemon, running as root. [14:31:17] T429310: Create a Debian package for the lvmd binary - https://phabricator.wikimedia.org/T429310 [14:38:47] indeed [14:41:26] ACK, I'll try the build approach then [15:03:28] the various approaches to CI-based deb building are hopefully well documented :) [15:08:45] I know **someon** who greatly contributed to those docs ;P [15:08:58] **someone** that is [15:14:23] <_joe_> various approaches? [15:14:39] <_joe_> there's two, for different use-cases [15:15:03] <_joe_> it's not that complex I think [15:16:30] inflatador: I would `go mod vendor` the code and from there you can use https://gitlab.wikimedia.org/repos/sre/CIDERGRINDER for reference [15:17:14] <_joe_> yeah go debian packaging the debian way is deranged IMO [15:17:50] I confess that I have yet to use even one of the CI-based deb building approaches, successfully. I would be very happy to see this package built this way. [15:17:55] <_joe_> it's a typical denial of reality - we will make the logic of golang bow to our tooling instead of having tooling that works with statically linked binaries [15:17:59] key files `.gitlab-ci.yml`, `.pipeline/blubber.yaml` (quick shout-out that modern blubber is really quite nice to write in, might start using it for personal projects), and 3) `debian/rules` (if you are an official Debian Developer you are not allowed to read this file) [15:18:15] <_joe_> ahahahaha [15:19:27] <_joe_> oh come on I thought it would be much worse [15:19:53] cdanis excellent, will check it out [15:20:03] does debian care about static linking if there are actually no libraries being linked? [15:20:18] We do have an opensearch plugins package that uses CI to build and it works quite well [15:20:27] (imagine a C project that doesn't even use libc, for example) [15:20:55] blblack: are you thinking about transpiling zig to C as part of a build step [15:21:05] <_joe_> cdanis: behave [15:21:20] _joe_: oh sorry, was that WP:BEANS? [15:22:03] <_joe_> blblack: no, but the Debian Way of building go packages is to switch out the versions of libraries the go package declares for whatever library versions debian has in it, so you still statically link the result, but to different sources than the ones the author of the software declared [15:22:28] <_joe_> cdanis: kind of yes [15:25:28] ewwww [15:26:21] cdanis: no, I was just thinking of Zig-based software in general, but was leaving out the distraction of Zig itself [15:26:30] :) [15:26:34] it's not hard to avoid library deps in that world [15:26:52] is embedded Zig a thing? at either the like "ESP32-S3" scale, or smaller? [15:27:12] yes, although I don't know how mature it is. it has been a target from the early days though [15:27:14] (my "ewwww" is in response to "the Debian Way of building go packages") [15:29:14] https://ziglang.org/download/0.16.0/release-notes.html#Additional-Platforms <- the ones ending in "freestanding" here imply embedded devel without an OS, e.g. [15:29:26] thumb[eb]-freestanding [18:47:10] pcc taking a holiday? [18:48:07] andrewbogott: it seems like the nodes were not specified so it's running on all hosts [18:48:10] you should probably cancel that one [18:48:19] https://integration.wikimedia.org/ci/job/operations-puppet-catalog-compiler-puppet7-test/7046/console [18:48:28] what the heck... [18:48:33] ok, you're right, will cancel [18:48:41] yeah it happens, you are not the first (or the last one :) [18:48:56] always happens to me when I put the host name in the wrong field [18:49:00] Yep yep, I'm def guilty too [18:49:19] you can also delete it, saves a bunch of disk space [18:50:46] I guess I added a list of hostnames but never typed 'git review' [18:50:50] deleted, and thanks! [18:51:25] the right solution is that there are probably a very handful (one hand) cases in which you need to run PCC on all hosts anyway [18:51:29] it should bork right there [18:53:24] yeah, the default is bad [18:53:47] for sure, basic should probably be the default [18:54:09] the default isn't *every* host btw, but, it's one host for each definition in site.pp, which, is a lot [18:58:10] right, good point [18:58:45] someone should (do something) fix that [18:59:28] I can try but I am out next week, so perhaps I will give a try tomorrow [19:00:35] sukhe: https://gerrit.wikimedia.org/r/plugins/gitiles/operations/software/puppet-compiler/+/refs/heads/master/puppet_compiler/nodegen.py and https://gerrit.wikimedia.org/r/plugins/gitiles/operations/software/puppet-compiler/+/refs/heads/master/puppet_compiler/controller.py#116 [19:01:00] I'm out tomorrow and Monday though [19:01:05] :D [19:01:15] yeah I will flag it in case someone in Traffic wants to take it [19:01:17] but just switching the "no hosts found" behavior to basic looks simple [19:01:19] I am not even sure who owns it [19:01:24] I/F [19:01:33] (it doesn't matter to me but just want to make sure no toes are stepped) [19:02:00] I think everyone would be happier with more people doing small improvements to PCC :D [19:02:06] ++ [19:37:37] here, looking at acme-chief [22:26:08] https://neal.fun/wiki-spy/